An Integrated Geospatial Framework for Climate-Resilient Honey Bee Habitat Mapping, Fragmentation Assessment, and Apiary Site Selection.

Agro-pastoral communities in East Africa are increasingly affected by recurrent natural hazards, particularly drought, which leads to severe pasture degradation and substantial losses of livestock—their primary livelihood asset. These challenges highlight the urgent need for adaptive livelihood diversification strategies that reduce dependence on climate-sensitive resources. Beekeeping has emerged as a viable and climate-resilient alternative; however, its successful implementation requires reliable information on suitable habitats and forage availability. Advances in remote sensing and geospatial technologies provide powerful tools to support such strategies by integrating multi-source satellite data to monitor vegetation dynamics, landscape structure, and climate variability, thereby enabling informed planning and sustainable livelihood development in agro-pastoral environments. This study presents an integrated geospatial assessment of honey bee habitat suitability, fragmentation dynamics, and climate-smart apiary site selection in Yabelo (Ethiopia) and Taita-Taveta County (Kenya). Using Google Earth Engine, multi-source satellite data, including Planet Scope, Sentinel-1 SAR, Sentinel-2 multispectral imagery, and SRTM Digital Elevation Model were analyzed to map honey bee habitats through advanced machine learning techniques. Four classifiers (Gradient Tree Boosting, Random Forest, Classification and Regression Trees, and Support Vector Machine) were evaluated and combined using an Ensemble Learning Approach, achieving the highest classification accuracy, significantly outperforming individual models.

To understand landscape structure and resource accessibility, habitat fragmentation was assessed using key landscape metrics (Shannon diversity, contagion, and splitting index) across multiple spatial scales (500–3000 m buffers). Results reveal pronounced scale-dependent fragmentation, with Yabelo characterized by high landscape heterogeneity but increasing patch disconnection at larger scales, while Taita-Taveta exhibits more continuous but less diverse habitats. Human-induced land-use changes and edge effects were identified as major drivers of fragmentation, with wetlands and water bodies being particularly vulnerable.

To support adaptive beekeeping under climate change, fuzzy Multi-Criteria Decision-Making methods incorporating current and future climate projections (SSP1-2.6 and SSP5-8.5) were applied for apiary site suitability analysis. Findings indicate a projected decline in highly suitable apiary areas under future climate scenarios, highlighting climate-driven shifts in bee forage availability. Overall, this integrated framework demonstrates how ensemble machine learning, landscape ecology, and climate projections can support evidence-based, climate-resilient planning for sustainable beekeeping in East Africa.

Keywords: Agro-pastoral livelihoods, Remote sensing, Honey bee habitat, Climate change, Ensemble learning, Apiary suitability