Multivariate geospatial modelling of drought exposure using multi-year remote sensing data

Understanding the temporal evolution and spatial distribution of drought stress is crucial to enhance climate resilience in Mediterranean vineyards. This study presents a geomatics-based approach to analyze monthly drought variability over ten years (2015-2025) in a Southern Italy vineyard. The method integrates multisource satellite data, vegetation indices, land surface temperature, and soil moisture, with meteorological variables from the AgERA5 reanalysis, including air temperature, dew point, humidity, wind speed, precipitation, solar radiation, and vapor pressure. These datasets were harmonized, normalized, and reduced via Principal Component Analysis to identify dominant drought drivers. A composite drought index was then computed and classified through K-means clustering into distinct stress levels. Results revealed three consistent drought stress classes: wet, transitional, and dry. The contribution of individual climate and satellite variables aligned well with their expected influence on drought conditions, supporting the classification's validity. Temporal patterns showed that high-stress periods mainly occurred during summer and autumn, reflecting seasonal drought peaks typical of Mediterranean climates. Conversely, winter generally exhibited low drought stress, though occasional exceptions suggest that even cooler months can experience moderate water stress. This spatiotemporal classification provides a concise, synthetic time series representing drought intensity and evolution over the decade. Such insights facilitate early identification of critical drought periods, aiding vineyard management decisions aimed at mitigating water stress impacts. By combining Earth observation with climate data, this scalable methodology offers a valuable tool for data-driven agriculture and climate risk governance in viticultural systems.