Work

Stochastic Optimization and Machine Learning for Intelligent Water Management in Precision Agriculture

Year: 2026
Type: article
Abstract: Efficient irrigation-water allocation requires accounting for climatic uncertainty and heterogeneous crop water requirements. This study develops an integrated framework combining machine learning prediction, climatic scenario generation, crop water-balance modeling, and risk-averse stochastic optimization. The framework is evaluated for four municipalities in Hidalgo, Mexico, using historical meteorological and agricultural records. Reference evapotranspiration and precipitation predictions are incorporated into 200 climatic scenarios, which are propagated through crop-specific water-balance calculations to estimate uncertain net irrigation requirements. A static mean–CVaR optimization model determines crop-level water allocations under exploratory water budgets defined relative to simulated mean demand. The objective minimizes expected weighted irrigation deficits and their upper-tail risk, rather than observed monetary losses or irrigation expenditures. Computational results show that the influence of risk aversion depends on the available water budget. Under the adopted production-value-based priority weights, allocations become concentrated in grain barley as water scarcity increases, and the tested risk-aversion settings yield coincident allocations under more restrictive budgets. Supplementary sensitivity analysis indicates that exclusive allocation to barley depends on the adopted crop-priority criterion. The results provide a reproducible, scenario-based framework for comparing risk-aware irrigation-allocation decisions, while remaining conditional on the climatic scenario, crop–demand assumptions, priority weights, and hypothetical water budgets.
Source: Agriculture
Language: English
Open Access status: gold
Estimated APC paid: $3,137
OpenAlex ID: W7235801883
FWCI:
Cites:
Cited by:
Related to: 10