GeoAI for Sustainable Futures: Integrating Artificial Intelligence, Earth Observation and GIS for Environmental Resilience

Abstract Geospatial artificial intelligence (GeoAI), the integration of artificial intelligence with Earth Observation (EO) and Geographic Information Systems (GIS), is reshaping how environmental resilience is monitored and managed. This paper reviews GeoAI as both a technical practice and a geographical problem, arguing that its value for resilience planning depends on respecting the spatial properties — dependence, heterogeneity and scale — that distinguish geographic data from ordinary tabular data. The paper synthesises published evidence across four resilience-relevant domains: land use/land cover (LULC) change detection, drought risk assessment, landslide hazard zonation and groundwater potential mapping and situates this evidence within the Deccan Plateau and Western Ghats margin of Maharashtra, a terrain that compresses escarpment landslide risk, valley land-use change and rain-shadow drought within a single administrative region. The review finds consistent but bounded accuracy gains from machine-learning methods over conventional GIS overlay, with the largest gains reported where training and validation data are collected locally rather than transferred from elsewhere. The paper closes with a phased adoption roadmap for a teaching-and-research college operating in this terrain, arguing that institutional investment in ground-truth data collection should precede investment in more elaborate models.