The Pennsylvania State University. The Graduate School. Intercollege Graduate Degree Program in Agricultural and Biological Engineering

Transcription

1 The Pennsylvania State University The Graduate School Intercollege Graduate Degree Program in Agricultural and Biological Engineering ECONOMIC AND PHOSPHORUS-RELATED EFFECTS OF PRECISION FEEDING AND FORAGE MANAGEMENT MODELED AT FARM AND WATERSHED SCALES: CANNONSVILLE RESERVIOR WATERSHED, NY A Thesis in Agricultural and Biological Engineering by Lula Tekle Ghebremichael 2007 Lula Tekle Ghebremichael Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy May 2007

2 The thesis of Lula Tekle Ghebremichael was reviewed and approved* by the following: ii James M. Hamlett Associate Professor of Agricultural Engineering Thesis Advisor Chair of Committee William J. Gburek Adjunct Professor of Civil Engineering C. Alan Rotz Adjunct Professor of Agricultural Engineering Tamie L. Veith Adjunct Assistant Professor of Agricultural Engineering Albert R. Jarrett Professor of Agricultural Engineering Roy E. Young Professor of Agricultural and Biological Engineering Head of the Department of Agricultural and Biological Engineering *Signatures are on file in the Graduate School

3 ABSTRACT iii Soil phosphorus (P) build-up in the New York Cannonsville Reservoir Watershed (CRW) and, in turn, the poor water quality of the reservoir, is largely due to more P being imported to farms as feed and fertilizer than exported in milk and other products. A precision feed management (PFM) program has been initiated to address the P imbalance problems while maintaining farm profitability. In this study, a model-based evaluation of PFM was performed at two management levels, farm- and watershed-level planning. For two CRW dairy-farms, Integrated Farm System Model (IFSM) simulation of reduced dietary-p integrated with increased grass-forage productivity and use in the diet resulted in reductions of farm P imbalances (78-100%) and soluble P losses (18%). Also, a decrease in feed supplement purchase was achieved, thus increasing farm profitability. Soil and Water Assessment Tool (SWAT) simulation for the same strategy also demonstrated appreciable decrease in soil P (7-8%) during the growing season, thereby indicating increased soil P removal by the improved grass-forage. Predicted soluble P loss reduction from cropland due to such a strategy was 15%, comparable to the IFSMsimulated soluble P loss reduction, which, at the watershed outlet, was 10%. For the watershed-level planning at the Town Brook Watershed (TWB), SWAT simulation of a strategy of converting 50% corn to grass integrated with reduced dietary- P and increased grass-forage productivity resulted in reductions of soluble and sedimentbound P losses from agricultural-crops (15%; 19%) and at the watershed outlet (13%; 16%). When this strategy was applied to a TBW-farm, IFSM predicted 83% reduction in the farm s P imbalance without negatively affecting farm profitability. However, converting corn to grass without strategic changes in grass-yield productivity and use was found to negatively affect the farm s profitability and P-balance due to increased P importation in grain-feed supplements. Model-based studies such as those performed in this study at farm and watershed levels provide a comprehensive tool for assessing the potential for long-term, cost-effective, and permanent reduction of P loss from dairy agriculture in the CRW. Similar approaches can be applied to dairy farms throughout the northeastern U.S. where implementation of PFM-strategies are of interest.

4 TABLE OF CONTENTS iv LIST OF FIGURES...viii LIST OF TABLES...xi LIST OF ABBREVIATIONS...xiv ACKNOWLEDGEMENTS...xv DEDICATION...xvii Chapter 1 Introduction General Introduction Overview of Previous Studies Used in This Study R-farm Watershed Town Brook Watershed Thesis Overview...13 Chapter 2 Literature Review Overview of Phosphorus Pollution Sources Control Strategies for Phosphorus Source Management Dietary phosphorus management On-farm forage management and utilization Transport Management Precision Feed Management Program Assessment Strategies for Precision Feed Management Review of Selected Simulation Models Summary of Literature Review...27 Chapter 3 Research Goal, Objectives and Hypotheses Overall Goal Specific Objectives Hypotheses...30 Chapter 4 Economic and Phosphorus-Related Impacts of Precision Feed Management Modeled at a Farm Scale Introduction Materials and Methods Integrated Farm System Model...35

5 4.2.2 IFSM Input Data Study Area and Farm Data Weather Data Alternative Farm Plan Scenario Development Farm Baseline Representations and Verifications Feed production and utilization R-farm W-farm Phosphorus balance Phosphorus loss Farm profit Model Representation of Farm Plan Scenarios Scenario Scenario Scenario 3 and Scenario Scenario Farm Plan Performance Evaluation Results and Discussion Scenario Scenario Scenario 3 and Scenario Scenario Comments on Modeling PFM Using the IFSM Model Summary and Conclusions...88 v Chapter 5 Effects of PFM Strategies on P losses Modeled at a Watershed Scale Introduction SWAT Model Description Materials and Methods Study Farm Watershed Description Description of PFM-Based Scenarios Baseline Data SWAT Representation of PFM-Based Scenarios Scenario General background information on manure production and phosphorus content Manure phosphorus content derivation for Scenario Manure phosphorus concentration input data used for Scenario Scenario Scenario SWAT Model Phosphorus Dynamics and Expected Effects of PFM Implementation...114

6 5.5 Results and Discussion Impacts of PFM on Surface Runoff and Streamflow Impacts of PFM on Sediment Losses Impacts of PFM on Phosphorus Losses and Soil Phosphorus Impacts of PFM on phosphorus loss Impacts of PFM on soil phosphorus Summary and Conclusions Chapter 6 Remarks on IFSM and SWAT Simulations in Evaluating PFM-Based Strategies Planned at Farm Level Introduction Comparison of IFSM and SWAT PFM Simulations Model Representation of a Farm and PFM Farm Plans Predictions of IFSM and SWAT from PFM-Based Strategies Predicted Phosphorus Losses from PFM-Based Strategies General Comments on Applicability of IFSM and SWAT in Evaluating PFM-Based Strategies Planned at a Farm-by-Farm Level The Role of IFSM Model in Evaluating PFM-Based Strategies Planned at Farm Level The Role of the SWAT Model in Evaluating PFM-Based Strategies Planned at Farm Level Chapter 7 Economic and Environmental Impacts of Precision Feed Management in the Town Brook Watershed, NY Introduction Materials and Methods Study Area PFM-Based Strategies and Model Representation Watershed-level PFM strategies PFM strategies for SWAT simulations Representation of watershed-scale PFM strategies for IFSM simulation Results and Discussion SWAT Modeling of PFM Effects for the Town Brook Watershed IFSM Modeling of Watershed-Level Planned PFM Strategies for a Farm Remarks on Applicability of SWAT and IFSM in Evaluating PFM-Based Strategies Planned at a Watershed Scale The Role of SWAT Model The Role of IFSM Model Summary and Conclusions vi

7 Chapter 8 Conclusions and Recommendations Conclusions Recommendations for Future Research Bibliography Appendix A SWAT-Predicted Field-by-Field Phosphorus Losses for R-farm Appendix B SWAT Model Land Use Data for the Town Brook Watershed (adopted from Gitau (2003)) vii

8 LIST OF FIGURES viii Figure 4.1: IFSM window interface showing various input data requirements...41 Figure 4.2: Study area and farm location within the Cannonsville Reservoir Watershed, NY...43 Figure 4.3: R-farm: IFSM-predicted average daily diet composition of lactating cows for winter (November to March) and for non-winter (April to October) for an example simulation year (1998) for all scenarios simulated Figure 4.4: W-farm: IFSM-predicted average daily diet composition of lactating cows for winter (November to March) and for non-winter (April to October) for an example year (1998) for all scenarios simulated...69 Figure 4.5: IFSM-predicted average forage portion (%) and total daily dry matter intake (DMI, kg) for lactating cows for winter (November to March) and for non-winter (April to October) periods of an example year (1998), for all scenarios simulated...70 Figure 4.6: Percent change of IFSM-simulated outputs for precision feed management (PFM)-based farm strategies relative to baseline scenario...73 Figure 5.1: Map showing location of study area, farm, and encompassing watershed...98 Figure 5.2: Methodology flow chart for data derivation and Scenario 1 representation in SWAT Figure 5.3: SWAT soil phosphorus pools and processes (adapted from Neitsch et al., 2002b) Figure 5.4: SWAT-simulated average annual surface runoff and total stream flow contributions from agricultural land uses for all scenarios simulated during simulation years 1993 to Figure 5.5: SWAT-simulated daily stream flows at the outlet of the study watershed for all scenarios simulated during the period of 1993 to Figure 5.6: SWAT-predicted average annual sediment loads for the study watershed (~163 ha) for all scenarios simulated during the period of Figure 5.7: SWAT-predicted average of particulate phosphorus (PP), soluble phosphorus (SolP) and total phosphorus (TP) losses for all scenarios

9 simulated from agricultural land use (grass, corn, pasture) losses for three year ( ) Figure 5.8: SWAT-predicted three year ( ) average particulate phosphorus (PP), soluble phosphorus (SolP) and total phosphorus (TP) losses for all scenarios simulated for the study watershed Figure 5.9: Watershed level: SWAT model-predicted losses of organic nitrogen (Org N) and soluble nitrogen in runoff (Sur Q N) Figure 5.10: Field-level SWAT-predicted movement of phosphorus from the labile to the active and from the active to the stable pools averaged over three years (1993 to 1994) Figure 5.11: Watershed scale SWAT-predicted movement of phosphorus from the labile to the active and from the active to the stable pools averaged over three years (1993 to 1994) Figure 5.12: SWAT predicted amounts of P in the mineral pools, active, solution (labile), and stable pools, for the 1993 simulation year for grass fields of Baseline, Scenario 1, and Scenario Figure 7.1: Map showing the location of Cannonsville Reservoir Watershed (CRW), the Town Brook Watershed (TBW), land use distribution, and crop fields for the study farm (data obtained from USDA/ARS, 2004) Figure 7.2: Map showing 66 subbasins for the Town Brook Watershed generated during baseline SWAT simulations Figure 7.3: Town Brook Watershed baseline data layers: (a) general land use (grid file), (b) corn land use (shape file), and (c) overlay of corn land uses and subbasins that had corn land use (shaded) Figure 7.4: For the TBW, diagrams showing (a) USGS digitized streams network, (b) GIS-created 30-m buffer along the stream network, (c) TBW corn land uses, and (d) corn land uses that intersect stream buffer zones Figure 7.5: For TBW subbasins that contain corn land uses that intersect stream buffers a) subbasins that had corn fields that were fully or partially within the stream buffers and b) corn land uses located within these subbasins Figure 7.6: Maps showing a) study farm crop fields and corn fields (shaded), b) selected subbasins that had corn fields intersecting 30-m stream buffer, and c) corn fields of the study farm located within the selected subbasins ix

10 Figure 7.7: SWAT simulated average annual phosphorus loses from crop fields within Town Brook Watershed for various precision feed management strategies simulated during the simulation period of 1987 to Figure 7.8: SWAT model-predicted annual average grass fields (fields used for hay production) soil phosphorus uptake over simulation period of 1987 to 1996 for the various scenarios simulated Figure 7.9: IFSM-predicted average lactating cows s feed composition: (a) daily ration mix; (b) forage to concentrate ratio; and (c) total daily dry matter intake Figure 7.10: Percent change of IFSM-simulated outputs for PFM alternative scenarios relative to the baseline scenario x

11 LIST OF TABLES xi Table 1.1: Nash-Sutcliffe coefficients of the SWAT calibration results for R-farm watershed for pre-bmp installation period, (adapted from Gitau and Gburek (2005))...10 Table 1.2: SWAT model performances for Town Brook Watershed hydrology calibrations (10/1/1997-9/30/1999) and validation (10/1/ /30/2001). Flows in m 3 /s (adopted from Gitau, 2003) Table 1.3: SWAT model performances for Town Brook Watershed sediment calibrations (10/1/1997-9/30/1999) and validation (10/1/ /30/2001). Sediment in mg/l (Adopted from Gitau, 2003)...12 Table 4.1: Summary of precision feed management (PFM) plans, assumptions, and practices implemented for each farm under each of the modeled scenarios Table 4.2: R-farm IFSM-predicted average crop yields and nutritive contents (CP and NDF) over a 25 year farm analysis Table 4.3: R-farm feed production and utilization, milk production, and phosphorus balance for a 25 year analysis of a farm with 102 cows and 40 other stock on 120 ha farm land, simulations using IFSM low-forage-diet and high-forage-diet options, and after adjustment Table 4.4: R-farm IFSM-predicted average daily ration of each animal group for winter (November to March) and for non-winter (April to October) seasons over a randomly selected simulation year...55 Table 4.5: W-farm IFSM-predicted average crop yields and nutritive contents over a 25 year farm analysis Table 4.6: W-farm feed production and utilization, milk production, and phosphorus balance for a 25-year analysis of a farm with 52 cows and 27 other stock on 65 ha of farm land...57 Table 4.7: W-farm IFSM-predicted average daily ration of each animal group for winter (November to March) and for non-winter (April to October) seasons over a randomly selected simulation year...59 Table 4.8: SWAT and IFSM model-predicted phosphorus losses from crops for the simulation period of 1993 to 1995 for the R-farm....62

12 Table 4.9: IFSM-predicted average crop yields, nutritive contents, and feed production and utilization (considering a 25 year farm analysis) under each modeled scenario for the R-farm and the W-farm...67 Table 4.10: IFSM-simulated outputs for a baseline scenario and changes in simulated outputs of alternative precision feed management (PFM) farm planning scenarios from the baseline scenario for the R-farm...71 Table 4.11: IFSM-simulated outputs of a baseline scenario and changes in simulated outputs of alternative precision feed management (PFM) farm planning scenarios from the baseline scenario for the W-farm...72 Table 5.1: List of simulated precision feed management (PFM) scenarios and their descriptions Table 5.2: Study watershed crop land uses used in SWAT for the period Table 5.3: Fresh manure production and characteristics per 1000 kg mass per day for a dairy animal (adopted from ASAE,1998a; Neitsch et al., 2002a) Table 5.4: IFSM-predicted manure production, total manure phosphorus content, and estimated amounts and concentration of mineral and organic components of manure phosphorus for baseline scenario and Scenario 1 of the study farm (R-farm) Table 5.5: SWAT model manure phosphorus concentrations for Scenario Table 5.6: SWAT-simulated annual sediment loss on field-by-field basis for R- farm watershed for 1993, 1994 and Table 5.7: SWAT-predicted average annual sediment load from crop fields for all scenarios simulated during the period of Table 5.8: Field-level SWAT-predicted particulate phosphorus (PP), soluble phosphorus (SolP) and total phosphorus (TP) losses for all scenarios simulated during the period of 1993 to Table 5.9: SWAT-predicted three year particulate phosphorus (PP), soluble phosphorus (SolP) and total phosphorus (TP) losses for baseline scenario, and percent change of predicted losses of three scenarios compared to the baseline Table 5.10: SWAT-predicted annual average phosphorus losses from crop fields (grass, corn, and pasture) for three years ( ) for all scenarios xii

13 simulated, and percent change of predicted phosphorus losses of these scenarios compared to the baseline Table 5.11: Watershed-level* SWAT-predicted three year ( ) average of particulate phosphorus (PP), soluble phosphorus (SolP) and total phosphorus (TP) losses for all scenarios simulated, and percent change of predicted phosphorus losses of these scenarios compared to the baseline Table 5.12: SWAT-simulated average phosphorus uptake of crops in the study watershed during the period of 1993 to Table 6.1: Phosphorus loss reduction of precision feed management (PFM) scenarios as determined from IFSM and SWAT model simulations over the simulation period Table 6.2: Sediment loss reduction of precision feed management (PFM) scenarios as determined from IFSM and SWAT model simulations Table 7.1: Land use distribution for the Town Brook Watershed (derived from land use shape file (USDA/ARS, 2004)) Table 7.2: Manure phosphorus concentration values used for SWAT simulations in the Baseline and Scenario 1 conditions Table 7.3: Dietary phosphorus levels for baseline and Scenario 1 of IFSM simulation for the study farm Table 7.4: Average phosphorus losses from crop fields as simulated by SWAT for various precision feed management (PFM) strategies for Town Brook Watershed for the simulation period of 1987 to Table 7.5: SWAT-predicted average annual phosphorus losses for TBW for all scenarios simulated, and percent change of predicted phosphorus losses of the PFM scenarios compared to the baseline for simulations period of 1987 to Table 7.6: IFSM-predicted average annual crop yields, nutritive contents, and feed production and utilization (considering a 25 year farm analysis period) for the baseline and alternative scenarios for the study farm in the Town Brook Watershed Table 7.7: IFSM-simulated average annual outputs (considering a 25 year farm analysis period) for baseline scenario and changes from the baseline scenario of simulated outputs of alternative PFM strategies for the study farm in the Town Brook Watershed xiii

14 LIST OF ABBREVIATIONS xiv BMP Best management practice CCE Cornell Cooperative Extension CP crude protein CREP Conservation Reserve Enhancement Program CRW Cannonsville Reservoir Watershed C-SL Corn silage DAFOSYM Dairy Forage System Model DEM Digital Elevation Model DM Dry matter DMI Dry matter intake GIS Geographical Information System HRU Hydrologic Response Unit IFSM Integrated Farm System Model K Potassium MUSLE Modified Universal Soil Loss Equation N Nitrogen NDF Neutral Detergent Fiber; NRC National Research Council NY New York NYCDEP New York City Department of Environmental Protection Org N Organic Nitrogen P Phosphorus PFM Precision Feed Management PG Purchased corn grain PP Particulate Phosphorus PS Purchased protein supplement; SL Forage silage; SolP Soluble Phosphorus Sur Q N Soluble runoff nitrogen SWAT Soil and Water Assessment Tool TBW Town Brook watershed TMDL Total Maximum Daily Load TP total phosphorus USDA United States Department of Agriculture USDA-ARS United States Department of Agriculture Agricultural Research Service USGS United States Geological Survey USLE Universal Soil Loss Equation WAP Watershed Agricultural Program

15 xv ACKNOWLEDGEMENTS I would like to extend my sincere gratitude to my advisors, Dr. James M. Hamlett and Dr. W. J. Gburek for their support, guidance, encouragement, and patience during my study, research, and writing up this Thesis. I would also like to thank my dissertation committee, Dr. C. Alan Rotz, Dr. Tamie L Veith, and Dr. Albert R. Jarrett for their advice and guidance throughout my study. I would also like to thank the USDA-ARS Pasture Systems and Watershed Research Unit (PSWMRU) at University Park, PA for helping fund this research and for providing great learning environment and the necessary resources needed to complete this research. Also, I would like to thank the Agricultural and Biological Engineering Department at Penn State for providing assistantship, which from the first beginning gave me the opportunity to pursue my PhD work. I would also thank the Watershed Agricultural Council (WAC) and Cornell Collaborative Extension of Delaware County of New York State for their financial contribution to this research. I would like to thank Dale Dewing and Paul E. Cerosaletti from Cornell Collaborative Extension of Delaware County, NY, for their advice and guidance with regard to formulating practical alternative farm strategies studied, and for providing the detailed farm data needed for this study. My gratitude also goes to the Delaware County Soil and

16 Water Conservation District (DCSWCD) and the Watershed Agricultural Council (WAC) for facilitating data collection and use. xvi To all of my family, specifically to: Eddie Romero, Mehret, Senait, Simret, and Awet thank you for your support, encouragement, and love. I must also thank Eddie again for always being there with me during the difficult moments of my life. Also, my special gratitude goes to my dear friend Dawit Semere for being consistent and generous in providing emotional support. Dawit: I like your sense of humor, and it played a great role in entertaining the stressed mind with good and loud laughter(s). Most importantly, I Thank you God for everything.

17 xvii DEDICATION This dissertation is dedicated to the memory of my late mother, Silas Ghebregziabher, who provided me with immeasurable love, care, and prayer.

18 1 Chapter 1 Introduction 1.1 General Introduction The Cannonsville Reservoir Watershed (CRW) in Delaware County, NY, is part of the New York City Water Supply watershed system. The New York City Water Supply watershed system consists of 510,000 ha (~1972 square miles) composed of a network of 19 reservoirs and three controlled lakes across parts of eight New York counties, north northwest of New York City (NYCDEP, 2005). The Hudson River, which flows from north to south in New York State, divides the New York City Water Supply system into two main water supply systems. The Croton system is located east of the Hudson River, and the Catskill/Delaware system is located west of the Hudson River. In total, these systems provide approximately 1.1 billion gallons of drinking water each day to nearly 9 million people in New York City, including commuters, visitors, and residents of neighboring counties (NYCDEP, 2005). The Catskill/Delaware watershed system is the major contributor to the New York City drinking water supply. For the year 2005, the Catskill/Delaware watershed system supplied about 98% of the total water supply to New York City (NYCDEP, 2005). Major concerns remain regarding continuing phosphorus (P) inputs to the Cannonsville Reservoir, one of the reservoirs in the Catskill/Delaware system that supplies drinking water to New York City. Historically, the reservoir has been experiencing eutrophication

19 problems caused by excess phosphorus loading. Much of the phosphorus loading to the 2 reservoir is assumed to be from agriculture within the CRW, specifically dairy agriculture (Delaware County Watershed Affairs, 2002). Research has also indicated that phosphorus impairment of the Cannonsville Reservoir has been worsened by soil phosphorus buildup in the areas within the contributing farms (Tolson and Shoemaker, 2003). Long-term land application of manure can result in the addition of large amounts of phosphorus to soils, thereby saturating the soil with phosphorus and thus increasing the risk of phosphorus runoff loss to waters (Sharpley, 1995). Key strategies to intervene in soil phosphorus build-up are to reduce phosphorus input to the soils and/or to promote increased utilization of available soil phosphorus by plants (Lanyon, 1994) Best management practices (BMPs), that are mainly structural, and management-based practices have been implemented throughout the CRW under the Watershed Agricultural Program (WAP) in an effort to reduce phosphorus losses to the water supply reservoirs. However, long-term water quality control efforts in the CRW are believed to be hindered by phosphorus build-up in the soils resulting from farm phosphorus imports exceeding exports. Farm phosphorus imbalance problems are caused when phosphorus imports in purchased feed and fertilizer exceed phosphorus exports in milk or meat or off-farm sales of farm products. For example, a study by Klausner (1993) on New York commercial dairy farms reported that on average the annual phosphorus imports exceeded phosphorus exports by

20 70 to 80%. Purchased animal feeds account for 65 to 85% of phosphorus imported 3 annually (Tylutki and Fox, 1997; Cerosaletti et al., 1998).When the phosphorus imports exceed the exports, imbalance of phosphorus on the farm occurs resulting in elevated soil P (Wang et al., 1999). Studies by Rotz et al. (2002) and Cerosaletti et al. (2004) have also reported 42 to 63% more phosphorus was imported than exported on CRW farms. These imported phosphorus excesses likely are the main contributors to increased accumulation of phosphorus in the CRW soils. Efforts to address the problem of phosphorus imbalance led to the development of a precision feed management (PFM) program by the Cornell Cooperative Extension (CCE) of Delaware County, NY. The PFM program is a whole-farm-scale BMP that directly addresses the phosphorus imbalance problem. This strategy deals with the farm phosphorus imbalance problem by managing phosphorus sources to the farm through precision feeding of nutrients to match the National Research Council (NRC, 2001) dietary recommendations and by improving on-farm forage production and utilization in the animal diet. Another part of the forage management effort is aimed at reducing erosion and associated nutrient loss problems by converting areas of land from corn to grass production. These practices combine to ultimately reduce purchased feed, phosphorus imports to a farm, and phosphorus excreted in manure; promote recycling and re-use of phosphorus; and reduce potential generation of erosion and associated phosphorus losses from corn land use.

21 Precise diet formulation with regard to phosphorus is a component of the PFM that 4 involves limiting phosphorus in animal diets by minimizing phosphorus overfeeding and manure phosphorus excretions. Dietary phosphorus levels are balanced to the NRC s recommendations for dairy cattle (NRC, 2001). Limiting phosphorus intake by animals to the NRC-recommended level helps to balance farm phosphorus inputs and outputs in a dairy operation because feed inputs are generally the major causes of phosphorus surpluses (Lanyon, 1992). Based on a study by Dou et al. (2003), most dairy farms in New York, Delaware, Maryland, and Virginia feed about 34 % more phosphorus than the cows need. Several studies have shown that dietary phosphorus can be reduced to the NRC-recommended values without sacrificing milk production, reproduction, and other physiological functions (Satter and Wu, 1999; Wu et al., 2000; Wu and Satter, 2000; Wu et al., 2001). Therefore, reducing excess dietary phosphorus is critical in controlling the farm phosphorus balance. Reducing excess dietary phosphorus levels has three important effects. First, the amount of phosphorus in manure decreases (Cerosaletti et al., 2004); therefore there is less manure phosphorus to manage (Lanyon, 1994; Powell et al., 2001). Second, the risk of phosphorus runoff from surface-applied manure is decreased when lower concentrations of phosphorus are fed (Ebeling et al., 2002). Third, farmers can benefit economically by purchasing less dietary phosphorus.

22 Improving on-farm forage production and utilization is also a component of the PFM 5 farm program. This program is designed to improve production and utilization of homegrown forage in the animal diet in an effort to reduce phosphorus feed imports to a farm. Moreover, increasing productivity of homegrown forage promotes recycling and reuse of phosphorus on the farm (Lanyon, 1992). The strategy of enhancing crop yield to increase crop phosphorus harvest has been widely practiced on dairy farms throughout Europe (Sibbesen and Runge-Metzger, 1995). Thus, reducing feed phosphorus importation as a result of increased on-farm feed utilization helps to control the farm phosphorus imbalance. Implementation of a PFM component, precision feeding, on two pilot CRW farms was reported to be successful in minimizing the magnitude of the phosphorus imbalance (Cerosaletti et al., 2004). As part of the continuing Delaware County PFM program, implementation is occurring on more farms in the CRW. To complement these encouraging results and to successfully implement the PFM efforts on more farms, a more comprehensive method is needed for the planning and evaluation of the farm PFM strategies. Such a method could help quantify the impacts of PFM efforts on milk production, farm profitability, and farm-level nutrient flows as well as assess farm system options. Currently, there are virtually no on-farm data available beyond the two CRW pilot farms with which to compare various PFM strategies. Thus, the most feasible method of analysis is through use of a whole-farm model.

23 Currently, personnel from Cornell University Cooperative Extension (CCE) of 6 Delaware County are testing the PFM farm plans on CRW pilot farms in an effort to document the economic and environmental impacts. However, more comprehensive evaluation of the PFM farm planning strategies, not just those developed for the pilot farms, is needed to quantify the impacts of PFM farm efforts on milk production, farm profitability, farm-level nutrient flows, and to assess farm system options for more farms. assessment of the environmental impacts of implementing the PFM farm program must also project results of the on-farm activity to edge-of-farm and watershed scales. This is needed so as to address the contribution of on-farm activities to the watershed outlet water quality. a more systematic quantification of environmental impacts of various PFM farm plans is important for the watershed-based water quality assessment studies, such as the total maximum daily load (TMDL) program, which are required as part of the Federal Clean Water Act. Of growing interest is also the need to establish a means to quantitatively assess the impacts of various BMPs at the watershed scale. Thus far, the results of these efforts are limited and, because executing such an evaluation on actual farms requires a great investment (both in time and resources) and is limited to few management options, such evaluations can only be made by development and application of model(s) appropriate to the questions being asked.

