Enhancing SWAP Simulation Accuracy Via Assimilation of Leaf Area Index and Soil Moisture under Different Irrigation, Film Mulching and Maize Varieties Conditions
Xi Huang,Yin Zhao,Tongkai Guo,Xiaomin Mao
DOI: https://doi.org/10.1016/j.compag.2024.108625
IF: 8.3
2024-01-01
Computers and Electronics in Agriculture
Abstract:Crop models are effective tools for guiding irrigation, predicting crop yield and enhancing water productivity. However diverse agricultural regions pose challenges in simulation accuracy due to variability in irrigation, agronomic practices, and crop types. Data assimilation offers an effective approach to reduce these uncertainties. In this study, we explored the performance of four data assimilation methods when integrated with the SoilWater-Atmosphere-Plant (SWAP) model for simulating scenarios involving varying film mulching, irrigation levels and maize varieties. The data assimilation methods included global search (GS), Ensemble Kalman Filter (EnKF) for leaf area index (LAI) (EnKF-L), EnKF for soil water content (SWC) (EnKF-S), as well as EnKF for both LAI and SWC (EnKF-LS). To establish a baseline reference, we utilized the SWAP simulation without data assimilation method (OP) and compared its performance with the four data assimilation methods. To assess their performance, field experiments were conducted in the Shiyang River Basin of Northwest China during 2021 and 2022, with two varieties of spring maize subjected to two drip irrigation levels and two film mulching conditions. The treatment with full irrigation and no mulching in 2022 was employed to calibrate the SWAP model, while the other treatments were used for validation. The results showed that OP method failed to accurately capture the effects of film mulching and deficit irrigation on soil water and heat transport, crop growth, and meteorological parameters. The performance of the GS method was unstable, with a large range of root mean square error (RMSE) for maize growth and soil water content and temperature simulation. In contrast, the EnKF method significantly improved simulation accuracy, particularly when both LAI and SWC observations were assimilated simultaneously. The EnKF-LS methods reduced the root mean square error of the simulations by 0.04 to 1.57 m2/ m- 2(- |-) for LAI, 0.24 to 4.74 t hm-2 for aboveground dry biomass, 0 to 0.071 m3/m-3(- |-) for SWC, 0 to 2.74 degrees C for soil temperature, 2.20 to 6.91 t hm-2 for yield and 36.84 to 98.16 mm for evapotranspiration, both in comparison with the GS and OP methods. In terms of computational cost, the runtime of the three EnKF methods depends on the number of observations, whereas, over the entire simulation period of the SWAP model, the time required for EnKF methods was significantly less than that for the GS method. We recommend combining EnKFLS with the SWAP model to enhance the awareness of crop and soil conditions for agricultural decision support.