Five High-Resolution Canopy Fuel Maps Based on GEDI: A Basis for Wildfire Modeling in Germany

Open access publication under review. Visit <strong>this Earth Engine app</strong> to explore the data interactively. Abstract: Forest fuels are essential components of wildfire behavior and risk assessments, but difficult to quantify accurately. As fire occurs more frequently in new regions, demand for fuels information increases. This study develops a methodology for mapping canopy fuels over large areas at high spatial resolution, exclusively using open data. We propose a two-step approach that predicts five canopy fuel variables at point-level using measurements from NASA’s GEDI instrument, before interpolating them to high-resolution raster maps. Instead of conducting field measurements, we generate plot-level estimates for Canopy (Base) Height (CH, CBH), Cover (CC), Bulk Density (CBD), and Fuel Load (CFL) by segmenting point clouds from airborne laser scanning and processing tree-level metrics with allometric crown biomass models. Plot-level canopy fuel estimates serve as reference for Random Forest models fitted with GEDI measurements. Models are tuned to find optimal hyperparameters and cross-validated using k-fold Nearest Neighbor Distance Matching. Canopy fuel predictions at &gt;1.6 M GEDI points and multiple biophysical raster covariates are combined with a Universal Kriging method to produce maps for Germany at 20-meter resolution. The Random Forest model for CC performed best (R² = 0.76; RMSE = 9.95 %), followed by CBH (R² = 0.68; RMSE = 2.41 m) and CH (R² = 0.67; RMSE = 3.86 m). Accuracy was slightly lower for CFL (R² = 0.63; RMSE = 4.03 t/ha) and CBD (R² = 0.55; RMSE = 0.03 kg/m³). Interpolation was more precise for CC (R² = 0.65) than for the rest (0.24 &lt; R² &gt; 0.42). Compared to linear regression, additional residual Kriging improved results for all models. We show that canopy fuel maps can be derived from openly accessible data at country level without extensive field measurements. The proposed workflow has the potential to support regions where wildfire is an emerging issue and fuels information is scarce.