The Fertilizer Forecaster: guiding short-term decisions in nutrient management

Transcription

1 The Fertilizer Forecaster: guiding short-term decisions in nutrient management Anthony Buda, Peter Kleinman, Ray Bryant, and Gordon Folmar USDA-ARS Patrick Drohan, Lauren Vitko, Jasmeet Lamba, Doug Beegle, Doug Miller, Brian Bills, and Paul Knight Penn State University This project was supported by Agriculture and Food Research Initiative Competitive Grant number , USDA National Institute of Food and Agriculture.

2 Partners Middle Atlantic River Forecast Center STATE CONSERVATION COMMISSION

3 Today s presentation 1. Runoff generation and nutrient management 2. Fertilizer Forecaster runoff prediction approaches 3. Fertilizer Forecaster timeline and progress

4 The need for daily recommendations Poor timing has consequences Applying fertilizers and manures at the wrong time increases the risk of surface water contamination. CDT/Nabil K. Mark Thursday, Feb. 12, 2009 Thousands of fish killed - Owner blames manure runoff from farm Centre Daily Times Site assessment tools are currently seasonal (e.g., P Index), but daily recommendations would be helpful. Dissolved reactive P in runoff (mg/l) no dairy manure 20 kg P/ha (P based) 70 kg P/ha (N based) 0 2 days 9 days Time since surface application

5 Weather forecasts can now support finer scale modeling Short range ensemble forecast (SREF) km grid, forecasts out to 3.5 d High resolution ensemble forecast (HREF) 4 4 km grid, forecasts out to 2 d

6 The Fertilizer Forecaster short-term runoff prediction for nutrient management

7 The Fertilizer Forecaster Project Empirical runoff prediction models Development of a web-based forecasting platform to guide field fertilizer and manure management decisions in the Chesapeake Bay Watershed. USDA-ARS, University Park, PA. Evaluate three runoff forecasting models Hydrologically Sensitive Areas model NOAA s River Forecast Center model Work with a project advisory team to select one model (or a suite of models) Test web-based system to identify when and where to apply fertilizers and manures

8 Project watersheds Anderson Creek Watershed Anderson Creek Allegheny Plateau Spring Creek Watershed / Rock Springs Spring Creek Ridge & Valley Piedmont Mahantango Creek Conewago Creek Mahantango Creek Watershed Coastal Plain Conewago Creek Watershed

9 Keeping it simple Empirical runoff prediction models Runoff = 1 No runoff = 0 vs. Wash off Wash in Logistic regression A useful technique for predicting the probability of a binary outcome (e.g., runoff occurrence).

10 Performance of logistic models depend upon forecasts and soil mapping of restrictive layers Moderately well to poorly drained Somewhat poorly drained Poorly-drained soil matrix. Photo: S. Dadio prism face Alluvial parent material Anthropogenic erosion contact. Colluvial parent material; clay or fragic properties; discontinuity present. Colluvial parent materials lower in the landscape; fragipan present.

11 Runoff generation mechanisms Three basic mechanisms that Rainfall rate Infiltration excess (IE) overland flow generate surface runoff runoff Infiltration rate Partial Infiltration area excess IE overland (IE) runoff flow runoff Rainfall Rainfall rate rate Rainfall Saturation excess (SE) overland runoff flow runoff Infiltration rate Infiltration rate Rainfall Rainfall Perched SE overland runoff flow return flow Rainfall Rainfall runoff return return flow flow Rainfall Rainfall

12 Well drained soils Logistic regression models of Probability of infiltration excess runoff runoff occurrence Rainfall intensity (cm hr -1 ) No fragipan Poorly drained soils Probability of saturation excess runoff Soil moisture (%) Fragipan PROS: High spatial resolution. Predicts wash in and wash off potential. CONS: Requires monitoring data to build models and predictions may range widely across sites.

