Coupling SWAT and DSSAT models for assessing climate change impacts on watershed-level crop production

Introduction

Consider a single wheat field in the Rhineland: its yield depends on rainfall, temperature, and nitrogen supply, but these are shaped by what happens upstream runoff generation, lateral flow, and nutrient leaching across the watershed. Traditional crop models like DSSAT treat each field as an isolated soil column, ignoring these landscape connections. Yet in undulating terrain, water redistribution can shift effective availability by 30-80 mm per season. SWAT was built to model watershed processes surface runoff, lateral flow, groundwater recharge, and stream routing at sub-basin resolution. But SWAT's built-in crop routines are simplified compared to DSSAT's detailed phenology and stress response modules. Coupling the two takes advantage of SWAT's spatial hydrology and DSSAT's crop physiology. Several coupling attempts targeted tropical systems in Asia, but applications in temperate European watersheds are fewer. This research aimed to: 1) Develop a SWAT-DSSAT coupled framework for the Sieg River watershed; 2) Validate simulated winter wheat yield against sub-basin observations; and 3) Project yield impacts under RCP 4.5 and RCP 8.5 for the 2050s.

Theoretical background

The coupling follows a sequential architecture. SWAT runs daily, generating spatially distributed soil water content, temperature, lateral flow, and plant-available nitrogen for each hydrological response unit. These outputs replace DSSAT's internal soil water balance, while DSSAT simulates phenological development and grain yield. Feedback from DSSAT to SWAT (canopy cover affecting runoff) was handled through updated leaf area index at weekly intervals.

Material and Methods

Material

The Sieg River watershed (2,860 km²) drains into the Rhine near Bonn (50°46′N, 7°11′E, elevation 45-480 m). It was delineated into five sub-basins from a 30 m DEM. Land use from CORINE 2018 showed winter wheat on 34% of agricultural area. Soil data came from the BK50 map. Daily weather (2000-2020) was obtained from seven DWD stations. Future climate used delta factors from a threemodel CMIP5 ensemble for RCP 4.5 and RCP 8.5 at the 2050s. Observed wheat yields (2015-2020) were from the Landwirtschaftskammer NRW.

Methods

SWAT was calibrated (2005-2014) and validated (2015-2020) against monthly stream flow at Eitorf using SWAT-CUP. DSSAT (v4.7.5) was calibrated for cultivar Julius using 2015-2017 field trial data from the University of Bonn. The coupled model was validated against 2018-2020 sub-basin yields. Performance was assessed using R², NSE, nRMSE, d-index, and PBIAS. For projections, 30-year simulations per scenario were run and aggregated by sub-basin. Under RCP 8.5, watershed-average yield declined by 17.3%, but spatial variation was striking: Sub-basin 1 (upper watershed) lost 22.0% while Sub-basin 4 (lower) lost only 11.0% (Table 2). RCP 4.5 showed a moderate average decline of 9.8%. The spatial gradient reflects differences in soil water-holding capacity and the buffering effect of lateral water redistribution in lower-lying areas. shows that all performance metrics remained above 0.73 even under the most extreme scenario, though a general decline in model skill was evident from baseline to RCP 8.5, reflecting greater uncertainty in simulating novel climate conditions.

Results

Performance evaluation SWAT's hydrological calibration achieved NSE of 0.81 and R² of 0.86 for monthly stream flow at Eitorf during validation. A 30-year watershed simulation completed in 47 minutes on a standard workstation. Data exchange between SWAT and DSSAT accounted for 12% of total run time. The weekly LAI feedback loop introduced negligible impact on seasonal yield estimates (<0.4% difference versus daily feedback).

Discussion

The coupled framework outperformed standalone DSSAT by 4-7 percentage points in R² across sub-basins, confirming that lateral water redistribution improves yield prediction in undulating terrain. The 22% yield loss in the upper watershed under RCP 8.5 reflects both higher drought exposure (shallower soils, rapid drainage) and reduced lateral water inflow. Lower Sub-basin 4 benefits from alluvial soils with high water-holding capacity, buffering it against moderate stress. The sequential coupling architecture is simpler than fully integrated approaches but sufficient for seasonal yield estimation and computationally manageable. A limitation is that farmer adaptation (variety change, irrigation expansion) was not included, which could offset some projected losses.

Conclusion

References

1. Jones JW, Hoogenboom G, Porter CH. The DSSAT cropping system model. European Journal of Agronomy. 2003;18(3):235-265.
2. Holzworth DP, Huth NI, deVoil PG. APSIM -evolution towards a new generation of agricultural systems simulation. Environmental Modelling and Software. 2014;62:327-350.
3. Western AW, Grayson RB, Blöschl G. Observed spatial organization of soil moisture and its relation to terrain indices. Water Resources Research. 1999;35(3):797-810.
4. Arnold JG, Srinivasan R, Muttiah RS. Large area hydrologic modeling and assessment part I: model development. Journal of the American Water Resources Association. 1998;34(1):73-89.
5. Hoogenboom G, Porter CH, Boote KJ. The DSSAT crop modeling ecosystem: advances in crop modeling for a sustainable agriculture. Cambridge: Burleigh Dodds Science; 2019.
6. Bieger K, Arnold JG, Rathjens H. Introduction to SWAT+, a completely restructured version of the Soil and Water Assessment Tool. Journal of the American Water Resources Association. 2017;53(1):115-130.
7. Luo Y, Zhang Z, Chen Y. Coupling SWAT and DSSAT models for simulating crop growth in an irrigated watershed. Transactions of the ASABE. 2019;62(5):12711284.
8. Srinivasan R, Zhang X, Arnold JG. SWAT ungauged: hydrological budget and crop yield predictions in the Upper Mississippi River Basin. Transactions of the ASABE. 2010;53(5):1533-1546.
9. Große-Stoltenberg A, Zeuss D,138 J. Land use change and wheat production in the Rhine-Sieg region, Germany. Regional Environmental Change. 2018;18(7):2087-2099.
10. Rosenzweig C, Elliott J, Deryng D. Assessing agricultural risks of climate change in the 21st century in a global gridded crop model intercomparison. Proceedings of the National Academy of Sciences. 2014;111(9):3268-3273.
11. Jacob D, Petersen J, Eggert B. EURO-CORDEX: new highresolution climate change projections for European impact research. Regional Environmental Change. 2014;14(2):563578.
12. R Core Team. R: A Language and Environment for Statistical Computing. Vienna: R Foundation; 2023.
13. Gassman PW, Reyes MR, Green CH. The Soil and Water Assessment Tool: historical development, applications and future research directions. Transactions of the ASABE. 2007;50(4):1211-1250.
14. Trnka M, Rötter RP, Ruiz-Ramos M. Adverse weather conditions for European wheat production will become more frequent with climate change. Nature Climate Change. 2014;4(7):637-643.
15. Seneviratne SI, Corti T, Davin EL. Investigating soil moisture-climate interactions in a changing climate: a review. Earth-Science Reviews. 2010;99(3):125-161.
16. Volk M, Liersch S, Schmidt G. Towards the implementation of the European Water Framework Directive? Lessons learned from water quality simulations in an agricultural watershed. Land Use Policy. 2009;26(3):580-588.
17. Ewert F, Rötter RP, Bindi M. Crop modelling for integrated assessment of risk to food production from climate change. Environmental Modelling and Software. 2015;72:287-303.
18. Webber H, Ewert F, Olesen JE. Diverging importance of drought stress for maize and winter wheat in Europe. Nature Communications. 2018;9:4249.