Integrating Sentinel-2 imagery and meteorological data to estimate leaf area index and leaf water potential, with a leave-field-out validation strategy in chickpea fields

Chickpea (Cicer arietinum L.) is a major grain legume, playing a crucial role in semi-arid systems. Its irrigation management is closely tied to its physiological condition, while environmental factors such as water availability and extreme heat significantly cause yield losses, especially in rainfed systems. The ability to estimate within field chickpea physiological traits may enhance understanding of its responses to environmental conditions and support agricultural management decisions. This study aimed to develop spatial estimation models of Leaf Area Index (LAI) and Leaf Water Potential (LWP), using Sentinel-2 imagery and meteorological data, mimicking practical application. Total of 404 and 361 measurements of LAI and LWP were collected from 14 and 17 fields, respectively (2022–2023). A leave-field-out validation strategy reflecting operational scenarios was employed. Partial least squares regression resulted in the highest accuracy for chickpea LAI, with coefficient of determination (R²) and root mean square error (RMSE) of 0.73 and 1.38 m² m⁻², respectively, while random forest excelled during early growth stages, prior to flowering. For LWP, Ridge Regression (RR) and support vector machine performed comparably overall (R²: 0.15; RMSE: 0.34 and 0.33 MPa); however, RR outperformed during the critical irrigation decision stage (post-flowering). Time-series commercial chickpea trait maps effectively captured LAI and LWP classical seasonal dynamics, demonstrating their relevance for farmers and their potential to elucidate critical relationships between canopy growth and irrigation onset in indeterminate crops such as chickpea. These LAI and LWP models establish foundational models that potentially enable precise and knowledge-based agricultural management tools for chickpea farmers.