Effectiveness of BMPs in Controlling Nonpoint Source Pollutant Losses from a CEAP Watershed

Transcription

1 Effectiveness of BMPs in Controlling Nonpoint Source Pollutant Losses from a CEAP Watershed Indrajeet Chaubey Purdue University CEAP Project No

2 Modeling team Model setup and execution Indrajeet Chaubey, Margaret Gitau, Lichi Chiang, Chetan Maringanti, Sayeed Mohammed Input data and review of results J. Pennington, J. Popp, E. Gbur, G. Rodriguez SWAT Team input Jeff Arnold, Nancy Sammons, Mike White

3 Study Location: Lincoln Lake Watershed Lincoln Lake Watershed

4 Key watershed management questions How do various combinations of BMPs improve water quality at various spatial and temporal scales? How does uncertainty in future weather and climate conditions affect BMP performance? In a dynamic watershed, how can we differentiate land use impacts from BMP impacts on water quality? How can we optimize the selection and placement of BMPs in the watershed?

5 Modeling challenges Land use: Pasture 36% Forest 39% Urban 12% Available data at various spatial and temporal resolution Watershed is very dynamic Various externalities have resulted in dynamics watershed management

6 BMP scenarios modeled in SWAT 3 BMP categories(total of 171 BMP scenario combinations) baseline => 172 BMP scenarios spring summer fall Nutrient management poultry litter alumamended litter No application 1, 1.5, 2 ton/acre 1, 1.5, 2 ton/acre 2, 2.5, 3 ton/acre grazing and pasture management no grazing optimum grazing over grazing Buffer width 0 m 15 m 30 m

7 Weather data Period of interest Based on the expected life of BMP and minimum data needed for economic analyses Weather data generated using WXGEN program based on historical measured data in the watershed 250 weather realizations used in this study Weather data were same for all 172 BMP scenarios

8 Computational challenges Number of runs 172 BMP scenarios, 250 weather realization data/bmp scenario = 43,000 SWAT runs Time/run = 8-10 mins (LINUX); 5,700 CPU hrs Space temporary storage for output Output files 1*59 Mb, 3*2.5 Mb per run Total 67.5 Mb/run minimum 2.5 Tb

9 Approaches - Condor Modified, configured SWAT to run on TeraGrid Condor network and produce minimum required output for each run Prepared input files for batch runs weather, management, options Developed scripts and postprocessing tools to handle model runs and outputs.

10 Baseline model results: flow and sediment 2004 model as the baseline model Detailed measured watershed data used to develop the model Model is calibrated minimally as not to change the parameter values drastically from the baseline

11 Annual N and P losses from pastures Box: 1 st and 3 rd quartile Whisker: 10 th and 90 th percentile Buffer width 0m 30m 15m

12 Annual N and P losses from watershed Buffer width 0m 30m 15m

13 BMP Interactions: TN Dash line is with no litter application Fall litter application can result in the greatest TN loss, except with optimum grazing. Overall, TN loss increases with the intensity of grazing.

14 BMP Rankings P<.0001 logts logtn logsolp logtp R-square Rank source (logts) % source (logtn) % source (logtp) % 1 buffer grazing buffer buffer*grazing 0.30 buffer 8.61 grazing grazing 0.19 time 5.00 buffer*grazing time*grazing 0.08 time*grazing 2.71 amount time*buffer*grazing 0.05 amount 1.98 type amount 0.04 buffer*grazing 1.60 amount*buffer type*buffer*grazing 0.03 type 0.77 type*buffer type*grazing 0.02 amount*grazing 0.58 type*grazing type 0.02 time*buffer 0.41 amount*grazing amount*buffer 0.02 time*buffer*grazing 0.24 type*buffer*grazing 0.01

15 How can we differentiate land use change effects from BMP effects? Hypothesis: In agricultural watersheds, land use changes can mask the water quality improvements from the BMPs implemented in the watershed SWAT Code modified to change land use dynamically with time

16 Challenge: Dynamic land use in the watershed Urban area has increased four fold (from 3% to 12% Pasture area has decreased steadily over time How to account for the dynamic land use conditions in the watershed?