24 Though the PFM program was initiated and coordinated by personnel from the CCE of 7 Delaware County, NY, the program involves scientific support groups from Cornell University, and USDA-ARS of University Park, PA. The USDA-ARS is leading a research effort to evaluate environmental and economic effectiveness of the PFM program. This dissertation is an evaluation of the economical and environmental impacts of the PFM program both at the farm and watershed scales. This study used farm-and watershed-scale models to assess effectiveness of several PFM variations in controlling phosphorus losses, reducing the phosphorus imbalance problem, and maintaining farm profitability. The two models used in this study were the Soil and Water Assessment Tool (SWAT; Neitsch et al., 2002b) and the Integrated Farm System Model (IFSM; Rotz and Coiner, 2006). Both models are well recognized and widely used by USDA and other government agencies. The IFSM was employed on selected CRW dairy farms to assess economic and environmental impacts of the PFM strategies. The SWAT model was applied on a singlefarm watershed, the Town Brook Watershed, and subwatersheds of the CRW, to evaluate environmental impacts of PFM strategies. Together, these models provided a more comprehensive evaluation than was possible through the use of either model alone. The IFSM model provided economic and environmental effects at a farm scale whereas the SWAT model enabled evaluation of environmental effects of the PFM strategies at the watershed outlet and field edges.

25 8 Implementation of PFM variations were approached though the perspective of farmlevel and watershed-level planning. First, PFM variations that have the potential to reduce phosphorus imbalances and losses, while maintaining or increasing profitability of farms, were applied to two pilot CRW farms. In this approach, management changes were assumed at a farm level, and evaluation of the PFM plan effectiveness, both environmentally and economically, was done at a farm level using the IFSM model. In addition, representation of IFSM farm-level designed management changes were made in the SWAT model for a single-farm watershed that encompassed one of the CCE pilot farms. Thus, evaluations of these PFM farm-level plans were made at the outlet of the watershed. For the second approach, implementation of the PFM variations was done from the perspective of watershed-level planning. Thus, PFM variations were applied on the entire Town Brook Watershed encompassing multiple farms. Using the SWAT model the environmental impacts of these PFM variations were assessed at the outlet of the watershed. The consequences of implementation of PFM variations on individual farm income, animal feed mix, phosphorus balance, and other factors were then assessed using the IFSM whole farm model.

26 1.2 Overview of Previous Studies Used in This Study 9 SWAT representations of two Cannonsville Reservoir subwatersheds, R-farm watershed and Town Brook Watershed (TBW), from previous studies were used in this dissertation R-farm Watershed The R-farm watershed (162 ha) is located in the headwaters of the Cannonsville Reservoir Watershed (CRW), Delaware County, New York. The watershed encompasses a single 102 cow dairy farm. The farm produces hay and corn silage and imports concentrate supplements to support dairy feed. The study farm watershed has been the site of considerable research. In 1993 a sampling station was established at the outlet of the watershed, as part of a paired watershed experiment (Bishop et al., 2005) designed to evaluate the effect of best management practices (BMPs) implemented at the farm watershed. Intensive stream water monitoring data were collected from this farm watershed for all runoff events for the periods of pre-bmp installation and post-bmp installation. In 2005, Margaret Gitau and William Gburek, USDA-ARS, University Park, PA, applied the Soil and Water Assessment Tool (SWAT) for the farm watershed in an effort to evaluate the effectiveness of the BMPs (Gitau and Gburek, 2005). SWAT representation of the farm watershed was made for two periods, before and after implementation of BMPs. In Chapter 5 of this study, the SWAT representation of the farm watershed (Gitau

27 and Gburek, 2005) for the pre-bmp installation period ( ) was used as a 10 baseline condition against which to compare the impacts of management changes. For the SWAT representation of the farm watershed, 10 m DEM (Digital Elevation Model) topographic and land use data of the farm watershed were used. Weather data, both precipitation and temperature, were taken from the Delhi, NY, meteorological station, NY. Hydrologic response units (HRUs) were defined by considering individual fields as distinct units and by reproducing SWAT crop parameters for the individual fields. This entailed renaming of land uses with field-distinct names to avoid lumping of fields that have the same land use (Gitau and Gburek, 2005). As a result of such representation, detailed field-level management data including crop rotations, planting, harvesting, and manure application were represented distinctively for each field. Model calibration was performed for stream flow, sediment loss, and phosphorus loss predictions. Performance measures of the calibration results obtained from Gitau and Gburek (2005) are presented in Table 1.1. These model calibration efforts were assumed to be adequate for the study described in Chapter 5. Table 1.1 Nash-Sutcliffe (NS) coefficients of the SWAT calibration results for R-farm watershed for pre-bmp installation period, (adapted from Gitau and Gburek (2005)). Monthly Annual NS coefficients Stream flow, m 3 /s Sediment load, tonnes Dissolved phosphorus, kg Total phosphorus, kg

28 1.2.2 Town Brook Watershed 11 The Town Brook Watershed (TBW) is a subwatershed of the Cannonsville Reservoir Watershed and is part of the watershed system that supplies water to New York City. Margaret Gitau, a Ph.D student at the Pennsylvania State University Department of Agricultural and Biological Engineering, completed a dissertation in 2003 that calibrated and validated SWAT on the TBW (Gitau, 2003) for assessing effectiveness of best management practices (BMPs) for use in the watershed. The SWAT model representation provided by Gitau (2003) for TBW was used in Chapter 7 of this study. For SWAT representation, land used data of the 10-m land classification grid derived from LandSat Thematic Mapper Imagery were used. The entire TBW was divided into 66 subbasins required for SWAT model setup. For each subbasin, the formulated SWAT hydrologic response units (HRUs) represent a lumped area for the same land use and soil type combination. Management data on planting, tillage, and harvesting obtained from farm planners were used to represent each land use. Amounts of manure applied to the fields were estimated based on the number of animals within the watershed and manure production related to animal type and size. Amounts and times of manure application for each agricultural crop were based on the common practices in TBW obtained from farm planning data. Model calibration and verification were performed for stream flow, sediment loss and phosphorus loss predictions. For stream flow, simulated stream flows were compared to observed stream flow data at the outlet of the watershed. For sediment, sediment was first calibrated at the HRU / land use level, followed by an evaluation of stream sediment concentrations of sample data. For phosphorus, simulated average

29 annual DP and TP losses were compared to values in the literature. Performance 12 measures of the calibration results obtained from Gitau (2003) are presented in Tables 1.2 and 1.3. These model calibration efforts were assumed to be adequate for the activities described in Chapter 7 of this study. Table 1.2 SWAT model performance for Town Brook Watershed hydrology calibrations (10/1/1997-9/30/1999) and validation (10/1/ /30/2001). Flows in m 3 /s (adopted from Gitau, 2003). Run name Time period Statistics Measured Default Calibrated Validation Average Annual Standard deviation NS* r 2** Monthly Average Standard deviation NS r Daily Average Standard deviation NS r *NS = Nash-Sutcliffe coefficient (Martinez and Rango, 1989); **r 2 = regression coefficient of observed and simulated flows Table 1.3 SWAT model performance for Town Brook Watershed sediment calibrations (10/1/1997-9/30/1999) and validation (10/1/ /30/2001). Sediment in mg/l (adopted from Gitau, 2003). Run name Time period Statistics Measured Default Calibrated Monthly Average Standard deviation NS* r 2 ** Daily Average Standard deviation NS r *NS = Nash-Sutcliffe coefficient (Martinez and Rango, 1989); **r 2 = regression coefficient of observed and simulated sediment losses

30 Thesis Overview The contents of this dissertation are presented in various chapters, the first of which is the general introduction. The second chapter presents a comprehensive literature review. The third chapter lists the overall goal, objectives, and hypotheses of the study. Chapter four presents details of IFSM model representation and evaluation of PFM strategies on two CRW farms. Chapter five provides SWAT-model representation of IFSM-designed PFM strategies in a single-farm subwatershed of the CRW. Chapter five also presents results of the SWAT-predicted environmental impacts of the PFM strategies. Comparisons of the PFM simulations for the same farm from both IFSM and SWAT models are presented in Chapter six. Chapter seven presents assessment of the PFM variations applied to the Town Brook watershed. Also impacts of these watershed-level planned management changes for a farm located in the Town Brook watershed are discussed. General recommendations for future study are presented in Chapter eight. Finally, a complete list of references is included in the Bibliography section.

31 14 Chapter 2 Literature Review 2.1 Overview of Phosphorus Pollution Sources Phosphorus is an essential nutrient required for plant growth. However, application of phosphorus either as commercial fertilizer or manure in excess of plant needs can lead to accumulation in the soil, which eventually increases the risk of phosphorus loss in runoff (Sharpley, 1995). Phosphorus is also essential for proper bone and muscle growth, metabolism, and reproduction of animals. The main sources of phosphorus for productive livestock enterprises are feed supplements and mineral dietary phosphorus often imported to the farms. There are current concerns therefore for excess phosphorus in the soil resulting from continual land application of the phosphorus-rich manure produced from intensive livestock production (Heathwaite et al., 2000; Lanyon, 2000; Sharpley and Beegle, 2001). Most dairy farms in New York, Delaware, Maryland, and Virginia were found to feed about 34 % more phosphorus than the amount required by the cows (Dou et al., 2003). Feeding animals with an amount of phosphorus above their requirements results in enriching manure with a higher phosphorus content. Application of manure with a higher phosphorus content to cropland has the potential to cause phosphorus accumulation in the

32 soil, if the rates applied exceed crop phosphorus requirements. This, in turn, may increase the risk for runoff phosphorus loss. 15 Most of the intensive animal agriculture in the Northeast U.S. is located in grain-deficient areas because of the unavailability of sufficient farm land and unfavorable conditions for animal feed grain production. Hence, these operations require that grain and supplements for animal feed be imported, primarily from the midwestern U.S. This type of production system creates farms with a phosphorus imbalance problem resulting from exceeding import of phosphorus (as animal feed supplements and fertilizer) over phosphorus export (as crop and animal products). Thus, farms are exposed to a great potential of over applying phosphorus to the soil, mainly in the form of animal manure, thereby leading to an increased risk of soil phosphorus build-up and loss. In New York commercial dairy agriculture, for example, purchased animal feeds account for 65 to 85% of phosphorus imported to a farm annually (Tylutki and Fox, 1997; Cerosaletti et al., 1998). A study by Klausner (1993) on New York commercial dairy farms showed that annual phosphorus imports exceeded phosphorus exports by 70 to 80%. Other farm studies within Delaware County, NY, (Rotz et al., 2002; Cerosaletti et al., 2004) have shown that 42 to 63% of the imported phosphorus remains on the farms. More specifically, annual phosphorus excesses (i.e., phosphorus import minus phosphorus export) for two farms studied by Cerosaletti et al. (2004) were 12 kg/ha and 7.8 kg/ha. Rotz et al. (2002) also determined annual phosphorus excesses of 6 kg/ha and 3.6 kg/ha for two other farms in the same area. A study by Wang et al. (1999) also

33 16 indicated that farm phosphorus imbalance is an important parameter in influencing soil phosphorus status; that is, the higher the phosphorus imbalance, the greater the soil phosphorus build-up. Phosphorus from the soil moves primarily in two ways (Sharpley, 1995; Sharpley and Beegle, 2001): transport in surface runoff flow and transport in lateral flow, which joins the stream through baseflow. Phosphorus in runoff is usually transported in dissolved and sediment-bound forms, the latter normally being the greater. This is because phosphorus is readily adsorbed onto soil particles. Continuous transport of phosphorus in runoff elevates the phosphorus level in water bodies, thus accelerating eutrophication (Carpenter et al., 1998; Daniel et al., 1998; Correll, 1998). Eutrophication is an enrichment of surface waters with nutrients, primarily phosphorus (Correll, 1998), which promotes increased algal growth and decreased dissolved oxygen; as a result leading to water quality degradation affecting use of water for drinking, fisheries, recreation, etc. In general, water quality degradation due to elevated phosphorus levels transported from agricultural lands is of major concern to those involved in environmental issues. Therefore, agricultural activities, whether livestock or crop production, should set a goal that primarily attempts to maintain the phosphorus balance by controlling the inputs and outputs of phosphorus to the farm. This would eventually protect farm soils from being

34 saturated with phosphorus and thus reduce the risk of phosphorus losses to the water 17 bodies Control Strategies for Phosphorus Pollution Management strategies, collectively called best management practices (BMPs), to reduce phosphorus losses from agricultural land to water bodies focus mainly on managing the source and transport of phosphorus (Novotny and Olem, 1994; Sharpley, 1995). The source management strategies attempt to minimize accumulation of phosphorus at the soil surface. The transport management strategies are efforts that interfere with phosphorus movement from soil to water bodies. With regard to phosphorus pollution control, a number of efforts have been made to control the movement of phosphorus through runoff. Currently, efforts are being directed towards managing phosphorus at the source. Managing phosphorus at the source is thought to be the critical part of the phosphorus pollution control Source Management As mentioned earlier, source management attempts to minimize phosphorus build-up in the soil so that an optimum agronomic level is maintained. This is accomplished by controlling the amount of phosphorus in manure and fertilizers applied to the agricultural land. Some strategies, but not all, include manipulating animal dietary phosphorus; reducing the amount of phosphorus input to the farm/watershed by producing more home

35 18 grown animal feeds and hence purchasing less off-farm animal feeds; applying manure to crops based on the manure and soil phosphorus content and crop requirement; and exporting excess manure nutrients Dietary phosphorus management Reducing the amount of phosphorus in manure through manipulating animal dietary phosphorus is a powerful, cost effective approach in reducing potential phosphorus losses from dairy farms (Delaware County Watershed Affairs, 2002). The goal of dietary phosphorus management is to avoid excess feeding of phosphorus to animals, which in turn prevents enrichment of livestock manure with excess phosphorus. Intake of phosphorus by dairy cows has been shown to have a significant impact on phosphorus excretion (Satter and Wu, 1999; Ebeling et al., 2002; Dou et al., 2002; Cerosaletti et al., 2004). Phosphorus requirement guidelines for dairy cattle in the U.S. are available from the National Research Council (NRC, 2001). The NRC recommends that the typical dairy cow diet contain between 0.32 and 0.38% P, depending on milk production of the animal fed. The dietary phosphorus requirement data published in the NRC (2001) are based on existing knowledge on animal nutritional needs and are reviewed by panels of scientists. The NRC-recommended dietary phosphorus level allowances include safety factors to ensure that minimal nutrient requirements are met under any circumstances.

36 19 Despite the fact that the published phosphorus requirements include safety factors, many dairy herds are fed dietary phosphorus levels that exceed NRC recommendations (Dou et al., 2003). The common practice of overfeeding phosphorus to livestock, particularly to dairy cows, originated from the belief that high phosphorus diets improved animal reproductive performance (Dou et al., 2003). Although severe phosphorus deficiency may negatively affect reproductive performance of cattle, there is no scientific evidence that indicate benefits from feeding phosphorus to dairy cows above the NRCrecommended values. In fact, most studies show that impairments of reproductive performance in dairy cows don t begin to occur until dietary phosphorus levels fall below 0.3% of dry matter intake (Wu et al., 2000, 2001). In addition, Satter and Wu (1999) reported that the average dairy diet in the U.S. is supplemented to contain 4.8 g P/kg, while only 3.8 g P/ kg is needed for optimum milk production and reproductive efficiency. According to a survey study conducted on dairy farms in New York, Pennsylvania, Delaware, Maryland, and Virginia (Dou et al., 2003), dairy producers feed 0.44% dietary phosphorus on average. This average phosphorus feed level is still in excess by about 25% when compared to the NRC-recommended phosphorus levels (NRC, 2001). For dairy, the effect of two levels of phosphorus diets, low (3.1 g/kg) and high (4.9 g/kg), on phosphorus runoff from corn fields with no till seeding was investigated by Ebeling et al. (2002). When dairy manures from low- and high-phosphorus diets were applied at the same manure application rate, soluble phosphorus losses in the runoff generated by simulated rainfall were 40 kg/ha and 108 kg/ha for the low- and high-phosphorus diets,

37 20 respectively. This indicates a lower soluble phosphorus loss in runoff from fields treated with the manure from cows fed a lower phosphorus diet. Hence, reducing such excessive levels of dietary phosphorus intake to the NRC-recommended values can substantially reduce the amount of phosphorus in manure, which in turn, reduces the amount of phosphorus in runoff. This also benefits farmers economically, as they buy less dietary phosphorus supplements On-farm forage management and utilization Managing forage production to increase forage production and utilization in the animal diet can directly affect the amount of imported feeds required for farm production. When forage production of a farm is increased and on-farm produced forage is utilized by the farm animals, less purchased feed supplement is required to satisfy animal feed needs. Increasing productivity of forage also promotes increased utilization of soil phosphorus. Hence, increasing productivity of homegrown forage promotes recycling and re-use of phosphorus on the farm (Lanyon, 1992). For areas with higher soil phosphorus contents, increasing crop yields to increase crop phosphorus harvest has been widely practiced on dairy farms throughout Europe (Sibbesen and Runge-Metzger, 1995). Overall, increasing the availability of homegrown forage on farms would help to reduce reliance on the importation of animal feed supplements. Over time, this also prevents on-farm accumulation of excess phosphorus resulting from imported animal feeds.

38 2.2.2 Transport Management 21 Phosphorus transport management strategies represent BMPs that reduce runoff volumes and velocity and soil losses from the land surface, thus controlling the movement of phosphorus from the soil to water bodies. Runoff and erosion control practices include conservation tillage, crop rotation, crop residue management, terraces, buffer strips, filter strips, riparian buffers, cover crops, and impoundments (such as settling basins). Runoff and erosion control practices, including cover crops, crop rotation, conservation tillage, and crop residue management, are agricultural land management practices implemented to limit or contain soil movement. Management practices such as terraces and filter strips are structural devices installed or constructed to trap runoff and thereby reduce the amount of sediment and pollutant entering streams. 2.3 Precision Feed Management Program Major concerns have remained regarding continuing phosphorus (P) inputs to the Cannonsville Reservoir. Historically, the reservoir has been known for experiencing eutrophication problems caused by excess phosphorus loading. Much of the phosphorus loading to the reservoir is believed to be coming from agriculture within the Cannonsville Reservoir Watershed (CRW), specifically dairy agriculture (Delaware County Watershed Affairs, 2002). The impairment of the reservoir is also believed to be exacerbated by continuous soil phosphorus build-up, which is caused by the imbalance between farm phosphorus imports and phosphorus exports (Wang et al., 1999; Tolson and Shoemaker,

39 2003). The BMP strategies being implemented under the Watershed Agricultural 22 Program (WAP), while expected to be effective in controlling off-field phosphorus losses, do not address the ultimate cause of the phosphorus imbalance problem within many of the dairy farm operations. The ultimate cause is the phosphorus accumulation within the soil profile. Farm phosphorus imbalance problems are caused when phosphorus imports in purchased feed and fertilizer exceed phosphorus exports in milk or meat or off-farm sales of farm products (Wang et al., 1999). Efforts to address the problem of phosphorus imbalance led to the development of a precision feed management (PFM) program. The PFM involves a whole-farm-scale BMP that directly addresses the phosphorus imbalance problem. The PFM approach was developed by the Cornell Cooperative Extension (CCE) of Delaware County and other agencies. This strategy deals with the farm phosphorus imbalance problem by managing phosphorus sources at the farm level through 1) precision feeding of phosphorus with phosphorus rations matching the National Research Council recommendation, 2) improving on-farm forage production and utilization in the animal diet, and 3) converting corn land use to grass. These practices combine to ultimately reduce purchased feed and phosphorus imports to a farm, which reduces phosphorus excreted in manure; to promote recycling and re-use of phosphorus on the farm; and to reduce potential generation of erosion and associated phosphorus losses from corn land. Implementation of PFM is being developed for dairy farms in Delaware County, NY, that are located within the CRW (Cerosaletti, 2006).

40 2.4 Assessment Strategies for Precision Feed Management 23 The assessment of PFM farm strategies can be conducted using data acquired through actual farm field experimentation and/or through modeling. On-farm evaluation of a farm-level BMP such as the PFM would involve monitoring and record-keeping of many aspects of a farming system before and after PFM implementation. Aspects of a farming system that are important in this region are: 1) onfarm produced forage (the type, the amount, and the quality), 2) amount and types of feed fed to livestock, 3) supplemental feeds purchased and their nutrient constituents, 4) milk production levels, 5) the amount of manure excreted and its phosphorus content, 6) amount of phosphorus contained in milk produced and in other animal products, 7) farm products sold off-farm, 8) cost expenditures for the production system, and 9) farm net profits. Any variations in farm management would affect many economic and environmental aspects of a farming system. Hence, on-farm evaluation of any variation in PFM requires careful monitoring and record keeping of the different aspects of a farming system for a time period where variation effects are observed. Thus far, efforts by Cornell Cooperative Extension (CCE) of Delaware County to assess the PFM program include a field-based implementation and evaluation of the PFM component and a balancing of dietary phosphorus to NRC levels on two CRW dairy farms (Cerosaletti et al., 2004). The study by Cerosaletti et al. (2004) involved balancing dietary phosphorus levels of lactating cows to match the NRC requirement. Data

41 representing before-and after-implementation of this practice were monitored for months. Some of the data monitored included cow diets, forage quality, cow performance, manure nutrient composition, and milk production levels. Evaluation of the PFM program relative to its productivity, field- and farm-scale nutrient flows, and farm profitability on an actual farm require extensive work related to data collection and may also take a longer time to assess effects of the management changes in question. Assessment of environmental impacts of implementing farm-scale forage management systems must project results of the on-farm activity to edge-of-farm and watershed scales. In addition, evaluation techniques must be able to represent the variety of farm management options under time-varying climatic conditions. Further, evaluations should include effects of management alterations on losses of phosphorus, farm phosphorus balances, and farm profitability. Such comprehensive evaluations can only be efficiently made by development and application of model(s) appropriate to the questions being asked. Models have been very useful tools in planning land use management and identifying control options for pollution at farm and watershed scales. 2.5 Review of Selected Simulation Models This section briefly reviews selected farm and watershed models and compares them with regard to their suitability for representing PFM farm programs. In addition, reviews are presented relative to applicability of selected models for evaluating the effectiveness of the farm plans in controlling farm phosphorus imbalance, phosphorus soil build-up, and

42 25 phosphorus losses while maintaining farm profitability. Reviews focus on the Integrated Farm System Model (IFSM) and the Soil and Water Assessment Tool (SWAT), which are a farm-scale model and a watershed-scale model, respectively. These models are well recognized and widely used by USDA and other government agencies. IFSM (formerly the Dairy Forage System Model DAFOSYM; Rotz et al., 1989) is a comprehensive farm-scale model that simulates long-term environmental and economic benefits of various technologies (like machinery and storage) and management strategies of a farm system (Rotz and Coiner, 2006). DAFOSYM has been widely used in evaluating farming systems (Borton et al., 1997; Rotz et al., 1999a, 1999b, 1999c; 2001, 2002; Andresen et al., 2001; Soder et al., 2001; and Sanderson et al., 2001). SWAT is a simulation model that predicts long-term effects of non-point source pollution on water quality at watershed and subwatershed levels (Neitsch et al., 2002b). The SWAT model has been widely applied in the study of the impact of land management on water quantity and quality (Arnold and Allen, 1996; Arnold et al., 1999; Bingner et al., 1997; Cho et al., 1995; Peterson and Hamlett, 1998; Fitzhugh and Mackay, 2000; and Santhi et al., 2001a). The IFSM whole farm model integrates modeling routines for crop and animal (dairy or beef) production to determine long-term performance and environmental and economic impacts of a farm enterprise. IFSM is more detailed in its representation of crop growth, dairy (or beef) performance, animal feeds and their nutrient constituents, nutrient (such as

43 26 phosphorus and nitrogen) balances from imports in animal feed and fertilizer and exports in animal and crop products sold off-farm, and economics of the farm. Thus, IFSM predicts economic and environmental impacts of farm management changes involving varying feed strategies, animal handing, and crop management for a farm enterprise. A study by Rotz et al. (2002) successfully applied IFSM to evaluate both economic and environmental status of farming systems in the Northeastern U.S. With regard to phosphorus loss predictions, however, IFSM is more aggregate in its geographic representation and phosphorus transport within the farm. Furthermore, IFSM doesn t take into account the possible environmental impact difference within a farm due to varying location and field specific management into account. Also, IFSM can not be used to predict water quality at a watershed level because the model doesn t accommodate the complex processes and interactions that take place in a watershed system. On the other hand, the SWAT watershed-scale water quality model utilizes natural or geographic-based boundaries. The SWAT model predicts the impact of field specific management on water resources; however, the model doesn t include farm economics nor does it include animal performance in simulating system processes. SWAT generally focuses on representing the transport of the water, sediment, and nutrients for assessment of the impact of the land use management on the quality of water resources. Assessment of the environmental impact of farm system changes, including dietary phosphorus changes, increasing productivity of on-farm forages, and manipulation of

44 animal diets to increase use of on-farm produced forage, while maintaining farm 27 viability, requires the integrated use of both farm-scale and watershed-scale models. The phosphorus loss prediction component of IFSM is comprised of equations taken from SWAT (Neitsch et al., 2002) with an additional surface phosphorus component to represent soluble phosphorus loss directly from surface applied manures (Sedorovich et al., 2005). The surface phosphorus pool component of the IFSM allows availability of freely draining portion of unincorporated manure (in solution form) for runoff before it interacts with the top soil layer and binds to soil particles. This is thought to lower the amount of soil-bound phosphorus concentration on the top soil layer, thus less availability for sediment-bound phosphorus loss with runoff. In the SWAT model phosphorus component, surface applied manure is added directly to the soil phosphorus pools (organic or inorganic). When runoff events occur, prediction of sediment-bound phosphorus and soluble phosphorus losses are computed as a function of the respective phosphorus concentrations in these two pools in the top soil layer. 2.6 Summary of the Literature Review Phosphorus pollution from agricultural operations has been well documented. Livestock farms in the northeastern U.S. have been importing large amounts of phosphorus, relative to the amount necessary, in animal feeds to support and maintain production. Most of the farms in this area have a phosphorus imbalance problem resulting from the excessive import of phosphorus (as animal feed supplements and fertilizer) relative to phosphorus

45 28 export (as crop and animal products). The resulting phosphorus imbalance has increased soil phosphorus levels, thereby leading to an increased environmental risk for phosphorus losses in runoff and erosion. Efforts to address the problem of phosphorus imbalance led to the development of a precision feed management (PFM) program. The PFM farm program involves reduced dietary phosphorus and increased forage productivity and utilization in the livestock diets and is aimed at directly addressing the phosphorus imbalance problem. The need to more comprehensively determine the effectiveness of the PFM strategies has been discussed. The importance of integrated use of farm- and watershed-scale models in the determination of farm plan effectiveness relative to animal performance, farm viability, phosphorus imbalance, and phosphorus losses from the farm and at the watershed outlet has also been discussed.

46 29 Chapter 3 Study Goal, Objectives, and Hypotheses This study involved model-based evaluation of precision feed management (PFM) strategies designed to control phosphorus imbalances and losses while maintaining farm profitability within the farms and subwatersheds of the Cannonsville Reservoir Watershed (CRW), a part of the New York City water supply basin. 3.1 Overall Goal The overall goal of this study was to assess the effectiveness of various PFM-based farm plan strategies of controlling phosphorus imbalances and losses within the CRW at farm and watershed scales while maintaining or increasing profitability of farms. 3.2 Specific Objectives Specific objectives of this study were: To determine the applicability of the Integrated Farm System Model (IFSM) for evaluating and quantifying the economic and phosphorus-related environmental impacts of the PFM strategies for CRW dairy farms, To determine the impacts of PFM farm plan strategies on milk production, farm profitability, and farm-level nutrient flows on CRW dairy farms,

47 30 To assess farm system options for controlling the root cause of phosphorus accumulations while maintaining the profitability of the CRW farms, To represent the PFM-based farm plans in the Soil and Water Assessment Tool (SWAT) model for the purpose of evaluating these farm-level plans at the watershed level, To quantitatively assess the effectiveness of the PFM-based farm plans with regard to controlling phosphorus, both at field and watershed scales, To compare PFM effectiveness related to phosphorus losses as determined from IFSM and SWAT model simulations, and To perform comprehensive assessment of PFM strategies designed at the Town Brook Watershed scale for controlling phosphorus losses at the watershed outlet and edge of the fields while simultaneously assessing consequences of PFM implementation on individual farm profitability, animal feed availability and purchases, phosphorus balance, and other factors. 3.3 Hypotheses Implementing PFM strategies planned on a farm-by-farm basis will reduce farm phosphorus imports and additionally, a) farm profitability can be maintained or increased, b) phosphorus levels in watershed discharge can be reduced, and c) milk production of dairy cattle can be maintained or increased. By designing PFM strategies at the Town Brook Watershed,

48 a) profitability of farms within the watershed can be maintained or 31 increased, b) phosphorus imports to and exports from farms within the watershed can be balanced, and c) phosphorus levels in the watershed discharge can be reduced.