13 Fragipan soils and zones of saturation Late August 2003 no fragipan fragipan

14 Fragipan soils and zones of saturation Late October 2003 no fragipan fragipan

15 Penn State restrictive layer mapping Soil descriptions and high resolution elevation data (LiDAR) are used to digitally model and map the spatial extent of soils possessing a restrictive layer (e.g., a fragipan). The model evaluated: slope, aspect, flow accumulation, curvature, parent material, several wetness indices. Investigation via ground penetrating radar transects and targeted soil sampling. Model of soils possessing restrictive layers in the Mattern watershed. Restrictive layers mapped at two depths 0.65 m 0.85 m

16 Hillslope monitoring supports model testing and development Plan view Cross-section view Site view (edge of field) Soil moisture sensor Well drained site Surface runoff Slotted pipe Piezometer nest Surface runoff routed to USDA-ARS tipping bucket Poorly drained site Surface runoff 10 ft 10 ft Climate station Surface runoff routed to USDA-ARS tipping bucket 10ft Slotted pipe

17 Tying the restrictive layer model to soil saturation and runoff occurrence Comparing wet boot saturation to electromagnetic induction (EMI) surveys in the Mattern watershed Wet boot EMI Mattern Depth to restrictive layer 0.65 m 0.85 m

18 Applying the soil restrictive layer model to runoff forecasting Mattern Watershed FD-36 Watershed Soils prone to infiltration excess Soils prone to saturation excess

19 Another field scale approach Cornell s Hydrologically Sensitive Area Model Daily water balance model and topographic wetness index PROS: High spatial resolution. Model less likely to require site-specific modification. CONS: Only forecasts antecedent soil saturation conditions and does not predict runoff.

20 Wisconsin Manure Advisory System a watershed scale approach Runoff risk forecasts are similar to fire danger forecasts issued by the U.S. Forest Service (e.g., Smokey the Bear) High risk of runoff North Central River Forecast Center

21 Wisconsin s Manure Mgmt. Advisory System The decision support tool leverages state-of-the-art weather and hydrologic forecasting from the North Central River Forecast Center. NOAA s Sacramento Soil Moisture Accounting Model (SAC-SMA) is run for all watersheds in Wisconsin Output from the SAC-SMA model is sent to the University of Wisconsin where it is converted to maps of runoff risk PROS: Employs an existing forecasting system. Little is needed to adapt to nutrient management. CONS: Scale does not distinguish between fields on a farm and only runoff potential is forecast.

22 Recent work with NOAA s MARFC exploring probabilistic runoff forecasts at finer scales Short-Range Ensemble Forecast (SREF) Sacramento Soil Moisture Accounting (SAC-SMA) Model Six models, 21 members Eta_BMJ (3 members) Eta_KF (3 members) RSM_SAS (3 members) RSM_RAS (3 members) NMM (5 members) ARW (5 members) 21 forecasts of surface runoff + interflow

23 Using SREF as input to SAC-SMA trial in Mahantango Creek SAC-SMA forecast grid points in east-central Pennsylvania WE forecast points overlaying Mahantango Creek SAC-SMA watershed outlets on Mahantango Creek USGS GK-27

24 Using SREF as input to SAC-SMA trial in Mahantango Creek Point 207 (centroid of WE-38) 7 out of 21 members predict runoff, yielding a 33.3% chance of runoff in WE-38.

