Coupling SWAT simulations with advanced machine learning models for district-level crop yield prediction in the Varuna River Basin
Journal of Water and Climate Change · 01.07.2026 · 17 (7) · 2156-2176
DOI: 10.2166/wcc.2026.549 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Water and Climate Change, online erschienen am 20.07.2026. Verfasst von Shailesh Kushwaha und 2 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 73 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Accurate crop-yield prediction under hydro-climatic variability is challenging in monsoon-dominated agricultural basins. This study develops a process-informed SWAT-machine learning framework for district-level crop-yield prediction in the Varuna River Basin (∼3,665 km2), India. The Soil and Water Assessment Tool (SWAT) was calibrated (R2 = 0.91; NSE = 0.84) and validated (R2 = 0.69; NSE = 0.57), indicating satisfactory hydrological simulation. Water-balance analysis showed that precipitation (754.1 mm) was mainly partitioned into evapotranspiration (531.9 mm; ∼71%) and surface runoff (207.16 mm; ∼28%), with limited groundwater flow (3.47 mm) and lateral flow (0.42 mm), indicating runoff-dominated hydrology. SWAT-derived hydro-climatic and crop-growth variables were used to train Random Forest, XGBoost, artificial neural network, and Ridge regression models. Ensemble models performed best, with RF and XGBoost achieving comparable accuracy (R2 ≈ 0.76; RMSE ≈ 0.33; MAE ≈ 0.24). Seasonal and crop-wise evaluation showed higher accuracy during Rabi (R2 ≈ 0.81) than Kharif (R2 ≈ 0.65), and better wheat than rice prediction. These differences may reflect lower monsoon-rainfall uncertainty and irrigation/residual soil-moisture dependence during Rabi, whereas Kharif (rice) prediction is more affected by rainfall variability and nonlinear crop-water responses. The framework supports climate-resilient crop planning, vulnerable-system identification, and district-scale water-allocation decisions.
Zitieren
APA
Kushwaha, S., Maurya, S., & Soni, P. (2026). Coupling SWAT simulations with advanced machine learning models for district-level crop yield prediction in the Varuna River Basin. Journal of Water and Climate Change, 17(7), 2156-2176. https://doi.org/10.2166/wcc.2026.549
Vancouver
Kushwaha S, Maurya S, Soni P. Coupling SWAT simulations with advanced machine learning models for district-level crop yield prediction in the Varuna River Basin. Journal of Water and Climate Change. 2026;17(7):2156-2176. doi:10.2166/wcc.2026.549
BibTeX
@article{Kushwaha2026,
author = {Kushwaha, Shailesh and Maurya, Saurabh and Soni, Pramod},
title = {Coupling SWAT simulations with advanced machine learning models for district-level crop yield prediction in the Varuna River Basin},
journal = {Journal of Water and Climate Change},
year = {2026},
volume = {17},
number = {7},
pages = {2156--2176},
doi = {10.2166/wcc.2026.549}
}
EUVIMED