49 32 Chapter 4 Economic and Phosphorus-Related Impacts of Precision Feed Management Modeled at a Farm Scale 4.1 Introduction The Cannonsville Reservoir, part of the New York City drinking water supply system, routinely exhibits signs of eutrophication. Excess phosphorus from dairy-dominated agriculture within the contributing watershed (Cannonsville Reservoir Watershed CRW) has been estimated as the main cause of eutrophication-related water quality impairment of the reservoir. Based on a 2002 Delaware County (NY) Watershed Affairs report (Delaware County Watershed Affairs, 2002), dairy-dominated agriculture contributes approximately 70% of the annual nonpoint source total phosphorus load entering the Cannonsville Reservoir. Impairment of the reservoir threatens the quality of New York City s drinking water. In addition, the economic growth of farms within the watershed may be negatively impacted from regulatory requirements being established to improve and protect the water quality of the reservoir. Phosphorus-related impairment of the Cannonsville Reservoir is believed to be exacerbated by continuous soil phosphorus build-up (Tolson and Shoemaker, 2003), which is caused by the imbalance between farm phosphorus imports (as purchased feed and fertilizer) and phosphorus exports (as milk or meat or off-farm sales of harvested crops). A study by Wang et al. (1999) also indicated farm imbalances to be an important

50 33 process in influencing soil phosphorus status. Purchased animal feeds account for 65 to 85% of phosphorus imported annually to the watershed (Tylutki and Fox, 1997; Cerosaletti et al., 1998). In addition, a study by Klausner (1993) on New York commercial dairy farming showed that annual phosphorus imports exceeded phosphorus exports by 70 to 80%. Other farm studies within the CRW (Rotz et al., 2002; Cerosaletti et al., 2004) have shown that 42 to 63% of the imported phosphorus remains on the farms. More specifically, annual phosphorus excesses (i.e., phosphorus import minus phosphorus export) for two farms studied by Cerosaletti et al. (2004) were 12 kg/ha for a 120 ha farm and 7.8 kg/ha for a 65 ha farm. Annual phosphorus excesses for two farms modeled by Rotz et al. (2002) were 6 kg/ha for a 181 ha farm and 3.6 kg/ha for a 2025 ha farm. Currently, best management practices (BMPs) are being implemented under the Watershed Agricultural Program (Walter and Walter, 1999; Delaware County Watershed Affairs, 2002) to address phosphorus-related impairment of the Cannonsville Reservoir. Those BMPs, which are mainly structural and management-based practices, are targeted to control off-field phosphorus transport to streams and do not address long-term phosphorus imbalances from phosphorus import surpluses in animal feeds. Over time the effectiveness of such BMPs may be limited, as phosphorus build-up within the soil continues. Hence, identifying and targeting the root cause of the phosphorus imbalance at the farm level is critical to the long-term health and quality of the reservoir.

51 34 Personnel from Cornell University Cooperative Extension (CCE) of Delaware County are currently testing a farm-scale BMP that directly targets the root cause of phosphorus build-up on farms. This BMP, precision feed management (PFM), addresses farm-level phosphorus imbalance by managing phosphorus sources to the farm through three key strategies. The first, precision diet formulation and delivery, involves feeding dairy cattle with rations matching the National Research Council s recommendation (NRC, 2001). The second, on-farm forage management and utilization, is designed to improve production and utilization of homegrown forages in the animal diets. Together these two efforts reduce both purchased feed phosphorus imports to a farm and phosphorus excreted in manure (Lanyon, 1992; Satter and Wu, 1999; Ebeling et al., 2002; Dou et al., 2002). Moreover, increasing productivity of homegrown forages promotes recycling and re-use of phosphorus on the farm. The third strategy, land use management, involves conversion of land used for corn production to production of grass forages. This is an important subset of the PFM farm planning effort, which helps reduce erosion and associated nutrient losses from farm fields, particularly those used in corn silage production. Thus far, the success of the Delaware County PFM farm program in addressing farm phosphorus imbalance has been substantial. Cerosaletti et al. (2004) implemented precision diet formulation and delivery on two CRW farms. Through modification of animal diets, they decreased feed phosphorus intake by 25% without negatively affecting milk production. Also, total mass of manure phosphorus excretions decreased by 33% and mass phosphorus balance (imports minus exports) decreased by 49%. As part of the

52 35 continuing PFM program, PFM implementation is planned on more farms in the CRW. However, to complement these encouraging results and to successfully implement the PFM efforts on more farms, a more comprehensive evaluation of the PFM farm planning strategies is needed. Such an evaluation would help to quantify the impacts of PFM efforts on milk production, farm profitability, and farm-level nutrient flows and to assess farm system options that have potential to control phosphorus imbalances and losses. Currently, beyond the two CRW farms there are virtually no on-farm data available with which to compare various PFM strategies. Thus, the most feasible method of analysis is through use of a whole-farm model. The purpose of this study was to evaluate and quantify the economical and phosphorusrelated environmental impacts of the PFM strategies for a farm-scale enterprise. This study employed the Integrated Farm System Model (IFSM; Rotz and Coiner, 2006) on two selected CRW farms. Using this modeling system, several PFM farm planning strategies were evaluated and quantified with regard to their relative impacts on phosphorus balance, off-farm phosphorus loss, and farm profitability. 4.2 Materials and Methods Integrated Farm System Model The Integrated Farm System Model, IFSM (formerly the Dairy Forage System Model, DAFOSYM; Rotz and Coiner, 2006) is a comprehensive farm-scale model that simulates

53 long-term environmental impact and farm profitability for various technologies and 36 management strategies applied to a farm system. DAFOSYM has been widely used in studying farm planning strategies mainly in the Northeastern and Central U.S. and Canada (Rotz et al., 1999a; Rotz et al., 1999b; Rotz et al., 2001; Andersen et al., 2001; Soder et al., 2001; Sanderson et al., 2001; Rotz et al., 2002). The IFSM model was chosen for this study because it reports relative environmental and economic benefits of various management strategies at a farm scale and because it has been successfully used to evaluate both economic and environmental status of farming systems in the Northeast U.S. (Rotz et al., 2002). The IFSM farm simulation model integrates models of crop growth, harvest, storage, feeding, animal (dairy or beef) production, and manure handling to determine long-term performance and environmental and economic impacts of a farm enterprise. The IFSM allows simulation of up to 25 years of weather data for a farm system. Model simulations are done over each individual year and are non-continuous. That is, the model doesn t continue its simulations from year to year. When multiple years are selected for simulation, simulated outputs reported by the model represent values for individual years and average values over the simulation period. In a given year, IFSM simulation of the main processes including crop production and harvest, animal feed allocation, hydrology, and phosphorus losses are determined on a daily time step using daily weather data. Simulations of production costs and farm netreturns are determined on an annual basis for each simulated year.

54 The model is comprised of different components that help estimate farm performance, 37 profitability, and potential nutrient accumulation and loss to the environment. The following are summaries of model components pertinent to this study. A complete description of the IFSM model can be found in Rotz and Coiner (2006). The IFSM model evaluates the performance of a farm enterprise by predicting crop yield and quality; on-farm feed, milk, and manure produced; feeds sold and supplemental feeds purchased; and resources expended, such as labor, fuel, and equipment use. Feed allocation in the IFSM model is based on the nutritive value of available feeds and the nutrient requirements of different groups of animals making up the dairy herd. A linear ration optimization program is used to obtain maximum herd milk production with minimum cost rations (Rotz and Coiner, 2006). When the nutrient requirement of the animal group is greater than the sum of nutrients contained in the feeds available on the farm, the model estimates supplemental feed purchases required to satisfy animal needs and maintain milk production. For simulations where a certain level of milk production is required, the model user sets a target milk production that is maintained when the nutritive value of available feeds is sufficient to meet nutrient requirements. However, when available feeds are limited, milk production level declines in response to the limited feeds. The IFSM model allocates feeds to the dairy or beef herd based on individual animal requirements for maintenance, growth, and milk or meat production. Animal feed sources for modeling can be from on-farm produced and/or supplemental off-farm purchased

55 feeds. On-farm produced feeds may include forages (hay and silage) and grains, while 38 off-farm purchased feeds may include protein supplements, corn grain, and hay. Also, the IFSM model allocates various amounts of forage for a cow diet based on different animal groups (early-, mid-, late-lactating cows, dry cows, older heifers, and younger heifers, and the amount of forage required in the diet. For adjusting the forage level of dairy cow diets, the IFSM also provides forage options ( high-forage diet and low-forage diet ) for forage feeding (Rotz et al., 1999b). For high-forage diets, a maximum amount of forage is fed while meeting the energy and protein requirements with supplemental feeds. For the low-forage diet option, a minimum amount of forage is included in the ration while meeting a specified minimum roughage requirement for maintaining good rumen function. The economic component of IFSM uses a simple enterprise accounting of production costs and incomes to compute net-return of a farm enterprise. The production cost includes costs of crop production, harvest, storage, feeding, and other production-related activities. The farm income includes receipts from sales of milk, animals, and crops. The environmental component of IFSM predicts nutrient balances (phosphorus, nitrogen, and potassium) as well as off-farm erosion and nutrient losses. The farm phosphorus balance in the model is calculated by considering the import of phosphorus in feed and fertilizer and the export of phosphorus in milk, animals, and crops. When specified in the model, the phosphorus intake by animals is based upon animal group requirements following NRC-recommended values (NRC, 2001). Additionally, the quantity and

56 39 characteristics of phosphorus produced in the manure is calculated as a function of the quantity and phosphorus content of the feed consumed. In other words, phosphorus that is consumed but not used within the body for maintenance, growth, milk production, or reproduction will be excreted directly in manure. Within IFSM a basic hydrology approach is used, with particular fields in a farm lumped together and all crop fields generalized by a single soil and slope representation. Thus, in IFSM, there is no spatial representation of fields within a farm. To predict phosphorus loss from agricultural fields of a farm, the model uses equations that build upon those used in the Erosion-Productivity Impact Calculator (EPIC; Jones et al., 1984) and the Soil and Water Assessment Tool (SWAT; Neitsch et al., 2002). The component model used to predict off-farm phosphorus loss in the IFSM model is described by Sedorovich et al. (2005). The phosphorus prediction component of the IFSM has a surface phosphorus pool component in addition to the phosphorus pools represented in the SWAT model. The surface phosphorus pool of the IFSM allows availability of freely draining portion of unincorporated manure (in a solution form) for runoff before it interacts with the surface soil and binds to soil particles IFSM Input Data The IFSM model requires three input data files (farm, machinery, and weather input data) to represent a typical scenario. The farm data consist of detailed information that describes a farm enterprise. These are crop types and their area, generalized soil type and

57 slope, type of animal (Holstein, Jersey, and others), number of cows of different ages, 40 manure handling strategies, equipment and structures used, and prices of farm commodities produced, purchased feeds, and farm products sold off-farm. The machinery file contains data for machinery used, including parameters related to machine type, size and associated costs. Finally, the weather file consists of weather data required by the IFSM model. These data include daily values of total precipitation, maximum and minimum temperatures, and solar radiation. The IFSM model requires daily weather data for a minimum of one year (365 days). Model simulations are limited to a maximum of 25 years. A diagram of the IFSM user interface with selected input windows is presented for illustration purposes in Figure 4.1. The IFSM model input windows include: crop and soil, tillage and planting harvesting, animal feeding, machinery, economic information, and manure handling information. For, example the crop and soil information window requires data related to types of crops grown, crop area, fertilizer and manure application to crops, and dominant soil types across the farm. The tillage and planting information includes data related to types of tillage equipment, dates of tillage, and planting dates. The harvesting information consists of data related to harvest time and appropriate method of harvesting (hay, silage, high moisture and dry grain) for crops. The animal and feeding information consists of data related to animal number, type, and size; milk production level; phosphorus feeding level; forage feeding levels (high- and low-forage diet); and list of supplemental feeds for purchase. Machinery information includes number and sizes of tractors and machinery used for various aspects of farm operations

58 and the costs associated with the machinery. Economic information includes costs for 41 crop establishment, commodities, feeds, labor, and custom operations. Finally, the manure information requires data on manure handling, storage, and application methods. Figure 4.1: IFSM window interface showing various input data requirements.

59 4.2.3 Study Area and Farm Data 42 The two study farms selected for this study are located in the Cannonsville Reservoir Watershed (CRW), NY (refer to Figure 4.2). The CRW (917 km 2 ) is located in Delaware County, NY. The CRW is one of the largest watersheds serving as the source for New York City drinking water reservoirs. Elevations in the watershed range from 333 m to 1,018 m above mean sea level. The CRW climate is characterized as humid continental with an average annual temperature of about 8 0 C and precipitation of approximately 107 cm/yr (20-year average). Based on the 2001 land use data (Lounsbury, 2001), major land uses of the watershed include forest (70%), agriculture (26%), water bodies (3%), and developed land (1%). Agricultural activities consist primarily of dairy farming, and agricultural land use is typically pasture, corn, and hay crops, which are grown to support dairy farming. Dairy farms in the CRW, similar to all Northeast U.S. dairy farms, typically import large quantities of nutrients in purchased feed grain, which is necessitated due in part to insufficient land quality and climate to produce the required feed grains on the farm, as well as the positive economic response from increased milk production resulting from feeding purchased grain. Hence, on-farm nutrient accumulation, particularly phosphorus, has been a persistent problem on CRW dairy farms.

60 43 General location of study farms N Delhi weather station N New York Km # Km Cannonsville Reservoir Watershed, NY. Figure 4.2: Study area and farm location within the Cannonsville Reservoir Watershed, NY. Two CRW dairy farms, referred to as the R-farm and the W-farm, were selected for detailed analysis because detailed data needed for modeling were accessible. Also, they are part of the ongoing PFM program and are similar to other farms in the CRW area. The general location of the farms is indicated in Figure 4.2. Detailed data for both farms were gathered with the help of farm planners from the Delaware County Soil and Water Conservation District (DCSWCD) and the Watershed Agricultural Council (WAC), Walton, NY.

61 44 R-farm The R-farm consists of 120 ha of crop area on predominantly shallow loamy soils. Crops grown include corn for silage (12 ha), alfalfa (9 ha), and grass (99 ha). Grass forage was harvested from 63 ha with 36 ha used as pasture. The farm maintained holstein dairy cows with 102 mature lactating cows (683 kg average body weight), 40 heifers over one year old (517 kg average body weight), and 37 heifers under one year in age (217 kg average body weight). Cows were housed in a tie-stall barn; and heifers were housed in both tie-stall and free-stall barns. Milk yield from the farm averaged 25 kg/lactating cow/day (Dewing and Cerosaletti, 2005). During May through October, lactating cows were fed a diet of dry grass hay, corn silage, and grazed forage (mostly grass) supplemented with a corn meal and soy hulls mix and a protein and mineral mix. During November through April, lactating cows were fed a winter diet containing dry grass hay, corn silage, grass silage supplemented with corn meal and soy hulls, and a protein and mineral mix. Grass received 78% of the manure produced on the farm, and 22% of the manure was applied to corn land. In addition, fertilizers applied to corn land included 34, 17, and 17 kg/ha of nitrogen (N), phosphate (P 2 O 5 ), and potash (K 2 O), respectively. Nitrogen fertilizer was also applied to grass at a rate of 100 kg N/ha in addition to the manure applied.

62 45 W-farm The W-farm included about 65 ha of cultivated crops on predominantly shallow loamy soils. Crops grown on the farm include corn for silage (8 ha), alfalfa (16 ha), and grass (41 ha). Grass was harvested for forage from 24 ha whereas 17 ha of grass were used as pasture. The farm had holstein dairy cows with 52 mature lactating cows (637 kg average body weight), 22 heifers over one year in age (470 kg average body weight), and 27 heifers under one year in age (198 kg average body weight). Average milk yield was about 22 kg/ lactating cow/day. During the months of May through October, lactating cows were fed a diet consisting of dry grass hay, corn silage, and mostly grass pasture forage supplemented with corn meal, a citrus pulp mix, and a protein and mineral mix. During the months of November to April, lactating cows were fed a winter diet containing dry grass hay, corn silage, and grass silage supplemented with a corn meal, protein and mineral mix. Cows were housed in a tie-stall barn with heifers in tie-stall and bedded-pack barns. About 60% of the manure produced on the farm was applied to grass fields, 20% was applied to corn fields, and the remaining 20% was applied to alfalfa fields. Fertilizer applied to corn included 40, 17, and 17 kg/ha of nitrogen (N), phosphate (P 2 O 5 ), and potash (K 2 O), respectively. In addition to the manure applied, nitrogen fertilizer was also applied to grass at an average rate of 50 kg N/ha.

63 4.2.4 Weather Data 46 For the two study farms, IFSM simulation results represent average annual predictions using 25 years of historical weather data. The weather data required by IFSM, including daily values of total precipitation, maximum and minimum temperatures, and solar radiation, were obtained from the National Climate Data Center database. For both the R- farm and W-farm, precipitation and temperature data from the Delhi station, NY, were used (1978 to 2002). The Delhi station is located in the center of the CRW (Figure 4.2). Because solar radiation data were not available from the Delhi station, they were acquired from the Cooperstown station, New York, which is located approximately 45 km north of the Delhi station Alternative Farm Plan Scenario Development In this study, alternative farm planning scenarios were developed with the help of personnel from CCE of Delaware County, NY. These farm planning scenarios were based on the ongoing PFM program underway within the CRW. PFM is a farm-scale set of best management practices (BMPs) designed to directly target the root cause of phosphorus build-up on the farms and to ultimately reduce phosphorus loadings to the Cannonsville Reservoir. Because maintaining the economic viability of the farms is an important part of the whole-farm planning effort, the potential for economic benefits from implementing these scenarios were also considered. The scenarios studied were defined by farm planners to reflect realistic farm planning strategies for CRW farms.

64 47 The PFM farm planning strategies considered in developing the farm scenarios include precision diet formulation and delivery, on-farm forage management and utilization, and land use management as described in the following sections. The precision diet formulation and delivery component of the PFM involves modifying animal diets so as to minimize overfeeding of nutrients and to decrease manure nutrient excretions. Dietary nutrient levels are balanced to NRC recommendations for dairy cattle (NRC, 2001). Precision diet formulation and delivery, especially with regard to phosphorus, has been identified as a critical component of the PFM strategy to reduce phosphorus loading to the CRW (Delaware County Watershed Affairs, 2002). Therefore, it is imperative that this approach be considered in all farm planning scenarios simulated. The on-farm forage management and utilization component of the PFM involves increasing productivity of on-farm produced forages and their utilization in the animal diets. Thus, the strategy involves increasing the yield and quality of grass forage through intensive crop management and the feeding of higher forage diets. Intensive crop management involves production and harvesting strategies that increase high-quality forage yield. This strategy thus helps to decrease feed phosphorus imports and to promote re-use of soil phosphorus on the farm by increasing productivity of homegrown forages. Finally, a land use management component of the PFM requires converting crop land used in corn silage to grass-based forage with the intent to reduce sediment and associated phosphorus losses from lands typically used in the production of corn silage.

65 48 Corn silage land use in the CRW has been identified as having high potential threats for erosion and associated phosphorus losses. A modeling study of the CRW by Tolson and Shoemaker (2004) reported that 58 % of the watershed phosphorus loss was contributed by the corn production area, which comprises only 1.2 % of the total watershed area. Hence, studying effects of this strategy on reducing off-farm sediment and phosphorus losses and purchased grain imports was of interest. Farm planning scenarios modeled include a baseline scenario and alternative farm planning scenarios. The baseline scenario is an IFSM representation of the initial farming system of each study farm without a PFM strategy. The alternative scenarios were developed by considering different PFM farm planning strategies, or combinations thereof, as discussed previously. These scenarios were formulated in such a way that each successive scenario included an additional management approach to the practices considered in the preceding scenarios. Table 4.1 summarizes the PFM farm plans, assumptions, and practices implemented for each scenario applied to both the R-farm and the W-farm.

66 Table 4.1: Summary of precision feed management (PFM) plans, assumptions, and practices implemented for each farm under each of the modeled scenarios. 49 Parameter Dietary phosphorus (P) Scenarios 1-4 Dietary forage-level Grass yield Corn land converted to grass Milk produced and forage grazed R-farm PFM Scenario Assumptions PFM Scenario average P feeding rate (% Baseline 1 DMI 2 ): lactating cows = 0.50, others = 0.35 average P fed kg/cow/yr: lactating cows = 29.6, others =11.4 average P feeding rate (% DMI): lactating cows = 0.38, others = 0.32 average P fed kg/cow/yr: lactating cows = 22.2, others =10.6 Baseline Scenarios 1-5 W-farm Assumptions average P feeding rate (% DMI): lactating cows = 0.48, others =0.31 average P fed kg/cow/yr: lactating cows = 29.8, others =10.2 average P feeding rate (% DMI): lactating cows = 0.37, others = 0.28 average P fed kg/cow/yr: lactating cows = 22.3, others =8.9 Baseline and Baseline and Scenario 1 low-forage diet 3 Scenario 1 low-forage diet 3 Scenarios 2-4 high-forage diet 4 Scenarios 2-5 Baseline and Baseline and Scenario tonnes DM/ha Scenario 1 Scenarios 2-4 high-intensity yield = 8.0 Scenarios 2- tonnes DM/ha 5 5 Scenario 3 50% of corn land converted Scenario 3 to grass Scenario 4 100% of corn land Scenarios 4 converted to grass and 5 All scenarios milk production level and forage grazed kept constant All scenarios high-forage diet tonnes DM/ha high-intensity yield = 8.0 tonnes DM/ha 5 50% of corn land converted to grass 100% of corn land converted to grass milk production level and forage grazed kept constant Alfalfa and corn harvest info. All scenarios Alfalfa: 3 cuttings (silage and baled hay); corn : (corn silage ) All scenarios Alfalfa: 3 cuttings (silage and baled hay) Corn: (corn silage) Grass harvest info All scenarios Grass: 3 cuttings (silage and baled hay) Baseline and Scenario 1 Scenarios 2-5 Grass: two cuttings (silage & baled hay) Grass: 3 cuttings (silage & baled hay) 1 Baseline = initial farming system; Scenario 1 = Precision diet formulation and delivery; Scenario 2 = Scenario 1 + increased grass productivity and cows fed a high-forage diet; Scenario 3 = Scenario % corn land converted to grass; Scenario 4 = Scenario % corn land converted to grass; Scenario 5 = Scenario 4 + extra forage production for off farm sale; 2 DMI = dry matter intake; 3 low-forage diet = lowest forage composition possible without compromising proper rumen function; 4 high-forage diet = highest forage composition possible while providing sufficient energy for maintenance and production; 5 A yield of 8.0 tonnes DM/ha was taken from Cerosaletti and Dewing personal communications (2005).

67 4.2.6 Farm Baseline Representations and Verifications 50 To perform the modeling analysis, a baseline scenario for each of the CRW study farms was created in IFSM. The baseline scenario is an IFSM representation of the baseline farming system of each study farm. The baseline scenarios represented the initial economic and environmental conditions of each of the study farms and were based on data representing these CRW dairy farms as gathered from the CCE planners. To effectively assess the impacts of the proposed farm planning scenarios, considerable effort was made to ensure that the baseline scenarios modeled in the IFSM reflect initial conditions of the farms Feed production and utilization R-farm Feed production. Table 4.2 shows average crop yields and nutrient content as predicted by IFSM and actual farm data as per Dewing and Cerosalett (2005). By adjusting the IFSM yield factor, which causes an adjustment of the crop growth rate curves and resultant yields, 25 years of average predicted yield and nutrient content values were made to closely represent actual average yield data obtained from farm planners (Dewing and Cerosaletti, 2005).

68 Table 4.2 IFSM-predicted average crop yields and nutritive contents (CP and NDF) over a 25 year farm analysis for the R-farm. 51 Crops Yield, tonnes DM 1 /ha CP 2, % of DM 1 NDF 3, % of DM 1 Predicted Actual 4 Predicted Actual 4 Predicted Actual 4 ALFALFA, 9 ha (3 cuttings) GRASS 5, 99 ha (3 cuttings) CORN, 12 ha (silage) DM = dry matter; 2 CP= crude protein; 3 NDF= neutral detergent fiber; 4 typical farm data based on farm planners estimates (Dewing and Cerosaletti, 2005); 5 Grass represents total land used for grazing and harvesting hay and silage; Alfalfa mainly used for hay and silage production IFSM-predicted crop yields and nutritive contents, shown in Table 4.2, represent average values based on simulations over 25 years, with each year predicted as a separate observation. Predicted crop yields are measured in tonnes DM/ha. The nutritive contents, crude protein (CP) and neutral detergent fiber (NDF), of crops are measured as percent of dry matter (DM). Both CP and NDF factors are forage quality indicators. Feed utilization. To represent the baseline forage feeding level of the cows in IFSM, both IFSM forage level options (low-forage diet and high-forage diet) were used to evaluate the amount of feed used (on-farm produced and grain and protein supplement purchased) for the milk production level prescribed. Table 4.3 represents feed production and utilization, milk production, and phosphorus balance for the R-farm. IFSM-predicted and actual farm data as obtained from farm planners are shown.

69 52 Table 4.3: R-farm feed production and utilization, milk production, and phosphorus balance for a 25 year analysis of a farm with 102 cows and 40 other stock on 120 ha farm land, simulations using IFSM low-forage-diet and high-forage-diet options, and after adjustment. IFSM-predicted annual values 1 using Actual farm low-forage high-forage data 3 Parameter option option Adjusted 2 Forage, tonnes of DM Alfalfa and grass hay Alfalfa and grass silage Total hay and silage (alfalfa and grass) Corn silage Grazed grass forage Forage sold/purchased 149 (sold) 40 (purchased) 41 (sold) 40 (sold) Milk production, liters/cow Purchased feeds, tonnes of DM Corn grain Protein supplement Mineral phosphorus and vitamin mix NA 4 Phosphorus inputs and outputs, kg/ha 5 Phosphorus imported Phosphorus exported Phosphorus balance mean values based on 25 years of simulation, with each year as a separate observation; 2 forage feed-level was adjusted by constraining the grain and protein feed rate per cow per day; 3 based on two years actual data obtained from farm planners (Dewing and Cerosaletti, 2005); 4 NA = data not available; 5 values were estimated using the nutrient mass balance approach by Klausner et al. (1997). When the IFSM low-forage option was selected, discrepancies occurred between the values predicted by IFSM and the actual values for the amounts of forage sold and the purchased grain and protein supplements (Table 4.3). The IFSM low-forage option overpredicted the amount of forage sold and the supplemental grain purchased and underpredicted the amount of protein supplement purchased.

70 53 When the IFSM-high forage level was selected, the model predicted a need to purchase forage. This seemed to be not consistent with the actual farm data (Table 4.3). Because the two IFSM forage level options (low- and high-forage diets) resulted in data inconsistent to the actual data (Table 4.3), adjustment of the IFSM forage-feeding rate was needed to allow a better match of the purchased feed (grain and protein supplements) and forage sold to the actual data. Therefore, to come into agreement with the forage sold, grain and protein supplements purchased, and farm phosphorus balance, forage feed-level was adjusted by constraining the concentrate (grain and protein) feed rate per cow per day (in the model). The feeding limit values used for grain and protein concentrates were 9.8 kg/cow/day and 3.8 kg/cow/day, respectively, based on typical feed rate data obtained from farm planners. Limiting these feed grain and protein concentrates forced the model to utilize more available farm silage and hay. After the adjustment, values for forage sold and supplements of grain and protein purchased were more representative to the actual farm values (Table 4.3). Predicted phosphorus balance of R-farm also matched closely to the actual farm data. Thus, this simulation run was used to represent the baseline conditions. Daily cow diets. Detailed IFSM-predicted average daily rations for each animal group for a representative single year selected at random are presented in Table 4.4 for demonstration purposes. Average daily rations are presented separately for winter (November to March) and non-winter (April to October) periods. Predicted dietary amounts of daily grains and the amount of protein were reviewed by the farm planners

71 54 and deemed to be reasonable. Predicted total daily dry matter intake (DMI) for lactating cows was also comparable to the NRC-recommended values for the animal size and production level of the farm. For cows with a milk production level of 36 to 39 kg/day and animal weight of 650 to 750 kg, the recommended DMI ranged between 20.4 to 22.2 kg/cow/day (NRC, 2006). The IFSM-predicted lactating cows DMI of 19 to 23 kg/cow/day (Table 4.4) were thus very reasonable.