25 Using SREF as input to SAC-SMA trial in Mahantango Creek In WE-38, there was a slight (35%) chance of surface runoff Rainfall (in) on May 28, About 0.25 in (or 6.4 mm) of rain fell in WE-38, with as much as 0.5 in (12.7 mm) elsewhere in Mahantango Creek. Probability of runoff occurrence (%) <

26 Runoff coefficient Creating meaningful runoff risk thresholds moving beyond simple runoff occurrence forecasts Runoff No runoff vs Moisture thresholds θ 0.36 m 3 m Vol. soil moisture (m 3 m -3 ) & Runoff contributing areas Small storm Large storm

27 Calculating runoff contributing areas borrowing a concept from the Pennsylvania Phosphorus Index WE-38 Watershed (7.3 km 2 ) Runoff volume (m 3 ) Precipitation depth (m) Runoff contributing area (m 2 ) Gburek et al., 1999; 2002

28 Runoff contributing distances derived from contributing area and stream length WE-38 Watershed (7.3 km 2 ) Runoff contributing area (m 2 ) Stream length (m) WE-38 stream length = 11,250 m Contributing distance (m) Gburek et al., 1999; 2002

29 Soil restrictive layers drive runoff generation inferring the contributing distance of saturation excess runoff Mattern Watershed runoff 22,000 L 90 L well drained soil poorly drained soil (restrictive layer) Soils prone to infiltration excess Soils prone to saturation excess On average, soil restrictive layers extend about 75 m from the stream, suggesting a saturation excess contributing distance of about 150 m.

30 Does SAC-SMA capture daily saturation patterns? comparing modeled and measured moisture in WE-38 Mean volumetric water content (m 3 m -3 ) Sacramento (SAC) Soil Moisture Accounting (SMA) model Volumetric water content (top 25 cm) versus SAC-SMA saturation ratios (top 25 cm) Vol. soil moisture = 0.17 (saturation ratio) r 2 = 0.70; p < SAC-SMA expresses soil moisture as a saturation ratio Saturation ratio = θ θr θs θr, where θ = volumetric water content θr = permanent wilting point θs = porosity SAC-SMA saturation ratio Saturation ratios predicted by the SAC-SMA model are a good proxy for surface (i.e., top 25 cm) moisture conditions in the WE-38 watershed.

31 Runoff risk thresholds proposed decision rules for runoff risk Low Risk SAC-SMA saturation ratio < 0.6 SAC-SMA interflow + surface runoff 0 cfs Moderate Risk SAC-SMA saturation ratio > 0.6 SAC-SMA interflow + surface runoff > 0 cfs and < 17.5 cfs High Risk T.S. Lee at WE-38 SAC-SMA saturation ratio > 0.6 SAC-SMA interflow + surface runoff > 17.5 cfs

32 The Fertilizer Forecaster a vision for what the prototype might look like Watershed scale view Runoff risk Low Mod High Zoomed in view The zoomed in view would show the extent of the moderate runoff risk buffer, defining areas expected to be hydrologically connected to the stream. A hypothetical runoff risk forecast showing a low to moderate runoff risk for the km forecast cells that make up the Mahantango Creek Watershed.

33 Summary and next steps Soil moisture and runoff contributing area thresholds express runoff risk in terms of variable source area hydrology. Initial results suggest that these thresholds provide a meaningful and skillful representation of runoff risk. Continued corroboration of the soil restrictive layer model is needed to refine areas prone to saturation excess runoff. A prototype of the runoff risk tool based on these thresholds is in development and will soon be available for real-time testing.

34 Report benefits and drawbacks of different modeling approaches Report results to Advisory Panel in Spring 2015 Publish salient findings in Journal of Soil and Water Conservation PA Conservation Comm. PA NRCS PA Dept. of Environ. Prot. PA Conservation Districts 8 farmers

35 Project timeline 2012, 2013 Monitoring and modeling 2013, 2014, 2015 Model testing and evaluation 2015, 2016 On farm evaluation 2017 Fertilizer Forecaster online Development of a web-based forecasting platform to guide field fe decisions in the Chesapeake Bay Watershed. USDA-ARS, Unive

36 Thank you For more information, contact: Anthony Buda Research Hydrologist USDA-ARS 3702 Curtin Road University Park, PA Phone: Patrick Drohan Associate Professor of Pedology Penn State, Ecosys. Sci. & Mgmt. 116 Ag. Sciences and Industries Bldg. University Park, PA Phone: