1. Economic Incentives to Improve Water Quality in Agricultural Landscapes: Some New Variations on Old Ideas Catherine L. Kling Department of Economics Center for Agricultural and Rural Development Iowa State University AAEA, Denver, 2010

2. Topics Introduction –Water quality snapshot –Current policy Policy: abatement actions, point-based Tools: evolutionary algorithms Case study: Boone River Watershed

3. Water Quality: Lakes The diverse aquatic vegetation found in the Littoral Zone of freshwater lakes and ponds. A cyanobacteria bloom in a Midwestern lake. Lakes, Reservoirs, Ponds: –42% assessed, 65% inadequate water quality to support uses –Over 11 million acres are “impaired” –Agriculture third highest source of impairment

4. Water Quality: Rivers & Streams Photos courtesy Iowa DNR Rivers and Streams: –26% assessed, 50% inadequate water quality to support designated uses –Nearly ½ million stream miles are “impaired” –Agriculture leading source of impairment (identified as cause of 22% unknown second highest)

5. Hypoxia = Dead Zone Depleted oxygen creates zones incapable of supporting most life 400 worldwide Stressed marine and estuarine systems, mass mortality and dramatic changes in the structure of marine communities (Diaz and Rosenberg, 1995). In other words ….

6. Nutrients and sediments from the Mississippi River enrich the water making it brown. Over 400 hypoxic Areas worldwide (Diaz and Rosenberg, 2008) Hypoxic Zone in the Gulf of Mexico Image courtesy of Nancy Rabalais (Louisiana Universities Marine Consortium) and can be found on the Southern Regional Water Program web site.

7. Frequency and Size: 1985 -Present www.gulfhypoxia.net

8. What abatement options exist? In field Management Practices –Reduced (no) tillage –Manure, fertilizer management/reduction –Cover crops, rotation changes –Land retirement Panoramic view of gamma grass-big blue stem planting http://www.fsa.usda.gov/Internet/FSA_Image/ia_767_15.jpg

9. What abatement options exist? Structural Practices –Buffers –Grassed Waterways –Denitrification, controlled drainage –Wetland restoration Photo courtesy Missouri NRCS

10. Efficacy and Cost of Practices Vary by –Pollutant –Field characteristics –Land use in watershed –Provision of other ecosystem services Ideally, all of these factors considered in efficient policy design

11. Agricultural Water Pollution Characteristics Nonpoint (Segerson, Shortle and Dunn, etc.) –Measurement: costly to trace nutrients to source –Randomness: stochastic events (weather) have large effect on damages Spatial Aspect (Montgomery, Baumol and Oates) –Location of release affects damages –Damages are non-separable between fields

12. Clean Water Act, 1972 States required to have water quality “standards” identifying goals for water Point sources required to obtain a discharge permit Nonpoint sources have no such obligations –Total Maximum Daily Loads are not a standard –Identify sources Current Policy Milieu

13. Cost share programs - voluntary –EPA’s 319 program, –Conservation Reserve Program, –Environmental Quality Improvement Program, –Conservation Security Program, and –Wetlands Reserve Program Water Quality Trading –Lack of standards on agriculture – ~ 475 of 700 watersheds agriculture contributes 90%+ of N loads! (Ribaudo et al. 2008) Programs to Support Policy

14. Bottom Line Current policy approach –Voluntary –Property rights with polluters Fundamentally different than pollution control other sectors Nonpoint source nature –Makes reversing property right seem hard –Don’t know how much externality generating activity is attributable to each source

15. An Alternative ? Reverse property rights Focus on practices (abatement actions) –Imperfect, but may still be welfare enhancing –Example: Abatement Action Permit System

16. An Abatement Action Permit System (AAPS) Based on Points Assign each practice/land use a point Set total points for watershed and allocate Include innovation options Allow trading Choose enforcement mechanism Adopt adaptive management

17. Features Not perfect, but a way to move forward? Puts property rights to clean water in hands of society Addresses fairness – early adopters rewarded Base on readily observable practices

18. Soil and Water Assessment Tool Watershed-scale simulation model developed by USDA - Agricultural Research Service Predicts ambient (instream) water quality associated with a spatially explicit set of land use/conservation practices Gassman et al. (2007) identify over 250 publications using SWAT

19. SWAT Team

20. Watershed a d b a b c a d a b a a a 13 Fields, 4 land use/abatement options: a, b, c, d SWAT simulates water quality under alternative land use, abatement activities SWAT: N, P, and Sediment

21. Least Cost Problem What is the optimal placement of conservation practices? Brute force strategy: –Using water quality/hydrology model, analyze all the feasible scenarios, picking cost-efficient solutions –But, if there are N abatement possibilities for each field and there are F fields, this implies a total of possible NF configurations to compare –30 fields, 2 options  over 1 billion possible scenarios!