72 55 Table 4.4: IFSM-predicted average daily ration of each animal group for winter (November to March) and for non-winter (April to October) seasons over a single, representative simulation year for the R-farm. Cow Group Forage as a Quantity of each feed fed (kg DM/cow/day) portion of feed (%) Grazed 1 SL 2 Hay 3 C-SL 4 PG 5 PS 6 Total feed fed(dmi 7 ) kg/cow/day Winter (November to March) Dry cows Cows in early lactation Cows in mid lactation Cows in late lactation Cows in lactation average Older heifers Young heifers Heifers; average Non-winter (April to October) Dry cows Cows in early lactation Cows in mid lactation Cows in late lactation Cows in lactation average Older heifers Young heifers Heifers; average Grazed =grazed forage; 2 SL= silage from grass and alfalfa; 3 Hay = hay produced from grass and alfalfa; 4 C-SL= corn silage; 5 PG=purchased corn grain; 6 PS= purchased protein supplement; 7 DMI= dry matter intake.

73 W-farm 56 Feed production and utilization. IFSM model parameter adjustments were made to represent IFSM-simulated crop yield and production. IFSM-predicted crop yield and nutritive contents for the W-farm, shown in Table 4.5, were within the values suggested by local planners (Dewing and Cerosaletti, 2005). The predicted yield value of 8.95 tonnes DM/ha closely represented the observed corn yield for the W-farm. Based on the local planners report, the production of corn silage has historically exhibited low yields for the W-farm, mainly due to lower soil productivity. Also grass and alfalfa average annual yields of the farm were typically 5 tonnes DM/ha. Table 4.5: W-farm IFSM-predicted average crop yields and nutritive contents over a 25 year farm analysis. Crops Yield, tonnes DM 1 /ha CP 2, % of DM 1 NDF 3, % of DM 1 Predicted Actual 4 Predicted Actual 4 Predicted Actual 4 ALFALFA, 16 ha (3 cuttings) GRASS 5, 40 ha (2 cuttings) CORN, 8 ha (silage) DM = dry matter; 2 CP= crude protein; 3 NDF= neutral detergent fiber; 4 typical farm data based on farm planners estimates (Dewing and Cerosaletti, 2005); 5 Grass represents total land used for grazing and harvesting hay and silage; Alfalfa crop lands mainly used for hay and silage production Similar to the procedure used in the R-farm, predicted feed production and utilization were adjusted for the W-farm. Predicted feed production and utilization, milk production, and farm phosphorus balances are presented in Table 4.6. Though actual farm data for W- farm were not as detailed as for the R-farm, predicted data matched to the actual milk

74 57 production, amount of forage sold, and purchased protein and grain supplements. Based on the local planners report, the W-farm rarely sold forage. The long-term predicted forage sold for the farm (Table 4.6) was comparable to the actual values. Because the farm has a plan to increase milk production from the 6413 kg/cow/year to 7884 kg/cow/year, IFSM runs with the same on-farm forage production were made for the target milk production level. Hence, the baseline scenarios were based on the target milk production. Table 4.6: W-farm feed production and utilization, milk production, and phosphorus balance for a 25-year analysis for 52 cows and 27 other stock on 65 ha of farm land. IFSM-predicted annual values 1 Parameter Actual farm data 2 IFSM run a IFSM run b Forage, tonnes of DM Alfalfa and grass hay Alfalfa and grass silage Total hay and silage (alfalfa and grass) Corn silage Grazed grass forage Forage sold/purchased Milk production, liters/cow Purchased feeds, tonnes of DM Corn grain Protein supplement Mineral phosphorus and vitamin mix NA Phosphorus inputs and outputs, kg/ha 4 Phosphorus imported Phosphorus exported Phosphorus balance mean values based on 25 years weather simulation, with each year as separate observation; 2 based on two years actual data obtained from farm planners (Dewing and Cerosaletti, 2005); 3 NA = data not available; 4 values were estimated using the nutrient mass balance approach by Klausner et al. (1997); IFSM run a = represents simulations with actual farm milk production level; IFSM run b= represents simulations with target milk production level.

75 IFSM-predicted average daily diet composition of all dairy groups is presented in 58 Table 4.7. Data in Table 4.7 represent yearly average values for the different groups of lactating cows during the winter (November to March) and non-winter (April to October) seasons. Predicted daily average total DMI for lactating cows, for example, was 19 kg/cow/day for winter and non-winter periods. Based on NRC-recommended data, average DMI ranged between 18.5 to 20.1 kg/cow/day for cows with a milk production level of 32 to 35 kg/day and animal body weight of 600 to 650 kg (NRC, 2006). Average body weight of W-farm lactating cows was 637 kg; and average milk production level of W-farm cows (early- and mid-lactating) was 33.6 kg/cow/day.

76 59 Table 4.7 IFSM-predicted average daily ration of each animal group for winter (November to March) and for non-winter (April to October) seasons over a single, representative simulation year for the W-farm. Cow Group Forage as a Quantity of each feed fed (kg DM/animal/day) portion of feed Total feed fed(dmi 7 ) (%) Grazed 1 SL 2 Hay 3 C-SL 4 PG 5 PS 6 kg/cow/day Winter (November to March) Dry cows Cows in early lactation Cows in mid lactation Cows in late lactation Cows in lactation; average Older heifers Young heifers Heifers; average Non-winter (April to October) Dry cows Cows in early lactation Cows in mid lactation Cows in late lactation Cows in lactation; average Older heifers Young heifers Heifers; average Grazed =grazed forage; 2 SL=forage silage from grass and alfalfa; 3 Hay = hay produced from grass and alfalfa; 4 C-SL= corn silage; 5 PG=purchased corn grain; 6 PS= purchased protein supplement; 7 DMI= dry matter intake.

77 Phosphorus balance The term farm phosphorus balance, in the context of this chapter, represents the amount of farm phosphorus imported minus exported excluding off-farm phosphorus loss with runoff and erosion. The farm-level phosphorus balances (expressed in kg/ha) predicted by the IFSM model are based on the simulated phosphorus exports subtracted from the simulated phosphorus imports, which in turn are based on the feeds produced, feed concentrates purchased, fertilizer-containing phosphorus purchased, and export from the farm of milk, animals, and farm-produced crops sold. When these farm-related factors are represented accurately, farm phosphorus balance predictions are expected to be closely related to actual values. Predicted net phosphorus balances (phosphorus imported phosphorus exported) were 9.6 kg/ha and 10.5 kg/ha for the R-farm and the W-farm, respectively, for the baseline scenario (Table 4.3 and Table 4.6). The predicted phosphorus balance closely matched the actual phosphorus balance of each farm based on measured data via a nutrient mass balance approach by Klausner et al. (1997) Phosphorus loss IFSM phosphorus loss predictions represent the total off-farm phosphorus loss leaving the farm from all crop and grass land. Baseline IFSM-predicted phosphorus losses for R- farm and W-farm therefore represent off-farm phosphorus loss values at baseline

78 conditions. Farm-based observed data however are not available to directly verify the 61 IFSM-predicted phosphorus losses from crop fields for the R-farm and W-farm. However, IFSM-predicted phosphorus loss rates from crop fields of the R-farm (Table 4.8) were indirectly compared to average phosphorus loss rates of crop fields generated by the SWAT model in a representation of the R-farm watershed for period in a study conducted by Gitau and Gburek (2005). The phosphorus loss rates predicted by SWAT modeling for the pre-bmp installation periods (with annual Nash- Sutcliffe coefficients of 0.60 for the dissolved phosphorus and 0.62 for total phosphorus) in the Gitau and Gburek (2005) study were used in evaluating the performance of the IFSM phosphorus loss predictions. The average phosphorus loss rates of grass, alfalfa, and corn silage extracted from the SWAT outputs of the R-farm watershed are presented in Table 4.8. For comparison purposes, weighted average losses for the crops were estimated to be 0.83 kg/ha and 1.9 kg/ha for the dissolved phosphorus and total phosphorus, respectively. The IFSM-simulated three year average phosphorus loss rates from the R-farm were 0.48 kg/ha and 2.04 kg/ha for dissolved phosphorus and total phosphorus, respectively. Compared to SWAT predictions, IFSM- losses were underpredicted for dissolved phosphorus and overpredicted for sediment-bound phosphorus. The SWAT-predicted phosphorus loss data were used to indirectly verify the overall magnitudes of phosphorus losses predicted by IFSM.

79 Table 4.8: SWAT and IFSM model-predicted phosphorus losses from crops for the simulation period of 1993 to 1995 for the R-farm. SWAT-predicted phosphorus losses, kg/ha IFSM-predicted phosphorus losses, Kg/ha Crops PP 1 SolP 2 TP 3 Crops PP 1 SolP 2 TP 3 Grass/alfalfa Corn Weighted Average Lumped values for all crops PP = sediment-bound phosphorus; 2 SolP = soluble phosphorus; 3 TP = total phosphorus; 4 weighted average was calculated for simulated cropland losses by adding 3/8 of the corn losses to 5/8 of grass/alfalfa losses assuming typical rotation data of a 3 year -corn and 5 year hay/alfalfa rotation Farm profit Baseline scenario farm net-returns for the two study farms based on IFSM simulations represent average annual values estimated under long-term relative prices and costs of production. Actual farm costs and profits for individual farms are not presented in order to protect farmers financial information. Predicted costs and returns for the various scenarios and farms were reviewed by the local planners and deemed to be representative of farms of this size and type in this region Model Representation of Farm Plan Scenarios In all scenarios, the level of milk production for the farm was kept the same as in the baseline scenario, and efforts were made to keep the amount of grazed forage at the same level as in the baseline scenario.

80 Scenario 1 63 Scenario1 involved application of precision diet formulation and delivery, with an emphasis on feeding phosphorus at a prescribed rate. By changing a phosphorus feeding level parameter within IFSM, dietary nutrient levels for dairy cattle were reduced from the phosphorus levels fed under the baseline scenario to the NRC-recommended phosphorus levels (NRC, 2001) (refer to Table 4.1). As depicted in Table 4.1, the reduction in dietary phosphorus rates was substantial for the lactating cows. The dietary phosphorus level for lactating cows was reduced from 0.50 to 0.38 % for the R-farm, and from 0.48 to 0.37 % for the W-farm. The dietary phosphorus level reductions were equivalent to 7.4 kg/cow/year (from 29.6 to 22.2 kg/cow/year) for the R-farm, and 7.5 kg/cow/year (from 29.8 to 22.3 kg/cow/year) for the W-farm. For both farms, the dietary phosphorus level was reduced by 25 % compared to the baseline scenario Scenario 2 Scenario 2 had two practices added to the conditions used for Scenario 1. These added practices were: 1) increased productivity of grass fields and 2) feeding cows a high forage diet. The yield goal for high-yielding grass production was set to 8.0 tonnes DM/ha (as suggested by CCE planners). For both farms, to increase the yield and quality of grass production, application rates of nitrogen fertilizer were increased by 100 kg N/ha for the R-farm and 130 kg N /ha for the W-farm from the baseline scenario. In addition, harvest schedule for the W-farm was increased from two cuttings in the baseline to three cuttings in Scenario 2. The increased rate of nitrogen application and number of harvests were

81 64 simulated by performing iterative IFSM runs with different application rates of nitrogen fertilizer and harvesting times to achieve the prescribed yield goal. To represent a high forage diet, a high-forage feed formulation option in the model was selected to increase forage utilization. A model parameter that represents cow ability of feed consumption was adjusted to keep the total dry matter intake the same as that of the baseline scenario Scenario 3 and Scenario 4 Scenario 3 included the practices used in Scenario 2 and an accompanying conversion of corn production areas to grass production, with the intent to reduce erosion and associated phosphorus losses. Half of the corn area was converted to grass production, and half of the manure that was previously applied to corn was applied to grass. In Scenario 4 the entire corn acreage was converted to grass, and all manure that had previously been allocated to corn fields (before conversion) was applied to grass fields. Both scenarios were suggested by the CCE planners so as to determine the effects that those scenarios would have on reducing off-farm sediment-bound phosphorus losses and purchased grain imports Scenario 5 Finally, Scenario 5, applied only to the W-farm, was a modification of Scenario 4 which included increased forage production on the under-utilized grass area so as to produce additional forage for off-farm sale. The W-farm had an excess of on-farm phosphorus and

82 65 under-utilized forage area production, as reported in the Scenario 4 results. An additional scenario (Scenario 5) was used to allow off-farm sale of forage as a means of reducing farm phosphorus excess Farm Plan Performance Evaluation Model simulation results obtained by changing IFSM input parameters to represent the alternative scenarios were compared to the baseline conditions of the farms. IFSMpredicted baseline conditions were used as references in assessing impacts of alternative PFM farm planning strategies on farm profitability and farm phosphorus balance and offfarm phosphorus loss.

83 Results and Discussion Table 4.9 shows simulation results related to crop yield, nutritive contents and feed use for all scenarios applied to the study farms. Figure 4.3 and Figure 4.4 (for R-farm and W- farm, respectively) present the average daily diet composition of lactating cows for one example year (1998) for demonstration purposes. Because no grazing occurred for the months of November to March, average daily diet composition results were calculated separately for the winter (November to March) and non-winter (April to October) seasons. In addition, Figure 4.5 presents forage levels in diets and total dry matter intake of lactating cows under all scenarios simulated. Table 4.10 and Table 4.11 depict values predicted for the baseline scenario and changes from the baseline values for each alternative scenario for the R-farm and W-farm, respectively. The changes were calculated as the differences in values between the alternative and baseline scenarios such that a negative change represents a reduction, and a positive change represents an increase, in the predicted value compared to the baseline condition. Thus, the direction and magnitude of changes in economic or environmental factors resulting from implementation of an alternative scenario are shown. In addition, for each alternative scenario, the percentage change from the baseline for phosphorus balance, erosion, phosphorus losses, nitrogen leaching, and income are shown in Figure 4.6. The percentage change was determined by subtracting the value for the alternative scenario from that of the baseline and dividing by the baseline value.

84 67 Table 4.9 IFSM-predicted average crop yields, nutritive contents, and feed production and utilization (considering a 25 year farm analysis) under each modeled scenario for the R-farm and the W-farm. Crops/Feeds Baseline 1 R-farm Scenario 1 (102 mature cows) W-farm Scenario 2 Scenario 3 Scenario 4 Baseline 1 alfalfa, ha yield, tonnes DM 2 /ha CP 3, % of DM NDF 4,% of DM total grass, ha harvested grass, ha pasture, ha Yield 4, tonnes DM/ha CP 3, % of DM NDF 4, % of DM hay produced, tonnes DM silage produced, tonnes DM grazed forage, tonnes DM corn silage, ha Yield, tonnes DM/ha CP 3, % of DM NDF 4, % of DM corn silage produced, tonnes DM forage sold, tonnes DM purchased protein suppl., tonnes DM purchased corn grain, tonnes DM purchased vitamins and mineral P, tonnes Milk production, kg/cow Scenario 1 Scenario 2 (52 mature cows) Scenario 3 Scenario 4 Scenario Baseline = current farming system; Scenario 1 = Precision diet formulation and delivery; Scenario 2 = Scenario 1 + increased grass productivity and a high-forage diet; Scenario 3 = Scenario % corn land converted to grass; Scenario 4 = Scenario % corn land converted to grass; Scenario 5 = Scenario 4 + extra forage production for sale; 2 DM = dry matter basis; 3 CP = crude protein; 4 NDF= neutral detergent fiber; 4 combined values for both grazed and harvested grass forage (for pasture alone, IFSM sets CP = 26% for spring and fall, and CP =23% for spring; and NDF values for pasture start at 52% in spring, increase to 55% in the summer, and drops to 53% in the fall).

85 68 Cow feed kg/day Winter Diets Baseline Scenario1 Scenario2 Scenario3 Scenario4 Grazed SL Hay C_SL PG PS Baseline Scenario Scenario Scenario Scenario Cow feed kg/day Non-winter Diets Baseline Scenario1 Scenario2 Scenario3 Scenario4 Grazed SL Hay C_SL PG PS Baseline Scenario Scenario Scenario Scenario Baseline = current farming system; Scenario 1 = Precision diet formulation and delivery; Scenario 2 = Scenario 1 + increased grass productivity and a high-forage diet; Scenario 3 = Scenario % corn land converted to grass; Scenario 4 = Scenario % corn land converted to grass; Scenario 5 = Scenario 4 + extra forage production for sale; Grazed = Grazed forage; SL = Grass and alfalfa silage; Hay = Grass and alfalfa hay; CSL= Corn silage; PG = Purchased corn grain; PS = Purchased protein suppl.. Figure 4.3 R-farm: IFSM-predicted average daily diet composition of lactating cows for winter (November to March) and for nonwinter (April to October) for an example simulation year (1998); for all scenarios simulated.

86 69 Cow feed kg/day Winter Diets Baseline Scenario1 Scenario2 Scenario3 Scenario4 Scenario5 Grazed SL Hay C_SL PG PS Baseline Scenario Scenario Scenario Scenario Scenario Cow feed kg/day Non-winter Diets Baseline Scenario1 Scenario2 Scenario3 Scenario4 Scenario5 Grazed SL Hay C_SL PG PS Baseline Scenario Scenario Scenario Scenario Scenario Baseline = current farming system; Scenario 1 = Precision diet formulation and delivery; Scenario 2 = Scenario 1 + increased grass productivity and a high-forage diet; Scenario 3 = Scenario % corn land converted to grass; Scenario 4 = Scenario % corn land converted to grass; Scenario 5 = Scenario 4 + extra forage production for sale; Grazed = Grazed forage; SL = Grass and alfalfa silage; Hay = Grass and alfalfa hay; CSL= Corn silage; PG = Purchased corn grain; PS = Purchased protein suppl.. Figure 4.4 W-farm: IFSM-predicted average daily diet composition of lactating cows for winter (November to March) and for non-winter (April to October) for an example year (1998), for all scenarios simulated.

87 70 Forage portion of cow's diet % W-farm non-winter W-farm Winter R-farm non-winter R-farm Winter Baseline Scenario1 Scenario2 Scenario3 Scenario4 Scenario5 23 Total Dry matter intake, DMI, Kg/cow/day W-farm non-winter W-farm Winter R-farm non-winter R-farm Winter Baseline Scenario1 Scenario2 Scenario3 Scenario4 Scenario5 Baseline = current farming system; Scenario 1 = Precision diet formulation and delivery; Scenario 2 = Scenario 1 + increased grass productivity and a high-forage diet; Scenario 3 = Scenario % corn land converted to grass; Scenario 4 = Scenario % corn land converted to grass; Scenario 5 = Scenario 4 + extra forage production for sale. Figure 4.5: IFSM-predicted average forage portion (%) and total daily dry matter intake (DMI, kg) for lactating cows for winter (November to March) and for non-winter (April to October) periods of an example year (1998), for all scenarios simulated

88 71 Table 4.10: IFSM-simulated outputs for baseline scenario and changes in simulated outputs from the baseline scenario for the R-farm for alternative precision feed management (PFM) farm planning scenarios. IFSM model output Baseline 2 Change in value 1 as compared to the baseline scenario Scenario Scenario Scenario Scenario Hay and silage produced, tonnes DM Corn silage produced, tonnes DM Grazed forage consumed, tonnes DM Forage sold, tonnes DM Total concentrate purchased 3, tonnes DM Milk produced, kg/cow/day P imported, kg/ha P exported, kg/ha P balance, kg/ha Manure produced, tonnes DM P in manure, kg Erosion sediment loss, kg/ha Soluble P loss, kg /ha Sediment-bound P loss, kg/ha Nitrogen imported, kg/ha Nitrogen exported, kg/ha Nitrogen leached, kg/ha Nitrogen concentration in leachate, ppm Cost and return expressed per mature cow, $/cow Milk and animal income Total production cost Machinery cost Fuel, electric and labor cost Storage facilities cost Seed, fertilizer, and chemical cost Land rental and property tax Purchased feed and bedding cost Animal facilities and other expenses Farm net return Standard deviation in net return change in value = alternative scenario value baseline scenario value; 2 Baseline = current farming system; Scenario 1 = Precision diet formulation and delivery; Scenario 2 = Scenario 1 + increased grass productivity and a high-forage diet; Scenario 3 = Scenario % corn land converted to grass; Scenario 4 = Scenario % corn land converted to grass; 3 total concentrate includes grain, protein, vitamins, and mineral phosphorus; P = phosphorus; DM = dry matter basis.

89 Table 4.11: IFSM-simulated outputs of baseline scenario and changes in simulated outputs from the baseline scenario for the W-farm for alternative precision feed management (PFM) farm planning scenarios. IFSM model output Baseline 2 72 Change in value 1 as compared to the baseline scenario Scenario 1 Scenario 2 Scenario 3 Scenario 4 Scenario 5 Hay and silage produced, tonnes DM Corn silage produced, tonnes DM Grazed forage consumed, tones DM Forage sold, tonnes DM Total concentrate purchased 3, tonnes DM Milk produced, kg/cow/day P imported, kg/ha P exported, kg/ha P balance, kg/ha Manure produced, tonnes DM P in manure, kg Erosion sediment loss, kg/ha Soluble P loss, kg /ha Sediment-bound P loss, kg/ha Nitrogen imported, kg/ha Nitrogen exported, kg/ha Nitrogen leached, kg/ha Cost and return expressed per mature cow, $/cow Milk and animal income Total production cost Machinery cost Fuel, electric and labor cost Storage facilities cost Seed, fertilizer, and chemical cost Land rental and property tax Purchased feed and bedding cost Animal facilities and other expenses Farm net return Standard deviation in net return change in value = alternative scenario value baseline scenario value; 2 Baseline = current farming system; Scenario 1 = Precision diet formulation and delivery; Scenario 2 = Scenario 1 + increased grass productivity and a high-forage diet; Scenario 3 = Scenario % corn land converted to grass; Scenario 4 = Scenario % corn land converted to grass; Scenario 5 = Scenario 4 + extra forage production for sale; 3 total concentrate includes grain, protein, vitamins, and mineral phosphorus; P = phosphorus; DM = dry matter basis.

90 R-farm Percent (%) Erosion Soluble P Sedimentattached P P balance N leaching Income Scenario Scenario scenario Scenario Percent (%) W-farm Erosion Soluble P Sedimentattached P P balance N leaching Income Scenario Scenario Scenario Scenario Scenario Baseline = current farming system; Scenario 1 = Precision diet formulation and delivery; Scenario 2 = Scenario 1 + increased grass productivity and a high-forage diet; Scenario 3 = Scenario % corn land converted to grass; Scenario 4 = Scenario % corn land converted to grass; Scenario 5 = Scenario 4 + extra forage production for sale. Figure 4.6: Percent change of IFSM-simulated outputs for precision feed management (PFM)-based farm strategies relative to baseline scenario.

91 4.3.1 Scenario 1 74 Complete results of the IFSM simulation for the precision diet formulation and delivery strategy for both farms are presented under Scenario 1 in Table 4.9, Table 4.10, and Table 4.11 and in Figure 4.3, Figure 4.4, Figure 4.5, and Figure 4.6. As shown in Table 4.9, all data related to crop production and feed use, with the exception of mineral phosphorus purchased, remained the same in Scenario1 as in the baseline scenario. Reducing the dietary phosphorus rations to NRC (2001) recommendations resulted in a decrease in the amount of mineral phosphorus supplements purchased. The amount of mineral phosphorus supplements purchased was reduced by 3.7 tonnes/yr and 1.7 tonnes/yr for the R-farm and the W-farm, respectively. With regard to dietary phosphorus particularly, the reduction for both farms was equivalent to 7.5 kg/cow/year of dietary mineral phosphorus level. For both farms, the dietary phosphorus level reduction was 25% compared to the baseline scenario (calculated from Table 4.1). Investigation of detailed ration compositions of cows was carried out. Based on predicted ration data (presented in Figure 4.3 and Figure 4.4), predicted average daily diet composition (excluding dietary mineral phosphorus) in Scenario 1 was the same as that of the baseline scenario. Similarly, the forage level and total dry matter intake of Scenario 1 was identical to that of the baseline scenario (see Figure 4.5).

92 75 The predicted phosphorus balance was reduced by 5.8 kg/ha and 4.8 kg/ha for the R-farm and the W-farm, respectively (see Table 4.10 and Table 4.11). These reductions were equivalent to 60% and 46%, for R- and W-farms, respectively, as compared to the baseline conditions (Figure 4.6). The 60% reduction in phosphorus balance for the R- farm is consistent with the reduction of 60% the farm planners observed for the actual farm (Cerosaletti, 2006). In a previous study, Cerosaletti et al. (2004) obtained a 49% reduction in mass phosphorus balance in a field study for two CRW farms in which feed phosphorus intake was reduced by 25%. When cows were fed a reduced-phosphorus diet, the amount of phosphorus in excreted manure was reduced. Manure phosphorus was reduced from the baseline scenario by 490 kg/year on the R-farm, and 207 kg/year for the W-farm (Table 4.10 and Table 4.11). These reductions correspond to 4.8 and 4.0 kg manure phosphorus /mature cow/year for the R-farm and W-farm, respectively. Again, compared to the baseline case, reductions in manure phosphorus were 25% (490 kg of the1961 kg depicted in Table 4.10) for the R- farm and 21% (207 kg of the 977 kg depicted in Table 4.11) for the W-farm. These results are slightly less than the 33% reduction in manure phosphorus concentration reported by Cerosaletti et al. (2004) for a dietary phosphorus reduction of 25%. Because the field-applied manure contained a lower concentration of phosphorus than that of the baseline farm scenario, off-farm phosphorus loss (mainly soluble phosphorus loss with runoff) was reduced for this scenario. The off-farm soluble phosphorus loss reductions predicted by IFSM were 0.07 kg/ha for R-farm and 0.06 kg/ha for W-farm

93 (Table 4.10 and Table 4.11). These represent 13% and 12% (for R-farm and W-farm, 76 respectively) reductions in off-farm soluble phosphorus losses due to reduced dietary phosphorus level (Figure 4.6). Similar results were shown by Ebeling et al. (2002) when dairy manures from two dietary phosphorus levels (dietary phosphorus of 0.49% and 0.31% of dry matter intake) were surface applied. In that case the soluble phosphorus loss with runoff was lower in the manure from cows fed a lower phosphorus diet. Predicted losses of sediment-bound phosphorus from the R-and W-farm were reduced by 0.03 and 0.02 kg/ha, respectively, compared the baseline conditions (Table 4.10 and Table 4.11). The reduction in desired feed phosphorus content and the subsequent decrease in purchased mineral phosphorus supplements also impacted both farms financially. The annual farm net-return (based on long-term averages) predicted by IFSM for Scenario 1 increased by $20/cow for the R-farm and $17/cow for the W-farm (Table 4.10 and Table 4.11) as compared to the baseline scenario. The result was equivalent to a farmlevel net-return increase of $2040 for R-farm and $884 for W-farm. Based on a $550/tonne price for vitamins and mineral phosphorus supplement, these farm net-returns represent the money saved as a result of a reduction of 3.7 tonnes for R-farm and 1.7 tonnes for W-farm in mineral phosphorus supplements purchased.

