Field-scale irrigation estimation using Sentinel-1 and Sentinel-2 data within a soil water balance model coupled to the Water Cloud Model

Irrigation is a significant component of the water cycle in agricultural regions: however, information about how much and where irrigation is applied is scarce and uncertain. Remote sensing-based products have been developed to retrieve irrigation across scales, but they exhibit limitations due to spatial uncertainties and confounding factors, particularly in highly fragmented agricultural landscapes. This paper presents a novel method to estimate irrigation at the field scale by integrating Sentinel-1 C-band Synthetic Aperture Radar (SAR) backscatter data and Sentinel-2 NDVI data within a modeling framework. The proposed approach involves the joint calibration of the Water Cloud Model (WCM) and a two-layer Soil Water Balance (SWB) model to simultaneously simulate backscatter, irrigation, surface and root-zone soil moisture, and other water balance components. The model is driven by reanalysis meteorological data, and selected parameters in the WCM and SWB are calibrated exclusively using Sentinel-1 backscatter data, thus making the framework independent of in situ ground measurements. The framework is tested on 38 fields within an irrigation district in the Po Valley, Italy, during 2018, for which irrigation data are available as reference. It is shown that the capability to estimate irrigation events depends on the treatment of the SAR data: after bias-correcting Sentinel-1 observations, average time series correlations of 0.43 to 0.95 could be obtained for daily to bi-weekly irrigation estimates, respectively. The associated bias amounts to 0.01 mm/day and 0.04 mm/15 days, and less than 1% for the annual estimates.