Mapping soil water properties using soil samples and satellite images in the irrigated area of Biota (Spain)

Abstract

Mapping Total Available Water (TAW) is particularly challenging in shallow, stony soils with limited water retention, as in the Monte Saso de Biota irrigated area (1185 ha, Zaragoza, Spain). TAW maps are needed to optimize water and energy for pumping. The objective was to elaborate and assess two types of maps: quantitative, using regressions of Sentinel-2 imagery with field data, and the classic qualitative approach, with soil units based on field data and terrain analysis. A field campaign obtained 172 soil samples and TAW was estimated from field capacity (FC), wilting point (WP), bulk density, stoniness (STO) and effective depth (Z). A database of 82 winter–spring bare-soil Sentinel-2 images (2018–2024) informed predictive models, in which bands B02, B08, B11, and B12 were frequent predictors. Single-date and multiple-date (combi) models were developed, using 70% of field data. Combi models provided better predictive equations than single-date models. The best models reached coefficients of determination (R²) of 92% for WP, 89% for Z, 84% for STO, 77% for TAW, and 72% for FC. Models were ranked by performance and applied at pixel scale to elaborate quantitative maps. Validation used 30% of samples, obtaining R² of 19–45%. In parallel, a qualitative TAW map was produced using four units. The quantitative TAW map was successfully compared with maize yield maps (correlation coefficient of 0.35). Both types of maps captured terrain-related patterns, but the quantitative TAW map offered finer spatial detail, providing greater value for irrigation management studies.