Recent work has identified semi-arid and savanna-type (SAST) ecosystems as a critical component of interannual variability in the Earth system, yet our understanding of the spatial and temporal patterns present in these systems remains limited. Leaf phenology - the timing of the appearance, presence, and senescence of plant leaves - is a major contributor to this variability. Plants grow and drop their leaves in response to a variety of cues, including soil moisture, rainfall, day length, and relative humidity, and alternative phenological strategies might often co-exist in the same location. Here we use Landsat data and the concept of ‘phenoregions’ – regions determined by their phenological similarity – across five sites in East Africa to ask which cues might drive variation in phenoregions at the local and regional scales.
We show that at the local scale, phenoregion distributions appear to be driven by a suite of variables, including topography, substrate, and land use. At the regional scale – looking from Kenya to South Africa – the drivers are similar but not identical, suggesting an interaction between local-scale drivers and regional climate patterns. This suggests that applying a broad ‘one size fits all’ phenology algorithm for dry ecosystems does not capture the local and regional variability across these ecosystems, lowering the predictive ability of models. Future modeling work should focus on striking a balance between model complexity and capturing the important functional diversity of semi-arid and savanna-type ecosystems.