17 BMPs implemented ( ) Percentage of total pasture areas 0% 1% 5% 9% 11% 53%

18 Pre-process of HRU fraction files as SWAT input Original HRU fraction files: 1992, 1994, 1996, 1999, 2001, 2004 Matched by 2004 HRU distribution (1461 HRUs): subbasin_soil_landuse Regrouped by soil and land use Minimum HRU fraction = Modified HRU fraction files: 1992, 1994, 1996, 1999, 2001, 2004 (all have 1461 HRUs) File1.dat : 1992 HRU fraction File2.dat : 1994 HRU fraction File6.dat : 2004 HRU fraction LUD.DAT file: 1990/1/1 reads file1.dat 1994/1/1 reads file2.dat 2004/1/1 reads file6.dat SWAT model

19 Scenarios: Matrix of land use and management practice rotation Simulation period: Warming up year: Measured: at Upper Moores Creek BMP impacts: (dynamic_18rot) (dynamic_1rot) LUC impacts: (dynamic_1rot) (constant_1rot) Constant land use (1992) Dynamic land use ( ) No BMP: Constant management practice (1-year rotation) Const_1rot Dynamic_1rot With BMP: Dynamic management practice (18-year rotation) Const_18rot Dynamic_18rot

20 Simulated TP(kg/ha) Preliminary results Simulated TP with Constant Land Use Measured TP(kg/ha) y = x R² = Series1 Linear (Series1) TP (kg/ha) Jan-97 Jul-97 Jan-98 Jul-98 Jan-99 Measured: at Upper Moores Creek Simulated_constant: with 2004 land use data Simulated_actual: with HRU fraction changes data TP in Jul-99 Jan-00 Jul-00 Jan-01 Jul-01 Jan-02 Jul-02 Jan-03 Jul-03 Jan-04 Jul-04 Jan-05 measured TP simulated_const TP Jul-05 Jan-06 Jul-06 Jan-07 Jul-07 Simulated TP (kg/ha) Simulated TP with Dynamic Land Use Measured TP(kg/ha) y = x R² = Series1 Linear (Series1) TP (kg/ha) Jan-97 Jul-97 Jan-98 Jul-98 TP in Jan-99 Jul-99 Jan-00 Jul-00 Jan-01 Jul-01 Jan-02 Jul-02 Jan-03 Jul-03 Jan-04 Jul-04 measured TP simulated_actual TP Jan-05 Jul-05 Jan-06 Jul-06 Jan-07 Jul-07

21 Impacts of LUC and BMPs on water quality throughout the period : Positive value means degradation; negative value means improvement in water quality The negative impact of LUC on TS was greater than that of BMP LUC had positive impact on TN and TP due to loss of pasture land LUC and BMP have contrastive impacts on TP watershed scale

22 BMP optimization SWAT used to get the initial sediment loads A BMP tool is used to obtain BMP efficiencies in reducing pollutant load Multi-Objective GA with objective functions to: 1) Maximize pollution reduction in the watershed 2) Maximize the returns ($) from the watershed from BMP implementation An allele set with all possible BMPs that can be applied in a field BMP cost Pareto-Optimal Solution

23 Optimization: Objective function Objective 1: Total Pollutant Reduction Objective 2: Total Cost f ( x) = TP * (1 BMPR ) * Area HRU = i g( x) = Cost *( NCI )* Area HRU = i i i i BMPi i i

24

25 BMP Optimization TP Results

26 Location of Optimal BMP Placement

27 Conclusions Grazing management followed by timing of litter application are important non-structural BMPs for reducing nutrient losses from pasture. Buffer strip is the most important structural BMP for reducing nutrient and sediment losses from pasture. Weather variability can result in significant uncertainty in BMP performance BMP optimization resulted in significantly reduced losses of sediment and nutrients from the watershed at a marginal cost increase Condor enables a large number of SWAT runs needed for comprehensive BMP evaluation

28 Thank you