94 4.3.2 Scenario 2 77 Scenario 2 added two practices to Scenario 1. Productivity of grass fields was increased through improved management, and cows were fed with a high-forage diet. For both farms, improvement of grass quality is indicated by the increased values for crude protein (CP; 1.2% and 0.8%) and the decreased values for neutral detergent fiber (NDF; 1.0% and 1.7%) for grass (Table 4.9). On both farms, application of Scenario 2 increased the amount and quality of homegrown forage and decreased the amount of imported feed protein and grain supplements, which affected the farm phosphorus balance. In the R-farm baseline scenario, 47 ha of low-intensity grass with an annual yield of 6.0 tonnes DM/ha produced the amount of forage required to feed all cows a relatively lowforage diet with an annual sale of 41 tonnes DM/ha (Table 4.9). In Scenario 2, IFSM simulated the production of a high-yielding grass (8.0 tonnes DM/ha annually), and the grass quality was improved through more intensive management. Due to the increase in the amount of homegrown forage fed, a grass area of 61 ha was needed to produce the forage required to feed the cows a high-forage diet and to provide the 39 tonnes DM of forage sold. This closely matched the amount of forage sold in the baseline scenario (Table 4.9). Similarly, for the W-farm, the baseline grass yield of 5.1 tonnes DM/ha was increased to 8.0 tonnes DM/ha and the quality was improved by intensive management. As a result, 21 ha were required to produce the forage required to feed the cows a highforage diet with no extra forage sold as compared to the 18 ha in the baseline scenario. In consultation with CCE planners, production of grass silage as the primary harvest and

95 78 storage method was increased from 198 tonnes DM in the baseline scenario to 331 tonnes DM for Scenario 2 for the R-farm, and from 103 tonnes DM in the baseline scenario to 156 tonnes DM for Scenario 2 of the W-farm. In the daily ration, the amount of silage was increased whereas the amount of protein supplement was decreased in Scenario 2 compared to the baseline scenario for both farms (Figure 4.3 and Figure 4.4), both in winter and non-winter diets. This was also associated with increased levels of forage in the diet for the equivalent total dry matter intake of the cows (Figure 4.5). For lactating cows, average forage to concentrate ratio for the R-farm was increased from 52:48 in the baseline scenario to 65:35 in Scenario 2 (Figure 4.5). For W-farm, average forage to concentrate ratio was increased from 48:52 in the baseline scenario to 67:33 in Scenario 2. Increasing the forage productivity and the proportion of forage in the diet reduced the need for purchased feed, mainly in the form of protein supplements. As a result, the annual imported feed in protein concentrates purchased declined by 106 tonnes for the R- farm and 70 tonnes for the W-farm (Table 4.9). Consequently, the phosphorus imbalances were reduced compared to the baseline by 9.6 kg/ha for the R-farm and 8.2 kg/ha for the W-farm (Table 4.10 and Table 4.11). Compared to the reductions achieved by Scenario 1, Scenario 2 reduced the phosphorus imbalance an additional 3.8 kg/ha for the R-farm, and 3.4 kg/ha for the W-farm. Scenario 2 incrementally reduced the overall phosphorus balances of the farms by 40 and 32% over Scenario 1 for the R and W farms, respectively (Figure 4.6).

96 79 More accurately modeling feeding of phosphorus in the diet in addition to increasing the productivity of grass forages and increasing the proportion of forage in the diet demonstrated a great potential toward balancing phosphorus inputs and outputs of a farm enterprise. The set of conditions assumed in this study including forage production, forage quality, forage feeding level, and feed mixes suggests that achieving a zero phosphorus balance is possible, as indicated in the simulation result of Scenario 2 of the R-farm. Predicted reductions of sediment-bound and soluble phosphorus losses off-farm for Scenario 2 were minimal (5% on average) compared to Scenario 1. Predicted N loss in leachate increased across the two farms for scenarios in which additional N fertilizer were required to increase forage productivity. For Scenario 2, predicted N loss in leachate increased by 5 kg/ha for the R-farm and 7 kg/ha for the W-farm compared to the baseline (refer to Table 4.10 and Table 4.11). Also, predicted N leaching concentration increased by 1.9 mg/l for the R-farm and 1.5 mg/l for the W-farm compared to the baseline (refer to Table 4.10 and Table 4.11). Predicted N concentrations in leachate for this scenario were equivalent to the 10 mg/l maximum contaminant level for drinking water. These N concentration values also represent offfield losses, not losses at the streams. Therefore, the magnitudes of these losses were not considered to pose a major concern to the environment. However, the need of better management practices to better match N availability to crop needs in order to control N- leaching losses and increase efficiency of N use for all forage production levels is recognized.

97 When costs of N fertilizer and other operations required to increase grass forage 80 productivity are lower than costs of feed supplements, the profit of the farm is increased. Based on the simulated results related to cost and net-returns, expenses of purchasing fertilizer, fuel, machinery, labor and others, were greater in Scenario 2 than in the baseline scenario (+$108/cow for R-farm, and +$110/cow for the W-farm; Table 4.10 and Table 4.11). However, purchased feed and related costs for Scenario 2 were much smaller than these expenses in the baseline scenario (-$345/cow for R-farm, and - $287/cow for W-farm; Table 4.10 and Table 4.11). These costs, therefore, contributed to an increase in annual farm net-returns of $237/cow and $177/cow in Scenario 2 for the R and W farms, respectively, compared to the baseline scenario Scenario 3 and Scenario 4 Scenario 3 and Scenario 4 combine a land use management activity of converting corn land to grass land along with the Scenario 2 practices of enhanced forage and dietary phosphorus management. With an objective of reducing erosion and associated phosphorus losses, in Scenario 3 50% of the corn area on a farm was converted to grass. In Scenario 4, all corn land was converted to grass. As expected, by converting corn fields to grass, erosion and associated phosphorus losses were reduced. Because land use conversion from corn to grass altered the amount of forage produced on the farms, the required area of high-yield grass for each farm increased compared to Scenario 2. Also, IFSM predicted that more supplemental concentrates, mainly corn grain, must be purchased to offset the reduction in corn silage produced.

98 81 For example, in Scenario 3 for the R-farm 67 ha of high-yield grass (8.1 tonnes DM/ha) as compared to 61 ha for the Scenario 2 was required to produce the amount of forage required to feed the cows with high-forage diets and still produce an equivalent amount of forage for off farm sale (41 tonnes DM) to that in Scenario 2. Under Scenario 4, with all corn fields converted to grass, an area of 72 ha of high-yield grass (8 tonnes DM/ha) was required. This was equivalent to 97% utilization of the entire grass area (63 ha) plus the 12 ha of corn converted to grass. Moreover, with regard to feeds, the R-farm must purchase 18 tonnes DM and 23 tonnes DM more supplemental grain, under Scenario 3 and Scenario 4, respectively, compared to Scenario 2 (Table 4.9). For the W-farm, under Scenario 3, 25 ha of high yield harvested grass (8 tonnes DM/ha) was required to produce the amount of forage required for the high forage diets. This assumed that, as in the baseline, no extra forage was produced for sale. The resulting grass area required utilization of 86% of the entire grass area for high intensity management. Under Scenario 4 with all corn fields converted to grass, an area of 28 ha of high yield grass (8 tonnes DM/ha) was required. This was equivalent to 88% utilization of the entire grass area. Under Scenario 3 and Scenario 4, IFSM predicted that the W- farm needed to purchase 8 tonnes DM and 13 tonnes DM, respectively, more corn grain supplement to offset the reduction in feed energy available in corn silage, as compared to Scenario 2 conditions (Table 4.9). Predicted daily cow rations also showed a decrease in the use of corn silage and an increase in use of purchased grain supplements (Figure 4.3 and Figure 4.4). Predicted

99 forage portions of diets and total dry matter intake of cows was comparable to those of Scenario 2, for both farms (Figure 4.5 ) 82 By converting corn acreage to grass, reductions of erosion and associated phosphorus losses was achieved. However, there was no appreciable change in the farm phosphorus balance due to this land use conversion. For instance, compared to the baseline scenario, for Scenario 3 the reduction in farm phosphorus balance for the R-farm was 9.6 kg/ha (compared to a 9.6 kg/ha reduction for Scenario 2). Similarly, the reduction in farm phosphorus balances for the W-farm was 8.6 kg/ha (compared to a 8.2 kg/ha reduction for Scenario 2). For Scenario 4, the reductions in farm phosphorus balance compared to the baseline were 9.8 kg/ha and 8.8 kg/ha for the R-farm and W-farm, respectively. The farm phosphorus balances for both farms under both scenarios were minimally impacted by converting land from corn to grass, as compared to implementation of a high-forage diet and precision diet formulation and delivery, which resulted in 78% and 100% decreases in farm phosphorus balances (Scenario 2). This was due to the increase in feed purchases required to offset energy lost in the diet from reduced feeding of corn silage. A study by Rotz et al. (2002) for two CRW dairy farms also reported a minimal effect of cropping changes on the farm phosphorus balance. The reduction of erosion and associated phosphorus losses resulting from switching corn fields to grass was directly proportional to the area of corn converted to grass. For example, under Scenario 3 for the R-farm, losses of sediment, soluble phosphorus, and sediment-bound phosphorus in runoff were reduced by 38, 18, and 24%, respectively,

100 83 compared to the baseline scenario (refer to Figure 4.6 for the R-farm). For the W-farm, losses of sediment via erosion, soluble phosphorus in runoff, and sediment phosphorus in runoff were reduced by 29, 22, and 26%, respectively, compared to the baseline scenario (Figure 4.6 for the W-farm). Under Scenario 4, sediment, soluble phosphorus, and sediment phosphorus were reduced by 55, 23, and 44%, respectively, for the R-farm compared to the baseline scenario. For the W-farm, the reductions were 45, 25, and 38%, for sediment, soluble phosphorus, and sediment phosphorus, respectively, compared to the baseline scenario. For the R-farm, the model predicted that converting each hectare of corn to grass reduced the average sediment-bound phosphorus losses by 9.3 kg/ha each year. For the W-farm similar reductions were predicted to be 6.1 kg/ha sediment-bound phosphorus. A study by Gitau (2003) using the SWAT model predicted 6.5 kg/ha particulate phosphorus losses for corn land use in Town Brook Watershed, NY. Generally, these reductions may be very important relative to water quality, particularly if the location of the converted corn fields is in close proximity to streams. There were no straightforward findings as to how reductions in corn acreages affected the financial conditions of the farms. This was because 1) corn land use conversion to grass was applied to the already re-designed farms, having increased forage production and utilization, and not to the baseline conditions and 2) changes occurred in various aspects of the farm production system, such as costs of buying feeds and fertilizer and crop production operating costs.

101 84 Operating costs for producing corn are usually higher than those for producing grasses. When the costs of purchased feed and fertilizer are higher than the money saved from reduced operation costs for grass production, the farms may not gain in profit. This case was evident from the IFSM results for Scenario 3 for both farms, where a reduction in farm net-return was predicted. Because more corn grain concentrates were purchased in Scenario 3 than Scenario 2, the annual farm net-return in Scenario 3 decreased by $2/cow for the R-farm and $8/cow for the W-farm, compared to the corresponding farm net-returns in Scenario 2 (Table 4.10 and Table 4.11). However, the farm net-returns under Scenario 3 were still greater than those in the baseline scenario and in Scenario 1. When all corn fields were switched to grass in Scenario 4, however, IFSM-predicted that farm net-returns increased for both farms. The increase was mainly due to reduced total cost of production attributed to the reduced cost of machinery. Machinery costs were lower in Scenario 4 than in Scenario 3 for both farms. Generally, with no corn production, fixed costs of owning corn production equipment as well as operation costs were eliminated. Based on the model simulation results, reducing corn silage from production would require that the farm 1) buy more supplemental grain to offset the reduction in corn silage produced, leading to increased farm phosphorus inputs; and 2) increase high productivity forage production, which would increase the utilization of more soil phosphorus on the farm. Moreover, the farm would also reduce farm phosphorus inputs through reduced use of starter phosphorous fertilizer that would have been applied to corn crops. These farm

102 phosphorus inputs and phosphorus outputs counterbalance, thereby resulting in a minimal effect on the farm phosphorus balance Scenario 5 This scenario was applied to the W-farm in order to increase forage production on the under-utilized grass fields as compared to Scenario 4 conditions. In Scenario 4, 4 ha of grass area were under-utilized (Table 4.9). Considering the availability of under-utilized land and the farm s phosphorus excess (1.7 kg/ha; calculated from Table 4.11) that remained in Scenario 4, Scenario 5 explored the option of increasing productivity of under-utilized forage fields as a means of reducing farm phosphorus excess. Excess forage was sold off-farm, thus potentially exporting phosphorus from the farms and increasing profits. As depicted in Table 4.9 for Scenario 4, 28 ha of intensively managed grass satisfied the forage need of the farm. In Scenario 5 the entire grass area of 32 ha (24 ha baseline grass, plus 8 ha corn converted to grass) was placed in high-intensity management. This entailed managing an extra 4 ha of grass as compared to Scenario 4. Reduction in the farm phosphorus imbalance attributable to Scenario 5 was 1.7 kg/ha. That is, compared to the baseline, the phosphorus balance in Scenario 5 was reduced 10.5 kg/ha. This reduction was due to increased utilization of soil phosphorus and export of phosphorus off-farm through the sale of 36 tonnes DM of high-quality forage (refer to Table 4.9 and Table 4.11). The sale of this forage also increased farm N export and ultimately reduced the farm N imbalance. When the forage sale price offsets the cost

103 86 required to produce the extra forage, the farm can also earn additional profit by selling the extra forage. For the W-farm, the farm net-return for this scenario increased by $35/cow compared to Scenario Comments on Modeling PFM Using the IFSM Model Various efforts have been made by the IFSM developers to model a variety of processes occurring in a farm enterprise and to represent a farm in its holistic form. Because of the numerous processes included in IFSM, many interactions occur between the different components. Hence, careful examination of a variety of model parameters related to these processes is required to avoid erroneous representation of farm systems. The following paragraphs focus on observations made during model representation of PFM farm plans. Increasing grass forage production in the model was achieved by applying additional nitrogen fertilizer. This management change, however, could not be administered to harvested grass separately from grazed grass because the model lumps total areas of grass and pasture into one land use. This additional nitrogen fertilizer could cause an overestimate the amount of nitrogen fertilizer applied and increase expenses related to nitrogen fertilizer, nitrogen losses, and pasture yield. Though nitrogen was not the study focus, future plans need to consider improvement in the model to allow better handling of management changes to enable the selective application of nitrogen fertilizer to the desired grass fields. In this study, an attempt was made to adjust nitrogen balances and losses outside the model to temporarily fix this problem. Because pasture yield was increased due to the increased nitrogen fertilizer, the quantities of grazed pasture were

104 also increased. To keep the amount of forage grazed the same as the baseline, 87 adjustments were made to the grazed pasture yield parameter of the IFSM. This parameter reduces the yield rate of the grass grazed. If this parameter is not adjusted grazed grass yield rate increases as a result of more nitrogen application. Crop yield and quality are two main components of IFSM predictions that need to be represented closely to the actual farm values. Crop yield and quality affect the milk production levels and the amount of purchased extra feed supplement that might be required. Though yield and quality of crops are largely determined by soil nutrient levels and weather conditions, there is also a yield adjustment model parameter that can be used to calibrate and closely represent predictions to the expected values. Hence, careful usage of the yield adjustment model parameter is important to avoid erroneous predictions. Moreover, predicted forage yield and quality data need to be compared with historical data and/or the values available from local farm planners to assess the practicality for the farms being evaluated. The IFSM model has two forage level options, high-forage diet and low-forage-diet, for designing forage feed composition. When the high-forage diet option was selected to represent increased forage level in the cows diets, the model was found to underpredict the total dry matter intake of the cows. The model prediction showed that cows would not consume more forage because of the limited rumen volume, but they would still be in need of supplements to fulfill their energy requirement. To overcome this problem, an

105 adjustment was made on an internal model parameter that resulted in increased rumen 88 volume and therefore increased forage intake capacity. Phosphorus loss data from cropland obtained from SWAT modeling was compared to IFSM predictions because of a lack of farm-level observed phosphorus loss data. The absolute phosphorus loss prediction values provided by this study were not verified against observed data. Therefore, future efforts need to include testing and validation of the IFSM phosphorus loss components. 4.5 Summary and Conclusions Phosphorus accumulation in soils within a farm has been a persistent problem for the Cannonsville Reservoir Watershed (CWR) dairy farms. To address farm nutrient imbalances related to phosphorus, while maintaining economic sustainability of farms in the CRW, personnel from Cornell University Cooperative Extension (CCE) of Delaware County are implementing a precision feed management (PFM) program for dairy farms. PFM is a farm-scale BMP that reduces dietary phosphorus inputs to the farm and increases the efficiency of utilization of phosphorus within the farm. CCE personnel recognize the high potential of this BMP to directly address the root cause of the phosphorus build-up on farms attributed to large imports of phosphorus via purchase of concentrates for feed. To augment the PFM efforts, this study provided a model-based evaluation of the impacts of various farm planning strategies on farm profitability and nutrient balance and on subsequent nutrient loss to the environment.

106 89 This study found that the Integrated Farm Systems Model (IFSM) could be used to conduct a comprehensive evaluation of various farm planning strategies prior to their implementation. Such model-based studies done on a farm-by-farm basis are useful in complementing farm planners efforts in exploring innovative farming systems that maintain or increase farm profitability while reducing nutrient imbalances and subsequent off-farm nutrient losses. Also, if a similar study is expanded to multiple farms within a watershed, the findings should aid in incorporating the most critical level of management, farm planning, into watershed-based management planning. This study applied IFSM to two CRW farms to evaluate impacts of several PFM farm planning strategies on phosphorus balance, off-farm phosphorus loss, and farm profitability. IFSM representation of the baseline condition of farms related to their financial and environmental status was verified using actual farm data and the knowledge and field experience of local planners, as well as results from another well-calibrated water quality model. Reducing the dietary phosphorus level to NRC (2001) recommendations resulted in declines in the amounts of mineral phosphorus supplement purchased. Subsequent reductions in the amounts of nutrient phosphorus imported to the farms, and hence reductions in farm phosphorus imbalances of 4.8 kg/ha and 5.8 kg/ha for the two farms studied were noted. Furthermore, the amount of phosphorus in manure excreted was reduced as cows were fed a reduced-phosphorus diet (average reduction of 4.5 kg manure

107 90 phosphorus/cow/year across the two farms). Additionally, simulated off-farm phosphorus loss, mainly soluble phosphorus loss with runoff, was reduced 13% due to the reduced phosphorus concentration in manure applied to fields. Increases in annual farm net-return of $17 and $20 per cow were achieved through reduced phosphorus supplement purchases in dairy feed for the two farms. Reduced phosphorus in diets, in conjunction with increased productivity of grass forages and increased proportion of forage in the diet, resulted in further reductions in farm phosphorus surpluses. When milk production per cow was held constant, the farm phosphorus balances due to this strategy, compared to the baseline scenario, were reduced by 8.2 kg/ha and 9.6 kg/ha for the two farms, indicating that major reductions and possibly no net phosphorus accumulation in soils are achievable when on-farm forage management and utilization is included in PFM. These reductions were due to the decrease in imported feed concentrates, grains, and mineral phosphorus, and a subsequent decline in imported nutrient phosphorus, as well as more efficient use of soil phosphorus by the increased grass forage production. Environmental benefits of improved forage production are based upon increasing the forage level in the diet and reducing grain purchases. In addition to reducing the surplus nutrient phosphorus, producing the same amount of milk with less feed nutrient phosphorus imports improved the efficiency of farm resource utilization and increased farm net-returns. Generally, as long as the costs saved by buying less feed supplements were higher than the costs of N fertilizer and other operations required to increase forage productivity, the profit of the farm was increased. However, because higher nitrogen fertilizer rates were required to increase grass forage

108 production, more study about the nitrogen loss related environmental impacts of higher fertilizer rate uses are necessary with regard to implementing this strategy. 91 Using the IFSM off-farm phosphorus loss predictions, the reduction of off-farm sediment-bound phosphorus due to corn land use conversion to grass was estimated. When corn acreage was converted to grass production, the farm sediment-bound phosphorus losses decreased substantially. For a 50% corn land use conversion to grass, reductions of 26 and 31% in off-farm sediment-bound phosphorus total loads were found (compared to the baseline scenario) for the two farms. When all of the corn area was converted to grass, the reductions increased to 38 and 44%. Converting corn land use to grass showed no appreciable change in the farm phosphorus balance because the availability and increased use of high-quality forage helped to counter balance the increased importation of phosphorus in feed grain purchases when corn silage production was reduced. Conversion of corn to grass, combined with use of high-forage diets and precision diet formulation and delivery, can potentially result in a dramatic cumulative environmental benefit by controlling farm phosphorus surplus and reducing off-farm phosphorus losses. Increasing the crop productivity of under-utilized forage fields and selling excess forage produced was found to reduce the farm phosphorus imbalance for the W-farm. Under this strategy, the farm can also reduce the farm N imbalance and earn profit from selling forages.

109 92 Results of this study confirm that PFM farm plans that include precision diet formulation and delivery, on-farm forage management and utilization and land use management have great potential to benefit farms economically as well as environmentally. Whereas this study focused on two farms, the model-based approach employed is widely applicable, as is the methodology of representing alternative whole-farm system strategies to evaluate and quantify impacts of these strategies on milk production, farm profitability, and farmlevel phosphorus flows and losses. Similar approaches can be applied to dairy farms throughout the northeastern U.S. where implementation of PFM strategies is of interest. Moreover, these results set a benchmark for potential benefits of PFM strategies, both economically and environmentally. Assessment of additional farm system options that meet specific farm objectives is possible using the methodology developed. The methodology developed is, hence, helpful in assessing farm system options prior to their implementation.

110 93 Chapter 5 Effects of PFM Strategies Modeled at a Watershed Scale 5.1 Introduction In the previous chapter (Chapter 4), a model-based method was established to quantitatively evaluate the effectiveness of precision feed management (PFM) in controlling farm phosphorus balance and runoff phosphorus losses, and to assess the economic viability of PFM at a farm level. The model employed to assess farm-level effectiveness of PFM farm planning was the Integrated Farming System Model (IFSM). This whole farm model (Rotz and Coiner, 2006) was applied to study relative benefits of various PFM farm management strategies in relation to both environmental and economic aspects. Similarly, Rotz et al. (2002) showed this model to be a successful tool for evaluating environmental and economic impacts of altering farm management strategies, although their study did not specifically focus on the ongoing PFM farm planning efforts in the Cannonsville Reservoir Watershed (CRW). Farm-level modeling studies are important in evaluating farm planning strategies for their economic and environmental impacts for a farm enterprise. There are, however, limited studies available that have assessed impacts of such farm-level planning within a watershed context. Therefore, a study was undertaken that extrapolated benefits of farmlevel strategies to a watershed scale. This watershed-scale model attempted to show how

111 BMPs impacted water quality at the larger watershed scale. Quantification of 94 environmental impacts of various alternative farm systems is also important for the watershed-based water quality assessment studies, such as those required under the Total Maximum Daily Load (TMDL) program. Also, there is growing interest in establishing quantitatively the impacts of BMPs at the watershed scale. Therefore, the objectives of the research reported herein were to 1) represent the PFMbased farm plans in the Soil and Water Assessment Tool (SWAT; Neitsch et al., 2002) model for the purpose of evaluating the farm-level management scenarios (developed in Chapter 4) at the watershed level; and to 2) to quantitatively assess the effectiveness of the PFM-based farm plans with regard to controlling phosphorus, both at field and watershed scales. 5.2 SWAT Model Description Soil and Water Assessment Tool (SWAT) is a hydrologic and pollutant model that was developed by the Agricultural Research Service of the United States Department of Agriculture (Arnold et al., 1998). SWAT is a process-based, distributed, and continuous daily time-step watershed model that simulates the transport of sediment, runoff, nutrients, and pesticides as a function of land use at subwatershed and watershed scales. SWAT has a long history of successful use in hydrology and in the study of the impact of land management on water quantity and quality (Arnold and Allen, 1996; Arnold et al.,

112 ; Bingner et al., 1997; Cho et al., 1995; Peterson and Hamlett, 1998; Fitzhugh and Mackay, 2000; Santhi et al., 2001a and 2001b; Gitau 2003; Gitua and Gburek, 2005). Most of the SWAT modeling work has been focused on simulating hydrology. Studies by Bingner et al. (1997); Fitzhugh and Mackay (2000); and Santhi et al. (2001a and 2001b) included simulation of constituents such as sediment and nutrients, as well as hydrologic responses. Within the USDA-ARS, SWAT is a model of choice for evaluating the effects of land use and climate on water resources. The SWAT model and its associated GIS interface have been integrated into the United States Environmental Protection Agency s modeling framework of Better Assessment Science Integrating Point and Non-Point Sources (BASINS), which is being used in several states for total maximum daily loads (TMDLs) analysis (Diluzio et al., 2002). In addition, a study by Gitau and Gburek (2005) using the SWAT model successfully assessed impacts of various BMPs including barnyard management, crop rotations, nutrient management plans, and strip cropping at a watershed scale. A study by Santhi et al. (2001b) used the SWAT model to simulate the effect of a dietary phosphorus reduction on phosphorus losses. Because of its wide applicability and previous use within the Cannonsville Reservoir Watershed, the SWAT model was selected for this study. The SWAT model allows a watershed to be divided into subbasins based on topographic criteria, with further subdivision of subbasins into hydrologic response units (HRUs) based on land use and soil type. In order to simplify SWAT runs, areas of a particular

113 land use and soil type within a subbasin are combined together to form one HRU without any consideration to individual fields and with no significance to their spatial location. 96 However, with some modification of the SWAT model input data format, individual fields can be distinctly represented in the SWAT modeling process. The distinct representation of fields is useful during the process of HRU formation to avoid lumping of similar land use and soil combinations of different fields within a subbasin into one HRU. Once the field boundaries are taken into account in the process of HRU development, amounts of runoff and associated sediment and nutrient loadings for each field can then be extracted from the outputs of HRUs that are distinct to each field. SWAT allows the user to define management practices in every HRU. The user can also define the amount and timing of manure and fertilizer application in addition to other management operations. In the SWAT model, the effects of varying concentrations of manure phosphorus and nitrogen on the amount of nutrient loss and water quality can be evaluated. The model also simulates crop growth and crop uptake of phosphorus for the specified management, soil, and weather conditions. SWAT represents phosphorus dynamics using six pools: the fresh (associated with crop residue), active (associated with humus), and stable (associated with humus) organic pools, and the solution, active, and stable inorganic pools. Phosphorus removed from the soil by plants is taken from the solution phosphorus pool. Depending on the total plant biomass grown, or yield rate, the mass of

114 97 phosphorus stored in the plant biomass for each growth stage is determined. In addition, the amount of phosphorus depleted from the solution pool as the result of plant biomass growth is predicted. 5.3 Materials and Methods Study Farm Watershed Description This study was conducted for an entire subwatershed (162 ha) within the CRW (Figure 5.1), which is represented as a single dairy farm for modeling purposes. The farm specifics were studied using IFSM in Chapter 4 (R-farm) to assess impacts of alternative PFM scenarios on farm phosphorus balances, off-farm phosphorus losses, and profitabilities of the farm enterprise. A previous study by Gitau and Gburek (2005) has also successfully applied SWAT to the watershed under conditions both before and after implementation of agricultural management practices. As seen in Figure 5.1, the 162 ha study farm watershed encompasses approximately 60% of the R-farm crop area. Elevations of the farm watershed range from 601 to 735 m above mean sea level. The climate of the area is characterized as humid continental with an average annual temperature of about 8 0 C and precipitation of approximately 107 cm/yr (20-year average).

