Disappearing wetlands in North Central Nigeria: a Google Earth Engine Z-statistics framework co-designed with local stakeholders

Wetlands in North Central Nigeria support fisheries, flood regulation and smallholder agriculture, yet local observations point to a rapid decline in extent and persistence. For many inland wetlands in sub-Saharan Africa there are no long-term, consistent spatial products to quantify habitat change in ways that are meaningful for biodiversity management. Focusing on Ibaji (Kogi State), we present a cloud-based Earth observation framework in Google Earth Engine (GEE), co-designed with the National Space Research and Development Agency, (Nigeria) to reconstruct three decades of wetland dynamics and derive indicators relevant to the Kunming–Montréal Global Biodiversity Framework. The work is developed as a space-based solution addressing a water-challenge within the UNOOSA Space4Water programme, linking EO-based wetland monitoring to ecosystem services and community risk. The method adapts and simplifies a land-cover time-series approach originally developed for protected European mountain landscapes, replacing habitat-level mapping with a stakeholder-derived land cover reference. A manually interpreted training dataset for a single reference year, produced by local experts from Sentinel-2A and very-high spatial resolution imagery, is used to classify that year with a Random Forest model in GEE. The classifier uses multi-season Landsat 4-9 best-available-pixel composites, SRTM topography and TerraClimate variables. We then apply a Z-statistics approach to propagate this information through time. For each other year (1985–2024), multi-season predictor stacks are compared to the reference class signatures and only pixels close to the class-specific multivariate mean are retained as pure and stable training samples. These annually updated training sets drive year-specific Random Forest models that generate consistent land-cover maps with explicit water and wetland classes. These maps serve to derive indicators of wetland extent and trajectories of conversion to other land cover areas. Current work focuses on accuracy assessment with independent very-high resolution validation data, uncertainty estimation and spatially explicit change detection. This Space4Water “space-based solution” will support discussion on co-producing EO-based wetland and biodiversity indicators under sparse in situ data and discontinuous time series.

Disclaimer: The views expressed herein are those of the author(s) and do not necessarily reflect the views of the United Nations.