Andean cities are particularly vulnerable to natural hazards due to their geographical location. Climate change, deforestation and extractive activities further increase these risks. Meanwhile, the Amazon rainforest is approaching a tipping point that could alter precipitation patterns and the water supply in South America. As Andean cities depend heavily on stable precipitation and the páramo ecosystem, this could make them more vulnerable to water scarcity and natural hazards. The project uses a multi-method approach with a focus on satellite remote sensing to understand how cross-scale environmental changes impact freshwater systems, urban populations and the resilience of Andean cities.

InSAR (Interferometric Synthetic Aperture Radar) describes several remote sensing techniques using radar signals to detect changes in the Earth’s surface at the millimeter scale, including subsidence, uplift and deformation resulting from natural or human induced activities. Persistent Scatterer Interferometry (PSI) is a specialized InSAR method that uses a large time series of SAR satellite images to identify specific and punctual ground movements. Among other things, this can be used to investigate the stability of terminal moraines and improve the risk assessment of Glacial Lake Outburst Floods (GLOFs). Furthermore, these methods are to be used in future to detect permafrost in Fennoscandia and quantify its thawing across the entire region.

Climate change and land degradation affect strongly smallholder farmers in Sub-Sahara Africa. One solution addressing this downward spiral is the implementation of land restoration activities, that rehabilitate water and soil nutrient cycles and improve resilience and food security in the long-term. Remote sensing can contribute to the participatory planning process of such interventions by complementing local knowledge through environmental evidence. Combining satellite information with Citizen Science principles, land rehabilitation works will follow informed decision-making processes, having more sustainable impact on local resilience.

In the field of Geospatial Artificial Intelligence (AI), algorithms are trained on geodata such as Earth Observation (EO) time series to analyze spatial and temporal patterns. Deep Learning (DL) offers a higher potential for modelling nonlinear relationships compared to traditional approaches. The research focus lies on the spatially explicit risk forecasting of future artisanal and small-scale mining (ASM) sites in the Amazon using various DL architectures. Based on EO time series, the goal is to predict where the risk of primary forest being converted into illegal mining areas is elevated. The primary data source is Sentinel-1 Synthetic Aperture Radar (SAR) imagery. The research aims to contribute to preventive governance of unregulated ASM.