115 98 Delhi weather station N R-farm watershed # # N 0 20 Kilometers Cannonsville Reservoir Watershed Km R-farm fields ~163 ha R-farm watershed Field ID General land use description and area N # , 40 Banyard (0.13 ha 30 Developed (1.92 ha) 21 *CREP (6.03 ha) 2,4,13,17,24,26,29 Forest (80.83 ha) 1, 5, 7, 8, 9, 14, 15, 16, 19, 20, 27, 28, 33, 24, 35, 36, 37 Grass (39 ha) 10,12,22,23,25,31 Pasture (28.5 ha) 32 Pond (1.05 ha) 38 Road (0.83 ha) 18 Shrub (1.09 ha) 3,6 Silage corn (2.8) Km *CREP = grass area under Conservation Reserve Enhancement Program Figure 5.1: Map showing location of study area, farm, and encompassing watershed.

116 Phosphorus pollution mainly from the application of manure has been a consistent 99 problem of the CRW and is applicable to the study farm watershed. To address phosphorus loss problems, BMPs were implemented on the farm between June, 1995, and November, 1996, as part of a study intended to evaluate the potential effects of BMPs on phosphorus control (Bishop et al., 2005; Hively, 2004). The study watershed was also monitored for stream flow and sediment and phosphorus concentrations for the period before ( ) and after ( ) installation of BMPs. In addition to phosphorus loss problems, on-farm nutrient accumulation, particularly phosphorus, has been a persistent problem of CRW dairy farms. The on-farm phosphorus accumulation, or soil phosphorus build-up, is mainly caused by the imbalance between excess farm phosphorus imports and phosphorus exports. The chosen farm is one of the pilot farms involved in the preliminary field study of the Delaware County PFM program (Cerosaletti et al., 2004) aimed at addressing the root cause of phosphorus accumulation within soils Description of PFM-based Scenarios In this study, alternative farm planning scenarios represented in the SWAT model are part of the PFM farm program developed by the personnel from CCE of Delaware County, NY. These scenarios are also similar to those studied in Chapter 4. These farm planning scenarios are farm-scale BMPs designed to directly target the root cause of phosphorus

117 build-up on the farms and to ultimately reduce phosphorus loadings to the Cannonsville Reservoir. 100 Table 5.1 presents a list of PFM scenarios studied and their description. The PFM-based scenarios include the following: baseline scenario that is the SWAT model representation of the study watershed without PFM farm plan strategies, Scenario 1, which involves reduction of manure phosphorus concentration as a result of feeding cows with dietary phosphorus matching the cows requirement, Scenario 2, which involves application of forage management and utilization in addition to the reduction of dietary phosphorus included in Scenario 1, and Scenario 4, which represents implementation of conversion of the entire corn land area to grass in addition to the practices included in Scenario 2. Scenario 3 studied in Chapter 4, which involved a conversion of the 50% area used for corn production to grass was skipped as the area of corn in the study watershed was small. Table 5.1: List of simulated precision feed management (PFM)scenarios and their descriptions. Scenario Baseline Scenario 1 Scenario 2 Scenario 4 Description Current farming system; conditions before management changes Precision diet formulation and delivery; dietary phosphorus reduction Scenario 1 + increased grass productivity Scenario % corn land converted to grass

118 5.3.3 Baseline Data 101 In this study, the SWAT representation of the farm watershed from a previous study by Gitau and Gburek (2005) was used as a baseline condition against which to compare the impacts of management changes. The study by Gitau and Gburek (2005) applied SWAT to the 162 ha farm watershed to assess the applicability of SWAT in modeling impacts of BMPs at a watershed scale. In doing so, they successfully represented the study farm watershed for pre- and post-bmp installation periods. The SWAT representation for the pre-bmp installation period ( ) was used as a baseline scenario representation for the farm watershed for this study. For representing the baseline conditions in the SWAT model, the 10 m DEM (Digital Elevation Model) topographic and land use data of the farm watershed were used. Weather data, both precipitation and temperature, were taken from the Delhi, NY, metrological station (Figure 5.1). Hydrologic response units (HRUs) were defined by considering individual fields as distinct units and by reproducing SWAT crop parameters for the individual fields. This entailed renaming of land uses with field-distinct names to avoid lumping of fields that have the same land use (Gitau and Gburek, 2005). As a result of such representation, detailed field-level management data including crop rotations (Table 5.2), planting, harvesting, and manure application were represented distinctively for each field. Such detailed representation of field boundaries in the process of HRU development also enabled the amount of runoff and associated sediment and nutrient

119 loadings of a particular field to be distinctively derived from the outputs of HRUs that directly represent the individual, specific fields. 102 Other baseline SWAT input data were: manure was applied to all agricultural crops; planting /beginning of growing dates : alfalfa = May 1; corn = May 15; grass = May 10; and pasture = May 1; harvest dates: alfalfa = June 1 (first cutting), July 15 (second cutting), August 25 (third cutting); grass = May 20 (first cutting), July 1 (second cutting), August 15 (third cutting); and corn = October 1; and grazing dates: May (10, 25), June (10, 15), July (1, 15, 25), August (15, 25), September 25.

120 103 Table 5.2: Study watershed crop land uses used in SWAT for the period. CROP Rotations Field ID* Area (ha) 1 Alfalfa Alfalfa grass-alfalfa Forest Forest Forest grass-alfalfa grass-alfalfa grass-alfalfa Forest Forest Forest grass-alfalfa corn grass-alfalfa** corn corn grass-alfalfa** grass-alfalfa grass-alfalfa grass Alfalfa Alfalfa grass grass grass grass pasture pasture pasture pasture pasture pasture pasture pasture pasture Forest Forest Forest grass grass IDL grass-alfalfa grass-alfalfa IDL grass grass IDL Forest Forest Forest Shrub Shrub Shrub grass grass IDL grass grass grass CREP CREP CREP pasture pasture pasture pasture pasture pasture Forest Forest Forest pasture pasture pasture Forest Forest Forest grass grass IDL grass grass IDL Forest Forest Forest Build-up Build-up Build-up pasture pasture pasture Pond Pond Pond grass-alfalfa grass-alfalfa corn grass-alfalfa grass-alfalfa grass-alfalfa grass-alfalfa grass-alfalfa corn grass-alfalfa grass-alfalfa corn grass-alfalfa grass-alfalfa grass-alfalfa Road Road Road Barnyard Barnyard Barnyard Barnyard Barnyard Barnyard 0.04 * location of the corresponding fields are also shown in Figure 5.1. ** newly established grass-alfalfa mix IDL = idle land; CREP = grass area under Conservation Reserve Enhancement Program; Build-up = areas used for building; P i l d fi ld d f i

121 5.3.4 SWAT Representation for PFM-based Scenarios Scenario 1 A model representation of Scenario 1, dietary phosphorus reduction, required modification of animal feed rations (feed phosphorus content) to match the NRC (2001) diet phosphorus recommendations. When dietary phosphorus content of cows varies, both total and water-soluble phosphorus contents of manure are expected to change (Satter and Wu, 1999; Ebeling et al., 2002; Dou et al., 2002; Cerosaletti et al., 2004). Therefore, in the SWAT model representation, manure phosphorus concentration was the only environmental input parameter that was varied to reflect changes due to implementing the precision feeding strategy. The total mass of manure applied to the crops, the amount excreted by the grazing cows, and dates of application were kept the same as in the baseline representation General background information on manure production and phosphorus content Determination of manure phosphorus content. Actual phosphorus concentration of manure produced from a particular farming system can be obtained by performing laboratory analysis of manure samples. In addition, there are several ways of estimating manure phosphorus concentration for a prescribed amount of phosphorus nutrient in animal feed rations. One of these ways is to determine manure phosphorus content using

122 105 already developed relationships between phosphorus feed levels and manure phosphorus contents. Such relationships have been developed though many studies (Wu et al., 2001; Dou et al., 2002; Dou et al., 2003; and Toor et al., 2005). These relationships (equations) were developed from dietary and fecal samples collected from individual cows in an effort to investigate the effect of dietary phosphorus modifications on phosphorus excretions by dairy cows. Since the relationships were developed for specific types and sizes of animals, feed conditions, and locations, these equations are most useful in conditions where experimental data are limited. Another way of determining manure phosphorus concentration is to use a process-based animal model such as the Cornell Net Carbohydrate and Protein System (CNCPS; Fox et al., 2003) that better quantifies the concentration of phosphorus in the manure excreted. The CNCPS uses a combination of mechanistic and empirical approaches to account for the effects of animal characteristics and feed carbohydrate and protein types on animal performance (Fox et al., 2003). The CNCPS model predicts nutrient requirements, annual feed requirements, and nutrient excretions for each group in the herd and for a whole herd. This process-based animal modeling approach is expected to represent the feedingdigestion-excretion processes closely because of the considerations of various physiological functions in cattle including intestinal digestion, metabolism of absorbed nutrients, maintenance, growth, pregnancy, and lactation. Manure phosphorus prediction by the CNCPS model are based on an input minus output approach; the amount of phosphorus in manure is calculated by subtracting the amount of phosphorus utilized in

123 production (milk, tissue, and pregnancy) from the total dietary phosphorus fed to the 106 animal. The IFSM model, detailed in Chapter 4 (Rotz and Coiner, 2006), uses equations to relate the effects of animal characteristics and feed types on animal performance. Predictions of phosphorus in manure are based on an input minus output approach. Similar to the CNCPS model approach, the quantity of phosphorus produced in the manure is calculated as a function of the quantity and phosphorus content of the feed consumed and the quantity of phosphorus content utilized by the cow in milk production, growth, and maintenance. Manure phosphorus database in the SWAT model. The SWAT model consists of a database with information on typical amounts of manure and phosphorus content for different types of animals. This (manure and fertilizer) database consists of data on levels of nitrogen and phosphorus in manure obtained from a wide range of manure production and analysis studies that are derived from manure production and characteristics data compiled by the ASAE (1998a). When site-specific field data on manure phosphorus are unavailable for a watershed of interest, manure phosphorus contents and/or fractions of the different forms of manure phosphorus from the database can be used to obtain estimated values for SWAT modeling purposes. However, these phosphorus concentration values don t provide specific manure phosphorus contents for varying dietary phosphorus levels.

124 For example, for dairy cows the levels of various phosphorus forms in manure as 107 reported in ASAE (1998a) and Neitsch et al., (2002a) are presented in Table 5.3. Table 5.3: Fresh manure production and characteristics per 1000 kg dairy animal per day (adopted from ASAE, 1998a; Neitsch et al., 2002a). Parameter Total manure kg mean 86 Std dev 17 Total solids kg mean 12 Std dev 2.7 Total phosphorus kg mean Std dev Orthophosphorus kg mean Std dev Comment Note that orthophosphorus accounts for 65% of the total phosphorus, and Organic phosphorus kg of the total phosphorus organic phosphorus accounts for 35% 1 estimated value, organic manure phosphorus = total manure phosphorus manure orthophosphorus In relation to manure phosphorus data, the SWAT model assumptions are: that the mineral phosphorus pool is equal to the value given for orthophosphorus; and the organic phosphorus pool is equal to total phosphorus minus the value for orthophosphorus. The model developers suggest that default values from Table 5.3 can be adopted or proportions thereof can be used to estimate missing data (Neitsch et al., 2002a). Based on the above assumption that orthophosphorus is equivalent to the mineral phosphorus, the mineral phosphorus pool is 65% of the total manure phosphorus. Hence the remaining 35% of the total manure phosphorus is in its organic form.

125 Manure phosphorus content derivation for the SWAT model 108 To use the SWAT model to evaluate the change of dietary phosphorus on phosphorus runoff losses in a given watershed, information was required on the concentrations of different phosphorus forms in the manure produced within the watershed for the prescribed dietary phosphorus levels. Because the study watershed encompasses a single dairy farm, IFSM simulation of the dairy farm was employed to determine the change in the manure phosphorus content as a result of the application of precision feeding. Refer to the flowchart in Figure 5.2 for a summary of manure phosphorus data derivation for SWAT model representation of Scenario 1. Feed phosphorus level adjustment Current farm data (Baseline) Other land uses IFSM Output Partitions of manure phosphorus into organic and mineral components (as suggested in SWAT (Neitsch et al., 2002a) SWAT representation of farm watershed (Gitau and Gburek, 2005) Manure phosphorus content Phosphorus losses at field-level and watershed-level Figure 5.2: Methodology flow chart for data derivation and Scenario 1 representation in the SWAT model.

126 109 IFSM predicts total manure phosphorus excreted from the animal to be the difference between the phosphorus contained in the milk produced plus animal tissue growth and the total phosphorus intake by cows. For the dairy farm studied (R-farm in Chapter 4), IFSM predicted the total manure dry matter produced and the total phosphorus content for both a baseline scenario and Scenario 1 (when dietary phosphorus was reduced). As shown in Table 5.4, compared to the baseline condition, the reduction in total manure phosphorus content of the farm in Scenario 1 was 25%. Table 5.4: IFSM predicted manure production, total manure phosphorus content, and estimated amounts and concentration of mineral and organic components of manure phosphorus for baseline scenario and Scenario 1 of the study farm (R-farm). Parameters Baseline Scenario 1 IFSM-predicted values for the farm in the study watershed Total manure produced, kg 275, ,000 Total phosphorus in the manure, kg 1,949 1,471 Total phosphorus partitioned as suggested in the SWAT users manual (Neitsch et al., 2002a) Mineral phosphorus in the manure, kg (65% of the total phosphorus) 1, Organic phosphorus in the manure, kg (35% of the total phosphorus) Calculated concentration (mineral or organic total manure produced) Mineral phosphorus, kg/kg Organic phosphorus, kg/kg Predicted manure phosphorus content values obtained from the IFSM model represent the amount of total phosphorus in the manure produced. SWAT simulation of phosphorus losses requires that the total manure phosphorus be divided into its mineral and organic components. Methods, as suggested in the SWAT model manual (Neitsch et al., 2002a),

127 110 were used to estimate the mineral and organic phosphorus contents from the total manure phosphorus produced on the farm. Hence, the total manure phosphorus predicted by IFSM was partitioned into 65% mineral and 35% organic phosphorus forms. Table 5.4 presents estimated values of the mineral and organic phosphorus in the manure produced on the farm. Table 5.4 also presents IFSM-predicted manure phosphorus concentration values for the baseline scenario and Scenario 1. Manure phosphorus concentration input data for mineral and organic phosphorus were calculated by dividing the amount of mineral (or organic) phosphorus by the total amount of manure produced. For example, the kg/kg mineral phosphorus concentration of manure for the baseline scenario was calculated by dividing the 1267 kg of mineral phosphorus by the 275,000 kg of total manure produced (Table 5.4). The same procedure was used to calculate Scenario 1 manure phosphorus concentration data Manure phosphorus concentration input data used for Scenario 1 Table 5.5 presents IFSM-predicted manure phosphorus concentration values for the baseline scenario and Scenario 1 (taken from Table 5.4). The IFSM-predicted values for mineral and organic forms were used as Scenario 1 manure concentration input data in the SWAT model. Manure phosphorus concentration data used in SWAT modeling for the baseline scenario (Gitua and Gburek, 2005) are also presented in Table 5.5. As shown in Table 5.5, baseline manure phosphorus concentration (mineral/organic) in the IFSM

128 modeling and the SWAT modeling by Gitau and Gburek (2005) were 0.004/0.002 kg/kg and 0.005/0.002 kg/kg, respectively. 111 Table 5.5: SWAT model manure phosphorus concentrations for Scenario 1. Parameter Manure phosphorus concentration, kg/kg Baseline condition Scenario 1 IFSM-predicted* SWAT-Baseline** IFSM-predicted* Mineral phosphorus Organic phosphorus *concentration values taken from Table 5.4 **concentrations of manure phosphorus used in the SWAT for the baseline (Gitau and Gburek, 2005) In summary, manure phosphorus concentrations used in the SWAT-Scenario 1 baseline representation were reduced from the baseline to reflect dietary phosphorus intake reductions, which translated into reduced phosphorus concentration in manure Scenario 2 Scenario 2, forage management and utilization integrated with precision feeding and delivery, involved increasing grass-forage productivity and feeding cows with a highforage diet, in addition to adjusting feed phosphorus to the prescribed NRC level. For proper SWAT representation of this scenario, first, the manure phosphorus concentrations were adjusted to reflect the balanced phosphorus diet (methods detailed in baseline input Section ), and second, an increase in grass crop yield was represented in the model.

129 112 For SWAT simulations where a certain amount of crop yield and biomass is required, the model user can force the model to meet this amount by setting a harvest index target and a biomass target (Neitsch et al., 2002a). However, this method is effective only if a harvest and kill operation is used to represent the harvest of crops (Neitsch et al., 2002a). The harvest and kill operation in SWAT specifies a harvest at one time followed by killing of plants after harvest. In this study, a harvest only operation, rather than the harvest and kill operation, was selected to represent cutting grass a multiple number of days. The harvest only operation in the SWAT removes plant biomass without killing the plant, and this operation is most commonly used to represent the cutting of hay or grass forage (Neitsch et al., 2002a). Therefore, increasing the crop yield by setting the SWAT harvest index target and a biomass target was not appropriate. Another way of attaining increased crop-yield in the SWAT model is through applying management changes that promote an increase in crop yield. The main factors affecting plant growth representation in the SWAT model are water, temperature, phosphorus, and nitrogen. Increasing the application rates of nitrogen fertilizer, particularly for grass fields, was expected to increase grass yield. Hence, in this study, application of nitrogen fertilizer was used to increase grass yield. In the SWAT modeling of Scenario 2, an additional 100 kg/ha of nitrogen fertilizer (elemental nitrogen fertilizer, N) as suggested by farm planners was applied to the grass land to attain increased prescribed forage production. The entire 100 kg N/ha was applied on May 11 when grass plants start growing (May 10 in the model). For simplification of

130 modeling a one time nitrogen fertilizer application was used, though splitting nitrogen application rates into two or more application rates was a practical approach. 113 Based on the study in Chapter 4, the area required for production of a high-forage diet was found to be 61 ha of grass (out of 63 ha of total grass on the study farm). This required an increase in the yield rate target of 2 tonnes/ha beyond the 6 tonnes/ha in the baseline. The additional 2 tonnes/ha yield target was suggested by farm planners as it reflects an attainable yield increase on the farm. To examine the maximum effect of intensified forage management on soil phosphorus status and losses, however, the entire grass fields in each crop year ( ) within the study watershed were involved in the intensive forage management (shaded color in Table 5.2). The total grass areas within the study watershed available for intensive management were 35 ha, 33 ha, and 26 ha for the respective crop years of 1993, 1994, and As a result, the SWAT-predicted average grass yield rate was increased to 4.3 tonnes DM /ha (from 2.4 tonnes DM/ha in the baseline scenario). This is close to the yield rate increase target of 2 tonnes/ha, an attainable yield increase suggested by the farm planners (Dewing and Cerosaletti, 2005).

131 Scenario 4 Scenario 4 of the SWAT modeling involved conversion of corn production areas to grass production, in addition to the practices used in Scenario 2. Representation of this scenario in the SWAT model required conversion of corn fields to grass crops in addition to the practices included in Scenario 2. Changes in the management file in the SWAT model such as planting and harvesting dates were made to reflect land use conversion of corn to grass. Corn crops used in crop rotations (Table 5.2) were changed to grass land use. For the years of 1993, 1994, and 1995, 1.58, 3.62, 3.38 ha, respectively, were converted from corn to grass. These areas represent fields with identification (id) number of 6 for 1993, 5 and 6 for 1994, and 33, 35, and 36 for 1995 (Table 5.2). 5.4 SWAT Model Phosphorus Dynamics and Expected Effects of PFM Implementation Prior to discussing SWAT simulation results for the PFM farm plans, a brief review of the SWAT model phosphorus dynamics is presented. The SWAT model represents phosphorus dynamics using mineral and organic pools (Figure 5.3). The mineral pools are further represented by solution, active, and stable inorganic pools. The solution pool is in rapid equilibrium (several days or weeks) with the active pool. The active pool is in slow equilibrium (several years; Neitsch et al., 2002b) with the stable pool. The organic pools consist of fresh (associated with crop

132 residue), active (associated with humus), and stable (associated with humus) organic 115 pools. Figure 5.3: SWAT soil phosphorus pools and processes (adapted from Neitsch et al., 2002b). Organic and mineral phosphorus bound to soil particles may be transported from the soil surface by surface runoff. Sediment-bound phosphorus loss in the SWAT model is thus estimated as a function of sediment-bound phosphorus concentration in the top soil layer, the sediment yield, and an enrichment ratio. Soluble phosphorus loss is computed as a function of soluble phosphorus concentration in the top soil layer and runoff volume. SWAT simulation results are expected to demonstrate that lowering of manure phosphorus content reduces the phosphorus loss in runoff. Applying manure with a lower phosphorus concentration on fields would affect availability of phosphorus in both the

133 active organic and the solution (labile) inorganic pools. Thus, both soluble and 116 sediment-bound phosphorus losses are expected to be decreased as a result of lowering the phosphorus concentration of manure inputs to the soil. Increasing the crop yield rate is expected to increase the demand for soil phosphorus and thus result in a decrease over time in the amount of phosphorus in the phosphorus pools. In addition, increasing crop cover caused by increased yield rate is expected to decrease soil erosion and associated phosphorus losses. The plant cover (C) factor in the erosion prediction equation in the SWAT model is updated every day that rainfall occurs as a function of the average annual C factor for the land cover and the amount of plant residue on the soil surface. The C factor is the land cover and management factor in the Modified Universal Soil Loss Equation (MUSLE; Williams, 1975). Updating the C factor values to reflect the crop cover predicts impacts of the increased-yield of grass on the amount of soil erosion and associated phosphorus losses. The SWAT model plant phosphorus equation calculates the amount of phosphorus in the plant biomass as a function of plant growth stage given optimal growing conditions (Neitsch et al., 2002b). The mass of plant phosphorus for any growth stage is determined by multiplying the phosphorus fraction factor at that growth stage by the total plant biomass (Neitsch et al., 2002b). Therefore, increasing the yield rate is expected to increase the total plant phosphorus need as a result of increased biomass growth. This would also increase total plant uptake because the total plant phosphorus demand for a given day in SWAT is determined as a function of total plant biomass.

134 Thus, the actual amount of phosphorus removed from a soil layer is estimated as a 117 function of the plant s phosphorus uptake. SWAT assumes phosphorus removal from the soil by plants to take place from the solution phosphorus pool (Neitsch et al., 2002b). Therefore, increasing crop yield increases plant phosphorus uptake and decreases the amount of phosphorus in the solution pool. Because the solution pool is also assumed to be in rapid equilibrium with the active pool (Neitsch et al., 2002b), increasing crop yield is expected to affect the amount in the active inorganic phosphorus pool. However, a minimal effect is expected to occur (within a few years of simulation) in the amount of phosphorus in the stable pool since the SWAT model assumes that the active pool is in slow equilibrium with the stable pool (Neitsch et al., 2002b). Converting corn silage land to grass-based forage is expected to reduce sediment and associated phosphorus losses from land previously used in the production of corn silage. Because sediment loss from corn fields is usually higher than that of grass fields, changing from corn production to grass production is expected to reduce sediment losses. In addition, because the SWAT model predicts sediment-bound phosphorus loss as a function of sediment loss, reduced loss of sediment-bound phosphorus is also anticipated. 5.5 Results and Discussion In this section, impacts of PFM-based scenarios on water quality were examined based on the results obtained from the SWAT simulations. Brief discussions on the simulated results of runoff and losses of sediment and nitrogen are also included.

135 Model simulation in SWAT was conducted for 13 years (1983 to 1995); however, 118 output data representing three years ( ) were used in the analysis because these years represent a period of time before best management practices (BMPs) were installed in the watershed. This also represents a period for which baseline conditions are available against which results of management changes were compared. As mentioned in Chapter 5, SWAT representation of the study watershed and predictions of hydrology and nutrient losses for this period ( ) were previously performed by Gitau and Gburek (2005). Results of their study were used in this study to represent baseline conditions. Field-level results represent predictions from agricultural lands, whereas, watershed-level simulation results include total predictions for all land uses within the watershed (agricultural, forest, and other lands). No channel processes were considered because the study watershed is small (163 ha) Impacts of PFM on Surface Runoff and Streamflow SWAT-simulated results of surface runoff and stream flow for different agricultural land uses and watershed levels are presented in Figure 5.4 and Figure 5.5. Responses of surface runoff and stream flow to the various scenarios simulated were examined and found to have a negligible difference compared to the baseline scenario, for both agricultural-field and watershed levels (Figure 5.4 and Figure 5.5). Exceptions are the slight changes in the average annual surface runoff depths when the practice of land use conversion to grass was applied (Scenario 4). However, in reality, variation in the

136 predicted surface runoff and stream flow for Scenario 2 compared to baseline would 119 be expected due to increased biomass production, which in turn affects rainfall interception, water storage capacity, infiltration, and ultimately runoff generation. averge annual surface runoff & Stream flow "mm" Average from agricultural fields surface runoff stream flow Baseline Scenario 1 Scenario 2 Scenario 4 Figure 5.4: SWAT-simulated average annual surface runoff and total stream flow contributions from agricultural fields for all scenarios simulated during simulation years 1993 to 1995.

137 Stream flow "cms" /1/1993 3/1/1993 5/1/1993 7/1/1993 9/1/ /1/1993 1/1/1994 3/1/1994 5/1/1994 Baseline Scenario 1 Scenario 2 Scenario 4 Note: essentially no differences in watershed response were observed for these various scenarios. Figure 5.5: SWAT-simulated daily stream flows at the outlet of the study watershed for all scenarios simulated during the period of 1993 to /1/1994 9/1/ /1/1994 1/1/1995 3/1/1995 5/1/ Impact of PFM on Sediment Losses SWAT-predicted sediment losses for all scenarios are presented for different data levels, for field-by-field, general agricultural land uses, and watershed-level, respectively. Table 5.6 presents simulated sediment losses from each field for simulation period of 1993 to 1995 (Table 5.2).

138 Table 5.6: SWAT-simulated annual sediment loss on field-by-field basis for R-farm watershed for 1993, 1994 and Field ID 121 Sediment loss, tonnes Area Simulation Runs 1 Simulation Runs Simulation Runs ha B S.1 S.2 S.4 B S.1 S.2 S.4 B S.1 S.2 S Simulation runs B = baseline, S.1 = Scenario 1, S.2 = Scenario 2, and S.4 = Scenario 4

139 122 Sediment loss summarized by fields: Simulated sediment loss results of HRUs were summarized by individual fields. This was possible because field boundaries were used in place of a general land use mapping, hence HRUs coincided with fields. For all simulation years, Scenario 1 had no effect on the amount of sediment loss prediction. This was deemed reasonable because Scenario 1 only involved changing the concentration of phosphorus in manure applied to the farm. For Scenario 2, an increase in grass yield resulted in a reduced amount of sediment loss (Table 5.6) for the fields in which yield was increased (fields shaded with color in Table 5.2). For example, sediment loss reduction in fields with id #1, #14, #27 can be observed. For Scenario 4, when corn land was switched to grass production (in addition to practices included in Scenario 2), SWAT-predicted sediment loss was reduced substantially for corn fields with id #6 (1993), #5 and #6 (1994), and #33, #35, and # 36 (1995) (refer to Table 5.6). Note that predicted sediment loss from fields with id #5 and #6 (1995) were reduced substantially though they were not planted with corn in the corresponding year. These two grass-alfalfa mix fields, however, had corn planted in previous year and were newly established (Table 5.2). During establishment of these fields (grass-alfalfa mix), tillage operation had been applied that had resulted in a more disturbed soil and hence higher sediment losses. When corn land use was replaced by grass for the previous year (1994), tillage operation was excluded as there was not need to re-establish grass (1995). Therefore, sediment losses from fields with id #5 and #6 (1995) were reduced more likely because of the exclusion of tillage operation.

