Combining Earth Observation and AI to advance multi-risk assessment of hot and dry events on crops in the Adige River basin

Cascading and compounding multi-hazard events pose increasing challenges, presenting serious direct and indirect threats to people, the environment, and economic assets. Addressing these events and building disaster risk reduction capacity is crucial. This requires not only leveraging novel technologies such as modern Earth Observation (EO) platforms and AI, but also integrating them into effective multi-risk assessment frameworks. This study, conducted within the ESA EO4MultiHazard project, aims to exploit EO data to deepen our understanding of how multi-hazard cascading impacts unfold in affected areas. Specifically, it focuses on cascading and compounding hot and dry events—namely, heatwaves and droughts—and their impacts on crop vegetation in the lower Adige River Basin, located in northeastern Italy. The Adige River serves as a critical resource for the area's intensive agriculture, as its waters supply a dense irrigation network, making it especially vulnerable to reduced water availability during hot and dry conditions. Multi-risk assessment methodologies involve several key steps, including the spatiotemporal identification of hazards and the assessment of exposure and vulnerability. The ultimate goal of this study is to use high-resolution EO data to enhance the understanding of the different risk dimensions and identify risk susceptible areas. The multi-hazard identification methodology was adapted from the Myriad-EU project and applied to the Adige River Basin to analyze hot and dry events over the past 74 years (1950–2023) using the E-Obs gridded dataset. This analysis enabled the identification of general drought and heatwave trends, as well as the most severe and relevant events to inform a more detailed EO analysis. The 2022 drought, a recent and highly severe event, was selected as a case study period. In situ data—such as information on the irrigation network, irrigation districts, river discharge, and crop species at the field level—were combined with EO data from Sentinel-2. This integration of high-resolution satellite imagery (up to 10 meters) with detailed ground information allowed for the detection of vegetation stress responses to hot and dry events, serving as proxies for crop impacts. This approach not only identifies the most susceptible areas to inform multi-risk assessments, but also lays the groundwork for applying AI methodologies to predict future impacts under various climate scenarios. By creating past and present-day susceptibility maps, this study advances our understanding of hot and dry event dynamics on crops, and it demonstrates the potential of integrating advanced analytical tools and EO data into a multi-hazard framework to pave the way for machine learning applications for future climate multi-risk assessment and adaptation strategies.