UAV-based instantaneous fuel load measurements in southern African savannas

In a world of rising atmospheric carbon levels, globally occurring natural fires are a substantial source of greenhouse gases. Savanna biomes are the most frequently burned landscapes, and as such constitute the major source of greenhouse gases from biomass burning. To date, emission estimates of these fires have been mostly derived from coarse scale inventories neglecting fine spatial variability in burned area and fuel loads. Although improvements have been made in recent years in remotely-sensed burned area data, large uncertainties remain in fuel load data. We have carried out extensive fieldwork in two savanna-dominated locations in southern Africa: one lower rainfall region (~660 mm year^-1) in North-Western Ngamiland (Botswana) and one higher (~940 mm year^-1) in Niassa National reserve (Mozambique). We measured fuel loads and used high spatial resolution (5cm) images taken using an UAV-mounted multispectral camera in combination with meteorological reanalysis data from the ERA-5 land dataset to model these. For fine surface fuel classes which constitute the main source of emissions in savanna fires, the model could reproduce AGB measurements (R² values of 0.845 and 0.762 for grass and total fine fuel, respectively) across both high and low rainfall areas. The improved characterisation of fuel loads can be used to improve large-scale emission estimates with unprecedented spatial detail.