140 123 Sediment loss summarized by agricultural land uses: Data in Table 5.7 consists of simulation results from agricultural fields only. This was because effects of management changes were expected from these agricultural fields where management was altered. For all alternative scenarios, other land uses within the watershed (such as road, barnyard, forest) were kept the same as the baseline. For Scenario 1, as described previously, no effect on the amount of sediment loss prediction was observed. For Scenario 2, an increase in grass yield resulted in a reduced amount of sediment loss for the grass fields (Table 5.7). Average sediment loss reductions compared to the baseline scenario for grass fields was 15% (Table 5.7; ( )/9.6) for the three years analyzed. The reduction in predicted sediment loss is likely because of interception of rainfall energy by the increased land cover, thus decreasing erosion. In the SWAT model, the plant cover (C) factor is updated daily during the growth cycle of the plant to represent plant biomass grown (Neitsch et al., 2002b). For Scenario 4, when corn land was switched to grass production (in addition to practices included in Scenario 2), SWAT-predicted sediment loss was reduced substantially (Table 5.7). This is due to greater crop vegetative soil cover by grasses. The annual crop C factor of corn (default annual C factor of corn ~ 0.2; Neitsch et al., 2002a) is larger than that of the grass crops (default annual C factor of grass ~0.003; Neitsch et al., 2002a), implying lower interception ability of corn crops to rainfall energy compared to grasses. In addition, grass crops cover the ground for greater part of the year.

141 Table 5.7: SWAT-predicted average annual sediment load from agricultural fields for all scenarios simulated during the period of Year land use Land use area (ha) Baseline scenario Scenario 1 Sediment yield, tonnes (dietary phosphorus reduction) Sediment Land use yield, (ha) tonnes Scenario 2 (dietary P reduction & increased grass yield by applying nitrogen fertilizer) Land use (ha) Sediment yield, tonnes Scenario 4 Scenario 2 + conversion of corn to grass Land use (ha) Sediment yield, tonnes Grass Corn Pasture Total crops Grass Corn Pasture Total crops Grass Corn Pasture Total crops Three year average sediment yield Grass Corn Pasture Total from agricultural fields % reduction from baseline condition 0% 7% 58% 1 includes fields with grass and grass mix alfalfa harvested for hay or silage; 2 includes grass fields used for grazing; Note that, due to crop rotation, areas of agricultural fields are different for the three years presented

142 For example, as shown in Table 5.7, predicted sediment loss in 1993 for the 42.5 ha 125 grass field (after 1.6 ha corn area was converted to grass) was 3.1 tonnes. In Scenario 2, annual predicted total sediment loss for the 42.5 ha grass fields (40.9 ha grass ha corn) was 5.3 tonnes (2.9 tonnes tonnes). This resulted in 42% (( ) 5.3) reduction in sediment loss for Scenario 4 compared to Scenario 2 for the 1993 simulation. Sediment loss in 1994 for the 42.5 ha grass field (after 3.6 ha corn area was converted to grass) was 2.4 tonnes. In Scenario 2, annual predicted total sediment loss for the 42.5 ha grass fields (38.9 ha grass ha corn) was 20.5 tonnes (2.0 tonnes tonnes). For the year of 1995, sediment loss for the 42.5 ha grass field (after 3.4 ha corn area was converted to grass) was 9.3 tonnes. In Scenario 2, annual predicted total sediment loss for the 42.5 ha grass fields (39.1 ha grass ha corn) was 22.2 tonnes (19.5 tonnes (grass) tonnes (cornfields)). Note that the grass fields in Scenario 2 had high sediment loss due to the fact that some grass fields were newly established during this year. With the replacement of corn fields to grass for the previous year, however, these fields were not newly established and hence tillage operations were excluded. This is an additional reason that predicted sediment loss for grass fields in Scenario 4 was much lower than the sediment losses from grass fields in Scenario 2. Sediment loss summarized for the study watershed: Figure 5.6 shows predicted average total sediment loads from the entire study watershed for the three years. At a watershed-level, average sediment loads for the three years simulated for Scenario 2 were

143 126 reduced to 39.7 tonnes from 41.7 tonnes in the baseline scenario (Figure 5.6). Hence, sediment loss reductions from grass fields contributed to a 5% (( )/39.7) reduction of total sediment loss at a watershed-level for the three years simulated. The average grass area over the three years (1992 to 1995) was 25% of the total study watershed area. At the watershed level, 12.3 tonnes ( tonnes) reduction in the predicted sediment loss for Scenario 4 compared to the Scenario 2 was attained due to switching 2.9 ha ((1.6 ha ha ha) 3) of corn land to grass for three years ( ) under Scenario 4 and due to excluding tillage operations from grass fields (Figure 5.6). Note that for the three years, the average corn area converted to grass (2.9 ha) under Scenario 4 represented 1.8% of the total watershed area.

144 127 Sediment load "tonnes" Sediment load Baseline Scenario 1 Scenario 2 Scenario 4 Note: Watershed-level results represent the sum of sediment prediction from all land use losses within the watershed, including crop fields, forests, build-ups, and others within the study watershed. Figure 5.6: SWAT-predicted average annual sediment loads for the study watershed (~163 ha) for all scenarios simulated during the period of Impacts of PFM on Phosphorus Losses and Soil Phosphorus Complete results of the SWAT-simulated phosphorus losses and soil-buildup for the scenarios studied are presented in the following two sections. Since assessment of phosphorus related impacts from implementing PFM farm strategies were the main focus of this study, a detailed discussion is presented.

145 Impacts of PFM on Phosphorus losses 128 Phosphorus loss summarized by fields: Predicted phosphorus losses for individual fields within the study watershed for each simulation year are presented in Appendix A. Land use type for each field in Appendix A identified with identification number (# 1, 2, 3, and 40) can be found in Table 5.2. For Scenario 1, a 25% reduction in manure phosphorus content to the SWAT baseline data resulted in reduction of the predicted particulate phosphorus (PP) and soluble phosphorus (SolP) losses for fields that received manure application. For all agricultural fields (Appendix A) that received manure, phosphorus losses (PP and SolP) were reduced due to the lower phosphorus content of manure applied to the fields. Application of Scenario 2, which involved increasing the productivity of grass fields through improved management and a 25% reduction in manure phosphorus applied, resulted in the PP loss reduction from the grass fields that were intensified. The decrease in the predicted PP loss was due to lower sediment loss (Section 5.5.2) to which PP could be attached. For Scenario 4, when corn land was switched to grass production (in addition to practices included in Scenario 2), SWAT-predicted PP loss was reduced substantially for corn fields with id # 6 (1993), #5 and #6 (1994), and # 33, # 35, and # 36 (1995) (Appendix A). However, predicted SolP was increased for these fields. For example, PP loss for field id #6 for 1993 was reduced from 1.78 kg (Scenario 2) to 0.23 kg in Scenario 4; and SolP loss for field id #6 for 1993 was increased from 0.87 kg (Scenario 2) to 1.10 kg in Scenario 4. Also note that, as shown in Appendix A, PP losses were reduced for grass

146 fields with id #5 and #6 (1995) due to exclusion of tillage operation and subsequent 129 lowering in sediment loss (Section 5.5.2) to which PP could be attached. Phosphorus losses summarized by agricultural land uses: Predicted phosphorus losses for agricultural land uses (summarized from the data presented in Appendix A) within the study watershed for each simulation year are presented in Table 5.8. Table 5.9 presents baseline-predicted values and percentage changes of phosphorus losses of simulated scenarios from the baseline separately for each simulation year. Table 5.10 and Figure 5.7 show average phosphorus loss predictions for general agricultural fields during the simulation period. Table 5.10 also presents three year average baseline-predicted values and percentage changes of phosphorus losses of simulated scenarios from the baseline. For Scenario 1, a 25% reduction in manure phosphorus content to the SWAT baseline data resulted in reduction of the predicted particulate phosphorus (PP) and soluble phosphorus (SolP) losses for fields that received manure application. Effects (Table 5.8 and Table 5.9) were observed from all agricultural land uses that received manure application. For each agricultural land use (grass, corn, and pasture), phosphorus losses (PP and SolP) were reduced due to the lower phosphorus content of manure applied to the fields (Figure 5.7). Average reductions from these agricultural land uses were 12% for PP and 13% for SolP (Table 5.10). A study by Santhi et al. (2001b) using the SWAT model to simulate the effect of a 29% reduction in manure phosphorus concentration reported a reduction of 12% in soluble phosphorus losses.

147 130 Table 5.8: SWAT-predicted particulate phosphorus (PP), soluble phosphorus (SolP) and total phosphorus (TP) losses from agricultural land uses for all scenarios simulated during the period of 1993 to SWAT-simulated phosphorus losses, kg/ha Baseline Scenario 1 Scenario 2 Scenario 4 Year Land uses Land use area, ha PP 1 SolP 2 TP 3 Grass Corn Pasture Land use area, ha PP SolP TP Land use area, ha PP SolP TP Land use area, ha PP Sol P TP Grass Corn Pasture Grass 1995 corn Pasture PP = sediment-bound phosphorus; 2 SolP = soluble phosphorus; 3 TP = total phosphorus. 4 includes fields with grass (grass mix alfalfa) harvested for hay or silage; 5 includes grass fields used for grazing; Note that, due to crop rotation, areas of agricultural fields are different for the three years presented

148 131 Table 5.9: SWAT-predicted three year particulate phosphorus (PP), soluble phosphorus (SolP) and total phosphorus (TP) losses from agricultural land uses for baseline scenario, and percent change of predicted losses of three scenarios compared to the baseline. SWAT-simulated phosphorus losses, kg/ha Percentage reduction (-) or increase (+) compared to the baseline Baseline Scenario 1 Scenario 2 Scenario 4 Year Land use Land use area, ha PP 1 SolP 2 TP 3 Land use area, ha PP SolP TP Land use area, ha PP SolP TP Land use area, ha PP SolP TP Grass Corn 1.6 Pasture Grass Corn 3.6 Pasture Grass Corn Pasture PP = sediment-bound phosphorus; 2 SolP = soluble phosphorus; 3 TP = total phosphorus; 4 includes fields with grass (grass mix alfalfa) harvested for hay or silage; 5 includes grass fields used for grazing; Note that due to crop rotation areas of agricultural fields are different for the three years presented

149 132 Table 5.10: SWAT-predicted annual average phosphorus losses from agricultural land uses (grass, corn, and pasture) for three years ( ) for all scenarios simulated, and percent change of predicted phosphorus losses of these scenarios compared to the baseline. SWAT-Simulated average phosphorus loss, kg Scenarios Particulate phosphorus Soluble phosphorus Total phosphorus (PP) (SolP) (TP) Baseline Scenario Scenario Scenario Percentage reduction (-) compared to the baseline scenario Baseline Scenario Scenario Scenario Phosphorus loss, kg PP SolP TP Baseline Scenario 1 Scenario 2 Scenario 4 Figure 5.7: SWAT-predicted average agricultural land uses (grass, corn, pasture) losses for three years ( ) for particulate phosphorus (PP), soluble phosphorus (SolP) and total phosphorus (TP) for all scenarios simulated.

150 133 Application of Scenario 2, which involved increasing the productivity of grass fields through improved management and a 25% reduction in manure phosphorus applied, resulted in a minimal effect on phosphorus loss compared to Scenario 1. Exceptions were observed at the PP loss reduction from the grass fields that were intensified (Table 5.8, Table 5.9, and Figure 5.7). Generally, at a field scale, for grass fields, average PP and SolP loss reductions for the three year simulated were 15 and 12%, respectively, compared to the baseline scenario (Table 5.9). Average PP and SolP loss reductions for these grass fields for the three year simulation in Scenario 1 were 9% and 13%, respectively, compared to the baseline scenario. The decrease in the predicted PP loss was due to lower sediment loss (Section 5.5.2) to which PP could be attached. Scenario 4, which involved conversion of corn area to grass in addition to the practices considered in Scenario 2, resulted in a substantial decrease in the predicted PP losses for fields where conversion of corn to grass took place. Crop rotation data for the analysis period can be seen in Table 5.2. With regard to its main objective, i.e. reduction of erosion and associated phosphorus loss, the strategy of corn land conversion to grass was deemed to be effective. This was supported by the findings of modeling results from the SWAT model (Table 5.8, Table 5.9, Table 5.10, and Figure 5.7). For example as seen in Table 5.10, predicted average PP loss for crop fields was reduced to kg from kg in Scenario 2. However, the predicted average SolP loss for crop fields was slightly increased from

151 kg in Scenario 2 to kg in Scenario 4. Compared to the baseline scenario the predicted average PP loss due to conversion of corn to grass (combined with practices in Scenario 2) declined by 53% while average SolP decreased by 10% for the crop fields (grass, corn, and pasture). Average reductions for agricultural fields in Scenario 2 were 15% and 14% for PP and SolP, respectively, compared to the baseline scenario. Phosphorus losses summarized for the study watershed: For Scenario 1, watershedlevel predicted phosphorus losses are depicted in Table 5.11 and Figure 5.8. Watershedlevel results represent the sum of predictions from all land uses in the study watershed (grass, corn, pasture, forest, roads, and other land). At the watershed-level, predicted PP and SolP reductions for the three years averaged 7% and 11%, respectively, compared to losses predicted in the baseline scenario (based on data of Table 5.11). Overall, reduction in field-level phosphorus loss due to applying a lower concentration of phosphorus in manure resulting from feeding reduced dietary phosphorus eventually manifested its effect at the watershed level. For Scenario 2, predicted PP loss from grass fields contributed to reductions in the predicted PP loss at a watershed-level for the three years simulated (Figure 5.8). Predicted PP loss as a result of increased grass production was only slightly reduced compared to Scenario 1. The percentage reductions for PP and SolP were 9% and 11%, respectively, compared to the baseline scenario (Table 5.11). The percentage reductions for PP and SolP in Scenario 1 were 7% and 11%, respectively, compared to the baseline

152 scenario (Table 5.11). Average grass area over the three years (1992 to 1995) was 25% of the study watershed area. 135 Table 5.11: Watershed-level* SWAT-predicted three year ( ) average of particulate phosphorus (PP), soluble phosphorus (SolP) and total phosphorus (TP) losses for Baseline, Scenario 1, Scenario 2, and Scenario 4, and percent change of predicted phosphorus losses of these scenarios compared to the baseline. SWAT-simulated average phosphorus loss, kg Scenarios Particulate phosphorus Soluble phosphorus Total phosphorus (PP) (SolP) (TP) Baseline Scenario Scenario Scenario Percentage reduction compared to the baseline, % Baseline Scenario Scenario Scenario * sum of all land use losses including crop fields, forests, roads, and other lands within the study watershed.

153 136 Sol P "Kg" Soluble Phosphorus (Sol P) loss Baseline Scenario 1 Scenario 2 Scenario 4 PP "Kg" Particulate Phosphorus (PP) loss Baseline Scenario 1 Scenario 2 Scenario 4 Total Phosphorus (TP) loss Total P "Kg" Baseline Scenario 1 Scenario 2 Scenario 4 Figure 5.8: SWAT-predicted three year ( ) average particulate phosphorus (PP), soluble phosphorus (SolP) and total phosphorus (TP) losses for all scenarios simulated for the study watershed.

154 137 For Scenario 4, reductions of the PP and SolP losses were 31 and 10%, respectively, compared to the baseline scenario (Table 5.11). In Scenario 2, the reductions for PP and SolP at the watershed outlet were 9% and 11%, respectively, compared to the baseline scenario. In general, converting corn areas to grass had a positive environmental impact by reducing erosion and the associated phosphorus loss, as evidenced from the outputs of the SWAT simulations. However, slight increases in soluble phosphorus losses both at field and watershed levels were indicated as a result of the corn land use conversion to grass. This is likely due to higher water trapping capacity of grass crops compared to corn crops, which in turn increases availability of soluble phosphorus. Overall, simulated results in both Scenario 2 and Scenario 4 revealed the importance of increasing the management and yield of grass production and the conversion of corn to grass to decrease phosphorus losses. However, an elevated level of nitrogen loss in runoff was observed due to increased application of nitrogen fertilizer to boost grass-forage productivity. To increase grass yield, nitrogen fertilizer at a rate of 100 kg N/ha was applied at one time for grass fields after planting. Figure 5.9 shows predicted losses on a watershed basis of organic nitrogen (Org N) and soluble runoff nitrogen (Sur Q N) for the scenarios simulated. Predicted nitrogen losses are also expected to be higher than they should be because nitrogen fertilizer of a 100 kg/ha was applied at one time. To avoid excess nitrogen loss while attaining the intended grass yield and quality, split application of nitrogen fertilizer rate during growth period is

155 138 normally recommended. Though the magnitude of the predicted nitrogen losses for the baseline scenario were not calibrated or compared against observed data, relative changes from baseline conditions provide a general indication of consequences of nitrogen losses from applying nitrogen fertilizer to increase grass production. Reductions in Org N losses for Scenario 2 and Scenario 4, as shown in Figure 5.9, were likely due to reduced sediment losses which carry the organic nitrogen. On the other hand, increased nitrogen loss (~38%) with the surface runoff was mainly due to the increased nitrogen fertilizer application to increase grass production. More generally, adoption of increased grass production in order to control phosphorus loss needs to be complemented with consideration of nitrogen management. 300 Nitrogen loss Nitrogen loss "Kg" Baseline Scenario 1 Scenario 2 Scenario 4 Org N Sur Q N Figure 5.9: Watershed level: SWAT model-predicted losses of organic nitrogen (Org N) and soluble nitrogen in runoff (Sur Q N).

156 Impacts of PFM on Soil-phosphorus For all scenarios simulated, SWAT-predicted phosphorus uptake by crops within the study watershed are presented in Table Also, predicted movement of phosphorus between the mineral pools and the amount of soil phosphorus in various pools are illustrated in Figure 5.10, Figure 5.11, and Figure As shown in Table 5.12, total phosphorus uptake was increased substantially in agricultural fields for Scenario 2 as a result of an increased phosphorus uptake by highyielding grasses. Such an increase in crop phosphorus uptake can, over the long term, aid in reducing excess soil phosphorus. Table 5.12: SWAT-simulated average phosphorus uptake of crops in the study watershed during the period of 1993 to Total phosphorus uptake by crops, kg Baseline Scenario 1 Scenario 2 Scenario 4 Crop area, 71 ha Watershed total phosphorus uptake, kg Total watershed area (163 ha) Figure 5.10 and Figure 5.11 show phosphorus movement between the mineral pools for all the scenarios simulated, averaged over the three year period ( ) at field and watershed scales. For the field-level baseline condition, the magnitudes of movement from labile (solution) to active and from active to stable pools for corn and grass land

157 140 uses were larger than those of pasture and forest land uses. This was deemed reasonable as most of the phosphorus application (manure or fertilizer) occurs in corn and grass land uses. As shown in Figure 5.10, baseline values for forested land use were negligible, implying slower transformation of phosphorus pools.

158 141 movement of phosphorus between pools "kg/ha" Corn labile to active active to stable Baseline Scenario 1 Scenario 2 Scenario 4 m ovem ent of phosphorus between pools "kg/ha" Grass labile to active active to stable Baseline Scenario 1 Scenario 2 Scenario 4 m ovem ent of phosphorus between pools "kg/ha" Pasture labile to active active to stable Baseline Scenario 1 Scenario 2 Scenario 4 movement of phosphorus between pools "kg/ha" Forest labile to active active to stable Baseline Scenario 1 Scenario 2 Scenario 4 Figure 5.10: Field-level SWAT-predicted movement of phosphorus from the labile to the active and from the active to the stable pools for the various land uses averaged over three years (1993 to 1995).

159 142 Phosphorus movement between pools "kg/ha" labile to active active to stable Baseline Scenario 1 Scenario 2 Scenario 43 Figure 5.11: Watershed scale SWAT-predicted movement of phosphorus from the labile to the active and from the active to the stable pools averaged over three years (1993 to 1995). Comparing to baseline scenario, a decline of both amounts of phosphorus moved from labile to active and from active to stable pools occurred in all agricultural fields (corn, grass, and pasture). This was expected because the amount of phosphorus in the manure applied was lower in Scenario 1 than the baseline scenario. For Scenario 2, the movement of soil phosphorus, mainly from the labile to active pools, was affected only for grasses with increased productivity. This was likely due to the increased phosphorus uptake by the higher-yielding grass, resulting in lowering the movement of phosphorus from labile to active pools. To demonstrate the effect of crop phosphorus uptake by the increased grass production, the amounts of soil phosphorus in

160 the various phosphorus-mineral pools are presented in Figure Data for the various phosphorus pools shown represent the first year of simulation. 143 Figure 5.12 shows an increased phosphorus depletion of the active and labile (solution) pools for Scenario 2 compared to Scenario 1 and the baseline scenario. For the growing season the reductions are 10 mg/kg for active and 5 mg/kg for labile pools. The annual percent reduction of active and solution phosphorus pools, as predicted by the SWAT model, were 7 and 8 %, respectively. This appreciable decrease in field-level soil phosphorus (particularly for solution and active phosphorus pools) during the growing season indicates increased soil phosphorus removal by the improved grass-forage productivity. For Scenario 4, a further decrease in soil phosphorus (particularly for solution and active phosphorus pools) was observed for fields originally in corn production. This was due to the replacement of the corn crop with higher yielding grass, which required greater phosphorus uptake.

161 144 Active Pool P (mg/kg) Baseline Scenario 1 Scenario Julian date 50 Solution Pool P (mg/kg) Baseline Scenario 1 Scenario Julian date 500 Stable Pool P (mg/kg) Baseline Scenario 1 Scenario Julian date Figure 5.12: SWAT predicted amounts of P in the mineral pools(active, solution (labile) and stable), for the 1993 simulation year for grass fields of Baseline, Scenario 1, and Scenario 2.

162 Summary and Conclusions Major concerns remain regarding continuing phosphorus inputs to the Cannonsville Reservoir, a drinking water supply to New York City. Much of this problem is believed to be exacerbated by 1) the persistent problem of phosphorus accumulation in soils and 2) erosion and associated phosphorus loss from corn silage land use within the CRW dairy farms. The continuous phosphorus build-up in the soils is due to farm phosphorus imports (primarily in purchased feed supplement and commercial fertilizer) exceeding exports (products sold off-farm or environmental losses). Addressing CRW's phosphorus imbalance problems while maintaining the economic viability of the farms necessitates farm-level management alterations to reduce the farm phosphorus surpluses and off-farm phosphorus losses. Hence, the initiation of the ongoing PFM farm program within the CRW and the present interest in evaluating the effectiveness of the program. This study successfully represented PFM-based farm plans in the Soil and Water Assessment Tool (SWAT) model for the purpose of evaluating the effectiveness of these farm strategies on the water quality at a watershed scale. The SWAT model representation of PFM farm plans was carried out in a single-farm watershed. This single-farm watershed allowed for the proposed PFM-based plan to be easily represented in the watershed for the purpose of evaluating the effectiveness of PFM-based strategies with regard to achieving phosphorus loss reduction objectives.

163 146 Representing reduction of dietary phosphorus in the SWAT model required adjustment of the manure phosphorus concentration in the baseline SWAT representation. In this study, manure phosphorus reduction as a result of dietary phosphorus change was derived from simulation results from IFSM applied to the same farm (see Chapter 4). To attain an increased grass yield for the farm watershed in the SWAT model, nitrogen fertilizer application at a rate of 100 kg/ha was applied to grass fields. Yields of entire grass fields in crop rotation within the study watershed were increased by applying the additional nitrogen fertilizer. This required modifying the fertilization operations, amount of nitrogen fertilizer, and date of application to each grass HRU in the management files of the SWAT model. Conversion of corn land use to grass in the SWAT modeling was probably the most straightforward task. This required replacing corn silage land uses with grasses by simply changing these representations in the management files of the SWAT model. In this study, the corn area located within the study watershed was relatively small, thus all corn fields were converted to grass. However, in conditions where partial conversion of corn to grass is required, selection of corn fields to be converted to grass would be needed. Such selection would likely consider the expected yield and environmental threat associated with various fields; with selection of fields targeted to those that posed the greatest environmental threat and/or those that offered the greatest benefit in yield changes. Using this modeling approach, impacts of farm-level planned PFM strategies at the watershed outlet were studied. In addition, application of the SWAT model was useful in

164 obtaining effects of such farm plan changes on phosphorus losses for individual 147 agricultural fields. The same procedures can also be employed to multiple farms within a larger watershed to evaluate aggregated effects of different management schemes at the watershed outlet. Lowering phosphorus content of manure applied to agricultural lands (caused by reducing the dietary P rations) resulted in less off-field phosphorus loss both in soluble and particulate forms (a reduction of 14% for SolP, and 12% for PP). The reduction in fieldlevel phosphorus loss also eventually manifested its effect at the watershed level accounting for the reduction of 11% and 7% for SolP and PP, respectively. In addition, investigation of the amount of soil phosphorus for this scenario revealed a decline in the amount of phosphorus moved from labile to active and from active to stable pools for all agricultural fields (corn, grass, and pasture). This was due to reduced manure phosphorus input to the soil. Implementation of more accurate feeding of phosphorus in diets (thus lower manure phosphorus concentrations) integrated with increased productivity of grass crops resulted in a lowering of phosphorus loss, particularly in particulate form. This resulted from more crop cover and interception of rainfall energy thereby reducing the impact which generates erosion. In addition, increased crop phosphorus uptake from soil was attained by increasing grass productivity. On average, an increase in grass yield of 1.8 tonnes biomass/ha resulted in an increase in phosphorus uptake of 7.7 kg/ha on the fields undergoing increased management and yield. This was also supported by the increased

165 depletion of soil phosphorus both for the labile (10 mg/kg) and active (5 mg/kg) soil 148 phosphorus pools. The labile phosphorus pool was decreased as a result of increased plant uptake; and the active phosphorus pool was reduced due to replenishment of the labile pool. The movement of phosphorus from labile to active and from active to stable pools was also reduced. This could be interpreted as an indication of the smaller extent of soil phosphorus accumulation over a period of time. Overall, this strategy could play a great role in controlling phosphorus from its source by simultaneously lowering the soil phosphorus and thus the phosphorus availability for runoff. Adoption of this strategy, however, requires complementary management options in order to control the risk of elevated nitrogen loss due to increased application of nitrogen fertilizer used to boost the grass-forage production in model simulation. Corn land use conversion to grass with regard to its main objective of reducing erosion and associated phosphorus loss was found to be effective. This was supported by the findings of SWAT modeling results, which showed reductions in sediment and sediment-bound phosphorus losses. Depending on the size of corn fields and their locations relative to streams, application of this strategy would reduce the amount of particulate phosphorus carried by erosion from corn fields. However, results also indicated an imminent increase in soluble phosphorus loss when corn fields were converted to grass, though the corn area converted to grass in this study was small. Generally, the strategy of converting corn to grass, which controls off-field transport of phosphorus, was found to be a sound practice to control phosphorus runoff losses.

