Accounting for scaling effects on irrigation optimization within a land surface model using satellite observations

Quantifying irrigation is a key challenge in hydrology due to its impact on water and carbon cycles. This is compounded by the scarcity of benchmark irrigation data and limitations of Land Surface Models (LSMs), which hinder accurate grid-scale irrigation estimation. In this study, we optimize a sprinkler irrigation scheme within the Noah-MP LSM, as part of the NASA Land Information System, using Sentinel-1-based irrigation estimates and a genetic algorithm. Experiments in an intensively irrigated region of northeastern Spain (0.01° resolution) compare two calibration approaches: one adjusting the root-zone soil moisture threshold (Thirr) which triggers irrigation, and another introducing a Scale Irrigation Coefficient (SIC) parameter to account for spatial heterogeneity in irrigation practices. The Thirr calibration shows limitation in the system’s flexibility, causing sparse irrigation applications with excessive water amounts that optimization cannot correct. In contrast, SIC calibration improves irrigation dynamics, reduces model errors, and better represents interannual surface soil moisture anomalies, outperforming the default scheme against in situ data. Results highlight that assuming full irrigation at resolutions equal or beyond 1 km is unrealistic due to two reasons: first, farmers cannot irrigate all fields within a grid cell simultaneously; second, the heterogeneous field mosaic further complicates uniform irrigation. Comparisons with satellite-based evapotranspiration (ET) and Gross Primary Production (GPP) datasets highlight inconsistencies between model estimates and satellite ET products, revealing persistent vegetation dynamics issues. Future efforts could leverage the calibrated scheme with satellite data assimilation to improve soil moisture and vegetation conditions, capturing complex interactions between irrigation and the water-carbon cycles.