County-Level Irrigation Water Demand Estimation Using Machine Learning: Case Study of California

Irrigated agriculture is the largest consumer of freshwater globally. Despite clarity influential factors and deriving forces, estimation volumetric irrigation demand using biophysical models prohibitively difficult. Data-driven have proven their ability to predict geophysical hydrological phenomena with only a handful input variables; however, lack reliable data in most agricultural regions world hinders effectiveness these approaches. Attempting estimate water demand, we first analyze correlation potential influencing variables water. We develop machine learning California’s annual, county-level based on statistical analysis findings over an 18-year time span. Input are different combinations meteorological geographical characteristics, cropped area, crop category. After testing various regression approaches, result shows that Gaussian process produces best results. Our suggest irrigated air temperature, vapor pressure deficit significant predicting demand. This research also can high accuracy (R2 higher than 0.97 RMSE as low 0.06 km3) variable combinations. An accurate use categories areas assist decision-making processes improve management strategies. The proposed model help policy makers evaluate climatological scenarios hence be used decision support tool for at regional scale.