Field-Scale Irrigation Source Attribution

Irrigated agriculture is a major freshwater user in the western United States, yet field-level information on whether irrigation relies on surface water, groundwater, or both remains limited. This lack of source attribution constrains water-scarcity assessments, curtailment analysis, and evaluations of irrigation efficiency, particularly where infrastructure and governance spatially decouple water sources from field locations. We present a data-driven framework for mapping field-level irrigation source access that integrates large-scale geospatial predictors with administrative water-rights information. The approach uses a two-stage probabilistic classification pipeline: first, estimating the likelihood of groundwater and surface-water access and then inferring conjunctive use from the structure and uncertainty of these probabilities. Preliminary findings suggest that single-source irrigation can be consistently identified at the field scale across diverse settings, whereas mixed-source conditions exhibit greater sensitivity to local context. This enables irrigation source information to be incorporated at the field scale, while explicitly identifying settings where local conditions govern source use.