166 149 Chapter 6 Remarks on IFSM and SWAT Simulations in Evaluating PFM-Based Strategies Planned at a Farm-level 6.1 Introduction In Chapters 4 and 5, the Integrated Farm System Model (IFSM) and Soil and Water Assessment Tool (SWAT) were used to perform a comprehensive evaluation of alternative farm-system strategies for Cannonnsville Reservoir Watershed dairy farms. The alternative farm-system strategies evaluated were part of precision feed management (PFM) efforts aimed at controlling nutrient imbalances and off-farm losses while maintaining farm profitability. The IFSM was used to predict farm-level costs of production, net-returns, and nutrient balances and losses (particularly phosphorus) for the various strategies studied. Application of this whole farm model also allowed for re-designing a farm system with regard to incorporating the PFM farm planning that takes farm economic sustainability into consideration. In addition, the farm model allowed farm profitability and the offfarm phosphorus losses and farm phosphorus balances to be evaluated as a result of implementing these alternative farm plans. The SWAT model was used to represent IFSM-designed PFM strategies in a single-farm subwatershed in the CRW. Also, SWAT model was used to assess field- and watershed-

167 level environmental impacts of PFM farm plans for one of the Cannonsville Reservoir Watershed (CRW) farms studied using the IFSM model. 150 Together these models provided a more comprehensive picture of economic and environmental impacts from implementing PFM strategies. For example, IFSM provided simulation of economic and environmental effects at a farm scale, whereas SWAT enabled evaluation of environmental effects of the strategies at the watershed outlet and at the field edges. Current interest is in determining how the simulations from both IFSM and SWAT models relate and, thus, is the focus of this chapter. Discussions focus on the CRW farm (R-farm) simulated using both models. Comparisons are made on representation of the farm by both models and the IFSM- and SWAT-simulated phosphorus reductions from implementing the PFM farm plans. I also discuss the roles of both models in exploring best farm management solutions for improved water quality and sustainable farm production. 6.2 Comparison of IFSM and SWAT PFM Simulations This section analyzes model representations and simulation results of PFM farm plan strategies performed using both IFSM and SWAT models, as described in Chapter 4 and Chapter 5, respectively. With respect to the effort made in evaluating phosphorus

168 reductions achieved from implementing PFM farm plans, data representation by both models and the corresponding simulated model outputs for the R-farm are analyzed Model Representation of a Farm and PFM Farm Plans The IFSM model uses input data that are based on farm land boundaries. The IFSM model representation of a farm enterprise doesn t take into account the spatial location of fields. Also, IFSM doesn t care whether farm fields are contiguous or not. Furthermore, farm fields represented in IFSM may or may not flow to the same hydrologic drainage area. For instance, in one of the CRW farms studied in Chapter 4, about 60 percent of the farm fields were contained within a single hydrologic drainage area (the study farm watershed) and the remaining fields were scattered and drained to different subwatersheds. In IFSM, the same land uses are lumped together regardless of their location relative to drainage area. The SWAT model employs input data that are based on natural or geographic boundaries. SWAT representation of the CRW farm was done for the fields contained within the study watershed. Since the SWAT model uses a hydrologic watershed boundary, the area represented by the model may cover a single farm or a number of farms, entirely or partially. In Chapter 5, the watershed studied covered a single farm. Thus, SWAT predictions at the outlet of a watershed were the results of the single-farm management strategies practiced on fields that were located within the boundary of the study watershed.

169 152 Model input data related to fields and management in the IFSM and SWAT models have different data resolutions. Fields and management data in IFSM were in their generalized mode. That is, farm fields in the IFSM model were represented by single soil and slope values with no difference normally attributed to varying spatial locations. IFSM allowed only one management per land use category, thus fields with the same land use category were managed together as a single land use. Fields and management data in the SWAT model were relatively detailed and varied spatially within the model representation. The SWAT model allowed a watershed to be divided into subbasins based on topographic criteria followed by further subdivision of subbasins into hydrologic response units (HRUs) based on land use and soil type considerations. Due to its size, the study farm watershed was represented by a single subbasin. However, with some modification to land use naming and the land use database within SWAT, fields in the SWAT model were distinctly represented; and during the process of HRU formation, lumping of particular land uses and soil types within a subbasin were avoided. Key fieldlevel management inputs, including crop rotation, planting, harvesting, and manure application were thus specified for each HRU that represented a particular unique field. IFSM-predicted off-farm phosphorus runoff losses represent a lumped value for potential generation of phosphorus losses from crop fields of a farm as a result of farm management strategies employed. However, the predicted environmental outputs from the SWAT model represent distinct data from each field and at the outlet of the

170 watershed. SWAT model-simulated outputs from individual fields of a farm situated 153 within the watershed were aggregated to estimate the loading from the study farm. To represent a dietary phosphorus reduction strategy in IFSM, feed phosphorus content was matched to that of the NRC (2001)-recommended dietary phosphorus values. Then, using IFSM, impacts of these feed phosphorus level changes on manure phosphorus concentration, feed import, farm net-income, farm phosphorus balance, and phosphorus losses were obtained through modeling (Chapter 4). For the SWAT modeling scenario, data on manure phosphorus concentration resulting from the reduced dietary phosphorus was needed as an input to the model. Since IFSM has an animal model component that predicts the amount of manure phosphorus for varying dietary phosphorus rations, the IFSM-predicted manure phosphorus concentration resulting from the reduced dietary phosphorus was used to adjust manure phosphorus concentration input data used in the SWAT model. In IFSM, representation of an improved forage system management approach was performed by increased grass yield and quality (for example, higher CP values) and by feeding cows diets that had an increased proportion of forages relative to protein and grain concentrates. An increased grass yield and quality was achieved by applying more nitrogen fertilizer. The rate of increase of nitrogen fertilizer required for the prescribed increase in grass yield was determined by performing iterative IFSM runs. Various IFSM simulations were performed by varying the rate of nitrogen fertilizer applications. An increase of 100 kg/ha nitrogen fertilizer application was found to give the prescribed increase in grass yield (2 tonnes/ha) and quality (1.2% increase in CP content). The

171 154 quality of grass forage was determined by IFSM-predicted values for crude protein (CP) content and neutral detergent fiber (NDF). Generally, improved quality for grass forage was indicated by the increased CP values (from 18.1 to 19.3% of DM) and decreased NDF values (from 50 to 49% of DM) compared to the baseline scenarios. The rate of nitrogen fertilizer application determined by the IFSM model to achieve the proposed increase in grass yield was then used as input data to the SWAT model when representing this scenario. To represent the practice of corn land use conversion to grass in IFSM, areas of corn and grass were adjusted by simply changing the amount of area in corn and grass production. That is, for the study farm with 12 ha of corn and 63 ha of grass land, conversion of half the corn fields to grass was done by changing the area of corn to 6 ha and the area of grass to 69 ha. This simply represented a reduction in the area in corn production without any significance to the spatial location of the corn fields under consideration. Further, the amount of manure previously applied to corn (before land use conversion) was applied to the grass fields after conversion. In the SWAT model, however, corn land use conversion to grass was spatially made in the management file by replacing specific corn fields (considering spatial location) with grass in the crop rotation Predictions of IFSM and SWAT from PFM-Based Strategies IFSM provides several model outputs of a farm including crop yield and quality, feed produced, feed purchased, extra feed sold, total manure produced from the herd, manure

172 155 nutrient content (phosphorus, nitrogen, and potassium), milk produced, farm costs, netreturns or profitability, nutrient balances (including nutrient imports and exports), crop nutrient uptake, and phosphorus losses. SWAT model outputs include HRU/field, subbasin, and watershed values for surface flow, groundwater and lateral flow, crop yields, sediment loss, and nutrient (phosphorus and nitrogen) losses. For the simulation strategy of reduced phosphorus in the animal diet (Scenario 1), IFSM predicted the manure phosphorus content for varying feed phosphorus levels. Moreover, both IFSM and SWAT models were able to show the effect of lowering manure phosphorus content in reducing the off-field phosphorus losses in runoff. In addition to the field-level effects, the SWAT model predicted the environmental impacts of Scenario 1 at the outlet of the watershed. On the other hand, the IFSM, in addition to providing input data to the SWAT model, predicted impacts of the strategy on farm phosphorus balance and profitability. In general, the combined application of both models revealed economic and environmental benefits of the PFM approach at both farm and watershed scales. For the Scenario 2 simulation (involving increased productivity of grass fields through improved management, feeding cows with a high-forage diet, and adjusting the dietary phosphorus content to match the NRC recommended values), the IFSM model predicted the effects of this strategy on crop soil phosphorus uptake, nutrient phosphorus imported in feed supplements, farm phosphorus balance, farm profitability, and farm phosphorus losses. The SWAT model enabled prediction of crop soil phosphorus uptake and

173 phosphorus losses at both field and watershed levels as a result of implementation of 156 Scenario 2. The SWAT model, unlike IFSM, doesn t account for nutrient balances that are based on the difference between nutrient imports (in animal feed and fertilizer) and exports in animal and crop products. When land use conversion of corn to grass was added to Scenario 2, the IFSM model enabled the prediction of the required additional corn grain feed supplement needed to offset the energy lost from the reduced corn silage production. In addition, IFSM provided costs associated with producing and buying supplemental feeds, and data related to farm profitability, farm nutrient balance, and off-farm sediment and phosphorus losses. On the other hand, the SWAT model enabled assessment of environmental impacts related to sediment, nitrogen and phosphorus losses at field, subbasin, and watershed levels Predicted Phosphorus Losses from PFM-Based Strategies The following discussion is based on the PFM simulation results from both models as described in Chapters 4 and 5. Because of the difference in model representation between the IFSM and SWAT models, direct comparison of the outputs was not made. Instead, the relative changes of model output parameters from implementing different PFM strategies were compared.

174 157 As presented in Chapter 4, the IFSM model provided effectiveness of PFM farm plans related to economic and environmental aspects. As described in Chapter 5, SWAT predicted only the environmental aspects of the PFM farm scenarios. The following comparisons of simulated results from the two models focus on environmental aspects, mainly phosphorus losses. Data discussed in this chapter are taken from Chapters 4 and 5. Predicted impacts from IFSM and SWAT models of implementing PFM were evaluated by compiling annual particulate phosphorus (PP), soluble phosphorus (SolP) and total phosphorus (TP) losses for all agricultural fields for the simulation period. For each PFM-based scenario, predicted phosphorus reduction was calculated. The percentage of phosphorus loss reduction was determined by subtracting PFM scenario loss from the baseline scenario loss and dividing this value by the baseline scenario loss. Table 6.1 shows PFM phosphorus loss percentage reductions predicted by the IFSM and SWAT models. Table 6.2 was also included to show PFM sediment loss percentage reductions predicted by IFSM and SWAT models.

175 158 Table 6.1: Phosphorus loss reduction of precision feed management (PFM) scenarios as determined from IFSM and SWAT model simulations over the simulation period. Reduction of phosphorus loss as a % PFM Scenarios of baseline IFSM SWAT SolP 1 PP 2 TP 3 SolP PP TP Scenario 1: Dietary P reduction (translated into reduced manure P) 13% 2% 4% 13% 12% 13% Scenario 2*: Dietary P reduction + increased grass yield 18% 5% 8% 14% 15% 14% Scenario 3**: Dietary P reduction + increased grass yield + 50% corn land use conversion to grass 18% 24% 28% NA 4 NA NA Scenario 4***: Dietary P reduction + increased grass yield + 100% corn land use conversion to grass 23% 44% 39% 10% 53% 26% *grass area intensified (increased yield) in IFSM was ~50% of the total crop area of the study farm; *grass area intensified in the SWAT model was ~50% of the total crop area of the study farm watershed. ** for IFSM simulation, corn area that was converted to grass accounted for 5% of the total crop area of the study farm. ***for IFSM simulation, corn area that was converted to grass accounted for 10% of the total crop area of the study farm; For SWAT model simulation, corn area converted to grass accounts for 4% of the total crop area of the study farm watershed. 1 SolP = soluble phosphorus, 2 PP = sediment-bound phosphorus; 3 TP = total phosphorus; 4 NA= simulation was not done Table 6.2: Sediment loss reduction of precision feed management (PFM) scenarios as determined from IFSM and SWAT model simulations. PFM Scenarios Reduction of sediment loss as a % of baseline IFSM SWAT Scenario 1: Dietary P reduction (translated into reduced manure P) 0% 0% Scenario 2*: Dietary P reduction + increased grass yield 2% 7% Scenario 3**: Dietary P reduction + increased grass yield + 50% corn land use conversion to grass 38% NA 1 Scenario 4***: Dietary P reduction + increased grass yield + 100% corn land use conversion to grass 55% 58% *grass area intensified (increased yield) in IFSM was ~50% of the total crop area of the study farm; *grass area intensified in the SWAT model was ~50% of the total crop area of the study farm watershed. ** for IFSM simulation, corn area that was converted to grass accounted for 5% of the total crop area of the study farm. ***for IFSM simulation, corn area that was converted to grass accounted for 10% of the total crop area of the study farm; For SWAT model simulation, corn area converted to grass accounts for 4% of the total crop area of the study farm watershed. 1 NA= simulation was not done.

176 Scenario For the reduced dietary phosphorus scenario (Scenario 1) both IFSM and SWAT models predicted matching SolP loss reductions (13%) (Table 6.1). This was deemed reasonable because SolP in both models is predicted as a function of runoff and available phosphorus concentration in the solution pool. However, based on data on Table 6.1, the IFSM-predicted PP loss reduction was only 2% as compared to the 12 % PP loss reduction predicted by the SWAT model. Also, IFSM and SWAT TP loss reductions compared to the baseline were 4% and 13%, respectively. This discrepancy could likely be due to differences in the equations used in the models to represent manure phosphorus transport processes. The phosphorus component of IFSM is comprised of equations taken from SWAT (Neitsch et al., 2002) with an additional surface phosphorus component to represent soluble phosphorus loss directly from surface applied manures (Sedorovich et al., 2005). The surface phosphorus pool in IFSM allows availability of freely draining portion of unincorporated manure (in a solution form) for runoff before it interacts with the top soil layer and binds to soil particles. This is thought to lower the amount of soil-bound phosphorus concentration on the top soil layer, thus less availability for PP loss with runoff. This is presumed to be the reason for minimal PP loss effect obtained by IFSM for the reduced dietary phosphorus scenario (which resulted in less phosphorus in manure excreted).

177 160 In the SWAT model phosphorus component, surface applied manure is added directly to the soil phosphorus pools (organic or inorganic). When runoff events occur, prediction of PP and SolP losses are computed as a function of the respective phosphorus concentrations in these two pools in the top soil layer. Hence, changing the concentration of manure phosphorus in SWAT would affect prediction of both PP and SolP. Scenario 2 For Scenario 2, IFSM predicted an 18% SolP loss reduction compared to the baseline condition (Table 6.1). The SWAT model predicted a 14% SolP loss reduction compared to the baseline condition (Table 6.1). When compared to Scenario 1, predicted SolP loss reduction from IFSM was 5% and SWAT predicted a minimal effect (1%) on SolP loss as a result of the additional practice of simulating increased grass yield. Generally speaking, predicted SolP loss reductions from both models showed minimal effects from increasing grass yield. Based on data from Table 6.1, IFSM-predicted percent reduction of PP loss due to the Scenario 2 strategy was 5% compared to the baseline (and 3% compared to Scenario 1). SWAT-predicted percent reduction of PP loss due to the Scenario 2 strategy was 15% compared to the baseline (and 3% compared to Scenario 1). In Scenario 2, IFSM- and SWAT-predicted incremental percentage reductions of PP losses from the strategy of a dietary phosphorus reduction were comparable though sediment loss reductions predicted by the SWAT model were higher that those predicted by the IFSM. The SWAT model

178 161 predicted a higher field-level sediment loss reduction (7%) for the increased grass yield scenario compared to a reduction of 2% predicted by the IFSM for the same scenario (Table 6.2). Prediction of PP in IFSM and SWAT models is computed as a function of sedimentbound phosphorus concentration and amount of sediment loss. SWAT estimates erosion and sediment using the Modified Universal Soil Loss Equation (MUSLE; Williams, 1975). The plant cover (C) factor of this erosion prediction equation gets updated daily as a function of average annual C factor for the land cover and the amount of plant residue on the soil surface. Updating the C factor values in the SWAT model to reflect the crop cover allows prediction of the impacts of intensified grass on the amount of soil eroded and the associated phosphorus losses. Similar to SWAT, IFSM uses the MUSLE form of the soil loss equation (Williams, 1975) to predict sediment losses from crop land uses. The crop factor of the soil loss equation determines cover management conditions that are also updated daily in response to plant growth. Scenarios 3 and 4 For IFSM simulations, both Scenario 3 and Scenario 4 combined a land use management activity of converting corn land to grass land along with the Scenario 2 practices of enhanced forage and dietary phosphorus management. With an objective of reducing erosion and associated phosphorus losses in Scenario 3, 50% of the corn area on a farm was converted to grass. In Scenario 4 all corn land was converted to grass. As described in Chapter 5, for Scenario 4 of the SWAT simulation all corn fields within the study

179 watershed were converted to grass as the corn area located within the study watershed was relatively small. 162 IFSM-predicted percent reductions of SolP loss resulting from conversion of corn land use to grass were 18% and 23% for Scenario 3 and Scenario 4, respectively, compared to the baseline (Table 6.1). SWAT-predicted percent reduction of SolP loss for Scenario 3 was 10% compared to the baseline scenario (Table 6.1). Compared to Scenario 2, IFSMpredicted SolP loss reduction resulting from conversion of corn land use to grass showed a slight increase (from 0% to 5%). On the contrary, SWAT-predicted SolP loss reduction under Scenario 4 was decreased compared to SolP loss reductions predicted for Scenario 2. Both the IFSM and SWAT models showed increasing trends in PP and sediment loss reductions resulting from conversion of corn land use to grass (Table 6.1 and Table 6.2). In the SWAT modeling, converting all corn acreage (accounting for 4% of the total crop lands) to grass resulted in a 38% PP loss reduction (53% vs.15%) compared to Scenario 2. In the IFSM modeling, converting 50% corn acreage (accounting for 5% of the total agricultural crop lands) to grass resulted in a 19% PP loss reduction compared to Scenario 2; while converting all corn land (accounting for 10% of the total crop land) to grass resulted in a 39% PP loss reduction compared to Scenario 2. When all corn fields were converted to grass, sediment loss reduction from IFSM and SWAT models were comparable (Table 6.2).

180 6.3 General Comments on Applicability of IFSM and SWAT in Evaluating PFM- Based Strategies Planned at a Farm-By-Farm Level The Role of the IFSM Model in Evaluating PFM-Based Strategies Planned at a Farm Level To successfully assess effectiveness of farm plan strategies, a comprehensive evaluation tool is needed to quantify the impacts of implementation of farm management changes on milk production, farm profitability, farm-level nutrient imports and exports, and nutrient losses. For this purpose, the IFSM model can be a helpful tool, particularly for efforts to explore farm system options needed to address the root cause of phosphorus build-up on individual farms (phosphorus imbalance). The IFSM model can be used as an aid in designing farm management systems that reduce phosphorus imbalances while maintaining or increasing the profitability of farms and controlling off-farm nutrient losses to the environment. In Chapter 4, different PFM-based farm plans were evaluated in relation to their effectiveness in reducing phosphorus imbalances on a farm while maintaining the profitability of farm enterprises. The analysis also included off-farm phosphorus loss predictions. Though more validation of the phosphorus loss component of the IFSM is recommended, as described in Chapter 4, the relative changes of phosphorus losses resulting from implementation of management changes were presumed to provide information on the effects of farm strategies on relative phosphorus losses to the environment.

181 164 IFSM can also be extended to assess future innovative farm plan options (in addition to PFM-based farm plans) that have the potential for farm sustainability with less harm to the environment. This is possible because the IFSM model was developed to predict relative environmental and economic benefits of various management strategies at a farm scale. Also, the model has been successfully used to evaluate both economic and environmental status of farming systems in the Northeast U.S. (Rotz et al., 2002). As far as farm-level planning of individual farms is concerned, the IFSM model can perform comprehensive analyses of farm strategies that include both economic and environmental aspects. For a watershed with multiple farms, IFSM can be applied individually on each farm located within the watershed, and results from individual farms can be aggregated to assess effects at a watershed scale. In this regard, the IFSM predictions related to phosphorus imbalance and profitability of farms are thought to provide representation of phosphorus imbalances and profits aggregated from multiple farms. With more validation of the IFSM off-field phosphorus loss predictions, predictions of phosphorus losses aggregated from multiple farms are expected to reveal the potential effects of varying management schemes of individual farms on the water quality of the watershed of interest. This is particularly true for smaller watersheds with insignificant channel processes because IFSM predictions of phosphorus losses only represent potential generation of upland nutrient losses from agricultural crop fields without consideration of sediment or nutrient transport to streams. Also, with IFSM runs for multiple farms within a watershed of interest, effects of changing land use other than agricultural crops can not

182 be studied as the model, at its present state, can only represent agricultural crops of a 165 farm. For example, impacts of land use changes of agricultural crop fields to residential or other land uses can not be assessed using IFSM. When fields of a farm are located in different hydrologic drainage units, predicted offfarm phosphorus losses provided by IFSM do not drain to the same stream or outlet. In such cases, the effects of different fields can only be seen in separate hydrological subbasins. Hence, IFSM-predicted water quality related effects (such as phosphorus losses) from implementing farm plan strategies are hard to interpret on an actual landscape basis and on specific stream segments. That is, from IFSM-predicted data it is hard to know how much of the off-farm pollutant losses will actually affect the water quality of a certain stream or stream segment. In all cases, however, when a farm-level planning approach is used, the IFSM model can be used to provide data and information on farm nutrient balances, farm profitability, and nutrient losses from farm crop fields under varying farm management practices. Data obtained from the IFSM model can also be used to complement water quality models with farm nutrient balance and economic data for a more comprehensive assessment of farm plan strategies at a watershed scale The Role of the SWAT Model in Evaluating Farm Management Strategies Planned at Farm Level When a farm-level planning approach is desired, the SWAT model can be used to determine the environmental impacts at field, farm, and watershed levels imposed by

183 166 farm plan changes. Because the SWAT model has been widely recognized for studying the impacts of land management practices on water, sediment, nitrogen and phosphorus losses, the model can be used to study the impacts of farm management plans at the outlet of the watershed. The SWAT model predicts the impact of field specific management on water resources; however, the model doesn t include the farm economics nor does it include farm system in simulating system processes. Unlike IFSM, the SWAT model requires input related to what happens to actual fields on the landscape when change in a farm system occurs. Changes in a farm system may involve manipulation of animal diets (including dietary phosphorus changes), increasing productivity of on-farm forages, and increasing use of on-farm produced forages. For example, as described in Chapter 5, the SWAT model was successfully used to evaluate environmental impacts of farm plan strategies for a single-farm watershed. However, it was necessary to obtain some data related to how manipulation of phosphorus in animal diets translates into the produced manure phosphorus content from the IFSM model. Hence, input data required by the SWAT model related to phosphorus content of manure applied to fields had to be obtained outside of the SWAT model. When field specific management data are available, the SWAT model predicts the transport of the water, sediment, and nutrients for assessment of the impact of these land use management practices on the quality of water resources. In Chapter 5, because field boundaries were used in SWAT representation in place of a general land use mapping,

184 HRUs coincided with field boundaries. This made studying the effects of management changes on individual fields possible. 167 Also, when the SWAT model is used jointly with the IFSM model, the SWAT model can complement the IFSM with detailed field level predictions. Also, the SWAT model provides watershed-level flow and nutrient loss predictions that can be validated against observed (measured) flow rates and nutrient measurements, at a scale where most data are measured and/or monitored.

185 Chapter Economic and Environmental Impacts of Precision Feed Management in the Town Brook Watershed, NY 7.1 Introduction The Town Brook Watershed (TBW) is located in Delaware County, New York, and is part of the larger Cannonsville Reservoir Watershed (CRW). As with the larger CRW, water quality in the TBW is at risk due to phosphorus loading in runoff from agricultural fields receiving manure and fertilizer (Cerosaletti, 2006). Ongoing efforts in controlling phosphorus losses include implementation of best management practices (BMPs) on a farm-by-farm basis. However, these BMPs, while expected to be effective in controlling off-field phosphorus losses, do not address the ultimate cause of phosphorus imbalance problems within the dairy farm operations. And, this excess phosphorus imbalance (more off-farm import of phosphorus than off-farm export) is believed to be the main cause for accumulation of phosphorus in soils. As part of the current CRW precision feed management (PFM) efforts, personnel of the Cornell University Cooperative Extension (CCE) of Delaware County, NY, are planning the implementation of PFM strategies on TBW farms. Prior to implementation of these PFM management changes, extension personnel and farm operators are interested in assessing quantitatively the effectiveness of potential PFM strategies. In the previous chapters several aspects related to this quantitative assessment included: 1) successful representation of management conditions for the CRW farms in the

186 Integrated Farm System Model (IFSM) and development of reference data for the 169 before management change condition (hereafter referred to as baseline scenario ); 2) representation of alternative precision feed and forage management systems as proposed by CCE planners for the CRW farms into IFSM and assessment of economic and environmental impacts of implementation of these systems as compared to the baseline condition; 3) representation of IFSM farm-level designs of these precision feed and forage management systems into the Soil and Water Assessment Tool (SWAT) model for a single-farm watershed that encompasses one of the CRW farms; 4) evaluation of these PFM farm-level plans as represented in the SWAT model at a watershed level, and 5) comparison of effectiveness of IFSM- and SWAT-predicted PFM farm plan scenarios relative to phosphorus losses in the baseline conditions. Environmental consequences at the watershed outlet of a single-farm PFM scenario were assessed and discussed in Chapter 5. As suggested, the same procedure developed and described therein could be used to extend the representation of multiple farms under a variety of management schemes to a larger-sized watershed that encompasses these farms. Thus, aggregated environmental impacts of all farms located within the watershed of interest could be assessed. Though watershed-level evaluation of these management changes was included in the farm-level evaluation, these planning and evaluation processes were approached from a perspective of farm-level planning. Thus, when such a study is applied to multiple farms within a watershed, the environmental effects of different management options relative to the production system and viability of each farm can be studied. Exploration and assessment of BMPs for sustainable farming, considering

187 environmental and economic consequences, could thus be explored for all farms within a watershed. 170 Watershed-oriented management approaches are needed for addressing water quality impairments. Federal regulations, such as the Clean Water Act, recognize watershed groups and their role in identifying impaired water bodies and in developing watershed plans, including the development of Total Maximum Daily Load (TMDL) programs to meet national water quality standards. With watershed-oriented management approaches, it is possible using various available methods to objectively identify pollutant sources (fields, farms) and allocate resources and time to source areas where changes will have the greatest effect on water quality. Considering the role of watershed-oriented management planning, this chapter evaluates the implementation of various PFM plans designed from the perspective of a watershed-scale implementation. Consequences of PFM implementation on individual farm income, animal feed mix, phosphorus balance, and phosphorus loss were assessed. The effects of PFM-based alternative management plans (planned at a watershed-level) on specific farm profitability, phosphorus balances, and phosphorus losses were assessed in addition to watershed-level evaluation of PFMbased plans. This study employed farm- and watershed-based modeling to assist in planning, implementing, and evaluating the PFM program. The SWAT and IFSM models were used jointly to quantify and evaluate the economic and environmental impacts of various alternative farming system scenarios that have the potential to reduce phosphorus losses

188 at minimal cost. The SWAT model was applied to the TBW to assess the impacts of 171 PFM-based strategies at the watershed level. In addition, IFSM was applied to assess the consequences of PFM implementations to one of the farms within the TBW. The application of these models to a watershed experiencing phosphorus-related water quality problems provided a means of evaluating and identifying cost-effective phosphorus loss reduction strategies at a watershed scale, while simultaneously assessing the consequences of implementation of the various alternative farm management systems to individual farms.

189 7.2 Materials and Methods Study Area Study watershed: The TBW, part of the larger CRW was selected for evaluation of PFM strategies suggested by CCE planners, as documented in the previous chapters. The TBW was suitable for this study for several reasons: 1) Efforts to implement the PFM strategies on individual farms is presently underway; in fact, two thirds of the TBW farms are enrolled in the PFM program (Cerosaletti, 2006). Efforts are also underway to achieve full enrollment of farms within the TBW. 2) Data including SWAT representation of the watershed, monitored flow, and water quality are available from previous studies. A SWAT representation of TBW by Gitau (2003), for example, was utilized to represent the baseline condition for the farms for conditions prior to PFM adoption. 3) Finally, the TBW is a United States Department of Agriculture Agricultural Research Service (USDA-ARS) Conservation Efforts Assessment Program Benchmark watershed, which means the TBW is being studied to provide basic research on effectiveness of conservation practices and to serve as a framework for evaluating and improving performance of assessment models. The TBW (37 km 2 ) located in Delaware County, New York (refer to Figure 7.1), is a tributary of the West Branch Delaware River that feeds to the CRW. About 50 percent of the study watershed is occupied by agriculture (mostly dairy farms) and about 49 percent is comprised of forests (Table 7.1).

190 173 Cannonsville Reservoir Watershed TownBrook Watershed ALFA CSIL FR SD FR SE HAY PAST URLD UTRN WETN N # 0 20 Km 0 3 Kilometers Crop fields of the study farm N Figure 7.1: Map showing the location of Cannonsville Reservoir Watershed (CRW), the Town Brook Watershed (TBW), land use distribution, and crop fields for the study farm (data obtained from USDA/ARS, 2004).