Earth observation, geospatial data, and AI: new tools for climate evidence

Integrated Assessment of Urban Flooding and Heat Island Interactions: Geospatial Tech, Machine Learning, and Microclimate Dynamics
Authors: M. Imroz1, M.P. Akhtar2*, M.K. Sharma3, F. Alshehri4

1,3 Department of Civil Engineering, School of Engineering (SCE), Manipal University Jaipur, Jaipur, India
2,4 Department of Civil Engineering, College of Engineering, University of Bisha, P.O. Box 551, Bisha, 67614, Saudi Arabia

Urban flooding and the Urban Heat Island (UHI) impact are crucial challenges driven by rapid urbanization and climate change. Understanding their linked dynamics is crucial for mitigation. However, there are gaps in analytical frameworks representing the spatio-temporal interplay among these phenomena, especially regarding microclimate variations and land use and land cover (LULC) changes. This review highlights LULC, microclimate variables, and GIS-based techniques, incorporating machine learning approaches, evaluating more than seventy-four studies. Approximately 30–32% of studies employed hybrid learning models (CNN-LSTM) for UHI, while 60–65% used ensemble techniques (Random Forest, Gradient Boosting) for flood analysis. Deep learning models nonlinear relationships, whereas ensemble models support interpretability and multi-source integration. However, the sector lacks scalable models leveraging real-time IoT data. Their interactions amplify hydrological and thermal stresses, threatening livability, infrastructure, and public health. Most studies examine UHI and flooding separately, overlooking microclimatic variations, urban morphology, IoT integration, and model transferability. A unified data-driven framework is proposed to simulate hazards. Studies (2010–2025) were reviewed using PRISMA 2020 from Scopus, Web of Science, ScienceDirect, and Google Scholar. Ensemble methods dominated flood modeling (61–65%), while CNN–LSTM models were used for UHI (30–32%). CNN–LSTM performed well for moderate-resolution data, whereas transformers and GNNs excelled for large-scale scenarios. Key gaps include interpretability and IoT integration.

From Risk Data to Early Action: Evidence from Digital and Spatial Technologies for Anticipatory Action in Nepal’s Mahakali Basin
Authors: Shudarshan Hamal, Pratistha Katwal, Dechen Dolma Sherpa, and Biddya Bhandari. Affiliation: DASTAA Project Consortium, Kathmandu, Nepal.

This published lessons report focuses on use of digital and spatial technologies as tools to facilitate anticipatory action during the monsoon season in 2024 in three flood-prone municipalities in the Mahakali Basin of Nepal. It is based on the Early Warnings for All (EW4All) framework, which evaluates an end-to-end system from household risk knowledge to flood modelling, hydro-meteorological monitoring, impact-based forecasting, multi-channel warning, preparedness, evacuation and anticipatory financing. Evidence was created through household risk assessment for 22,803 households, HEC-RAS flood hazard modelling, localized INFORM-based risk profiling, hydro-meteorological monitoring, records of impact-based forecasts, warning dissemination, preparedness activities and implementation learning. The programme reached 117 local youth, who were mobilised for digital data collection, and 414 government officials who were trained. During the monsoon period, 18 impact-based forecast bulletins were released, 127,667 IVR messages were sent out, and 41,500 SMS alerts were sent.

Technology was found to be most effective when nested in local institutional setting, when used in community engagement, and when used in trusted communication. An early action message combining co-designed messages, multi-channel warning, real-time monitoring, evacuation decision and pre-arranged cash was illustrated in July 2024 cloudburst in Dodhara Chandani where 624 mm of rainfall was recorded. Some factors that were a problem were limited forecasting data and lead time, connectivity and mobile access gaps, communication costs, lack of trust and insufficient evacuation infrastructure. The report provides lessons learnt from the process of developing more inclusive and locally owned and scalable anticipatory action systems in climate-vulnerable settings.

Geospatial Impact Evaluation of Climate and Environmental Interventions: Evidence from Three Country Case Studies
Governments and development partners invest billions annually in climate adaptation, ecosystem restoration, biodiversity conservation, and disaster risk reduction. Yet a key barrier to evidence-informed climate action is the inability to tell whether environmental changes are attributable to interventions or would have occurred anyway. Conventional monitoring mainly tracks implementation and outputs, offering limited evidence on effectiveness or investment value.

This Open Access Handbook, peer-reviewed and co-edited by Kunwar K. Singh (AidData, William & Mary), Ariel BenYishay (Chief Economist, AidData; Associate Professor of Economics, William & Mary), and Claire Zanuso (Head of Emerging Tech Lab, AFD; former Head of Impact Evaluation), reflects the effort of 80+ researchers, practitioners, and evaluators across five continents, notably from AFD, KfW, IFAD, FAO, and the World Bank (DIME, IEG). It shows how geospatial impact evaluation (GIE) integrates Earth observation with counterfactual evaluation to generate credible evidence of climate intervention effectiveness.

Drawing on published evidence in climate adaptation, landscape restoration, biodiversity conservation, and nature-based solutions, it presents a framework combining satellite data with causal inference methods (matching, difference-in-differences) to estimate impacts where monitoring is limited by incomplete records and heterogeneity.

GIE moves beyond descriptive monitoring by attributing environmental change to specific interventions, enabling rigorous assessment across projects and portfolios, from protected areas to dryland rehabilitation and climate-resilient agriculture, in a scalable, replicable approach for data-constrained settings.

Turning satellite data into policy evidence, GIE strengthens accountability and adaptive management, embedding rigorous evaluation into climate policy as the Paris Agreement, Kunming-Montreal Framework, and UN Ecosystem Restoration Decade advance.