Mapping forest canopy fuel parameters at European scale using spaceborne LiDAR and satellite data

Spatially explicit data on forest canopy fuel parameters provide critical information for wildfire propagation modelling, emission estimations and risk assessment. LiDAR observations enable accurate retrieval of the vertical structure of vegetation, which makes them an excellent alternative for characterising forest fuel structures. In most cases, fuel parameterisation has been based on Airborne Laser Scanning (ALS) observations, which are costly and best suited for local research. Spaceborne LiDAR acquisitions overcome the limited spatiotemporal coverage of airborne systems, as they can cover much wider geographical areas. However, they do not provide continuous geographical data, requiring spatial interpolation methods to obtain wall-to-wall information. We developed a two-step, easily replicable methodology to estimate forest canopy fuel parameters for the entire European territory, based on data from the Global Ecosystem Dynamics Investigation (GEDI) sensor, onboard the International Space Station (ISS). First, we simulated GEDI pseudo-waveforms from discrete ALS data about forest plots. We then used metrics derived from the GEDI pseudo-waveforms to estimate mean canopy height (Hm), canopy cover (CC) and canopy base height (CBH), for which we used national forest inventory data as reference. The RH80 metric had the strongest correlation with Hm for all fuel types (r = 0.96–0.97, Bias = −0.16-0.30 m, RMSE = 1.53–2.52 m, rRMSE = 13.23–19.75%). A strong correlation was also observed between ALS-CC and GEDI-CC (r = 0.94, Bias = −0.02, RMSE = 0.09, rRMSE = 16.26%), whereas weaker correlations were obtained for CBH estimations based on forest inventory data (r = 0.46, Bias = 0 m, RMSE = 0.89 m, rRMSE = 39.80%). The second stage was to generate wall-to-wall maps for the continent of Europe of canopy fuel parameters at a resolution of 1 km using a spatial interpolation of GEDI-based estimates for within-fuel polygons covered by GEDI footprints. GEDI observations were not available for some of the polygons (mainly Northern latitudes, above 51.6°N). In these cases, the parameters were estimated using random forest regression models based on multispectral and SAR imagery and biophysical variables. Errors were higher than from direct GEDI retrievals, but still within the range of previous results (r = 0.72–0.82, Bias = −0.18-0.29 m, RMSE = 3.63–4.18 m and rRMSE = 28.43–30.66% for Hm; r = 0.82–0.91, Bias = 0, RMSE = 0.07–0.09 and rRMSE = 10.65–14.42% for CC; r = 0.62–0.75, Bias = 0.01–0.02 m, RMSE = 0.60–0.74 m and rRMSE = 19.16–22.93% for CBH). Uncertainty maps for the estimated parameters were provided at the grid level, for which purpose we considered the propagation of individual errors for each step in the methodology. The final outputs, which are publicly available (https://doi.org/10.21950/KTALA8), provide a wall-to-wall estimation for the continent of Europe of three critical parameters for modelling crown fire propagation potential and demonstrate the capacity of GEDI observations to improve the characterisation of fuel models.