Satellite-based optimization of irrigation in a land surface model accounting for scaling effects

Irrigation strongly influences land-atmosphere interactions and the terrestrial water cycle, yet its representation in land surface models (LSMs) remains highly uncertain. These uncertainties arise from both the scarcity of reliable irrigation benchmarks and the challenge of representing heterogeneous irrigation practices within coarse model grid cells (e.g., kilometer-scale resolutions). 

In this study, we examine structural limitations in the representation of irrigation within the Noah-MP LSM, implemented in the NASA Land Information System, by testing different calibration strategies. A sprinkler irrigation scheme is optimized using Sentinel-1-derived irrigation estimates and a genetic algorithm over an intensively irrigated region of northeastern Spain at a 0.01° spatial resolution. Two calibration approaches are evaluated: (i) adjusting the soil moisture threshold (Th_irr) that triggers irrigation, and (ii) introducing a Scale Irrigation Coefficient (SIC) to account for sub-grid heterogeneity in irrigated area and applications’ timing. 

Results show that calibrating Th_irr alone provides limited flexibility, resulting in unrealistic irrigation peaks and excessive water application. By contrast, the optimized SIC-based parameterization substantially improves irrigation dynamics, reduces model errors relative to benchmark in situ observations, and better captures interannual variability in surface soil moisture. Findings demonstrate that assuming uniform, full-grid irrigation at resolutions of ~1 km or coarser is physically unrealistic due to both operational constraints on irrigation practices and the fragmented structure of agricultural landscapes. Comparisons with satellite-based evapotranspiration and gross primary production datasets also reveal inconsistencies in simulated vegetation responses, highlighting remaining limitations in vegetation parameterization.  

Overall, this work underscores the importance of explicitly accounting for scaling effects in irrigation schemes and points toward future integration of satellite data assimilation to enhance representation of irrigation-water-carbon interactions.