22. Evolutionary Algorithm --- SPEA2 –Zitzler, Laumanns, and Thiele. “SPEA2: Improving the Strength Pareto Evolutionary Algorithm,” TIK-Report 103, May 2001, Errata added September, 2001 – Other water quality applications: Srivastava et al. (2002); Veith et al. (2003); Muleta and Nicklow (2005); Lant et al. (2005), Arabi et al. (2006)

23. Strength Pareto Evolutionary Algorithm Search technique to approximate pareto optimal frontier –Integrate Evolutionary Algorithm with water quality model –Search for a frontier of cost-efficient nutrient pollution reductions

24. Terminology “Individual” = specific assignment of practices to fields “Population” = set of individual watershed configurations b c a a d b a d d c a a

25. SPEA2 Applied to Optimal Watershed Design Step II: Run Swat and compute costs Step VI: Create offspring Step III: Identify best individuals Step IV: Evaluate stopping rule Step V: Choose parents Step I: Generate initial population Pareto frontier

26. Creating Offspring by Crossover Choose random position for cross Swap “heads” and “tails” Parents: abababababab cdcdcdcdcdcd Children: ababcdcdcdcd cdcdabababab Other variations: multiple points, random points, etc. Crossover point

27. Mutation Randomly replace an abatement option with another Example Child1:ababcdcdcdcd might become:ababacdcdcd Usually use a low mutation rate ~.003

28. Boone River Watershed Iowa ~586,000 acres tile drained, 90% corn and soybeans 128 CAFOs (~480,000 head swine)

29. Natural Environment: Boone Some of the highest N loads in Iowa TNC priority area biodiversity Iowa DNR Protected Water Area

30. Common Land Unit Boundaries 16,430 distinct CLUs Detailed data related to: land use, farming practices, production costs, slope, soils, CSRs, etc. Weather station data

31. The Land use/Abatement Set For each CLU –Current practice –Land retirement –No tillage –Reduced fertilizer (20%) –Cover crops –Sensible combinations

32. Evolutionary Algorithm: Uniform Seeds Description/ Uniform Seed Cost ($1000) % N Reduction % P Reduction Baseline 0 Reduced Fertilizer 2,199 6 0 No till 2,589 28 44 NT, RF 4,788 34 45 Cover Crop 13,181 24 32 CCr,RF 15,380 29 32 CCr, NT 15,770 48 42 CCr,NT, RF 17,969 53 43 Land Retirement 110,294 82 92

34. Gains from Optimal Placement Practice Allocation (%) Cost ($1000 dollars)% N% P NTNT, RF CC, RF CC NT RFOther Cover Crops, Red. Fert 15,3802932 100 Same N reductions 2,77829448413<1 3 Same Cost 15,3654745823<1645

36. Per acre average costs of abatement actions needed to achieve equal percent reductions in N and P

37. Thanks for your attention! Much appreciation to Todd Campbell, Phil Gassman, Manoj Jha, Becky Olson, Sergey Rabotyagov, and Adriana Valcu for superb research and presentation support and to Marca, Silvia, Lyubov and loads of other thoughtful people too numerous to mention for insightful discussions. Financial support from the U.S. Environmental Protection Agency under their targeted watersheds grant program, the USDA Conservation Effects Assessment Program, and CSREES is very gratefully acknowleged.

38. Cost Data MeanMinimumMaximum Corn Price 1 $3.08/bu No – Till 2 $4.09/ac--- Reduced Fertilizer 3 $4.06/ac ($3.6) 0$22 Land Retirement 4 $206/ac ($24) $13$240 Cover Crops 5 $25/ac 1.Average corn price from 2004-2009. 2.IFIP data. 3.Corn yield response curves to nitrogen fertilizer (Sawyer et al.(2006) ) and corn price data used to estimate reduced profits 4.The cost of land retirement is estimated Iowa CSR and cash rental rates (Edwards and Smith, 2009). County level rents were converted into CLU level data by weighting county estimates by the share of CLU area located in a county. 5.Estimated Costs of Crop Production in Iowa – 2008 (FM-1712) and Estimating Farm Machinery Costs (PM-710), actual seed costs, and bulk generic glyphosate in 2008. No fertilizer applied for cover crop growth

39. Least Cost for N and P Reductions Target % Decrease Cost Reduction (%)Watershed practices (counts of HRUs) ($1,000)($/acre)NPBaselineNTCC CC, NTRF NT, RF CC, RF CC, NT, RF Retire Land 101,1582.1911211781795402311342 202,0643.90213358023104211202 303,3896.4130441239813338251072 408,07215.264045794903217356083 5020,81539.3650 51051112966111635247 6039,65174.9860 6353921381828827 7079,194149.75708146123692417532039 80104,993198.5380894839171622780

40. Fitness Assignment