Fig. 2 | Scientific Reports

Fig. 2

From: Assessment of spatial autocorrelation and scalability in fine-scale wildfire random forest prediction models

Fig. 2

Effect of removing predictor variables on burn severity prediction accuracy shown for the Hermit’s Peak/Calf Canyon fire. (a) Observed Sentinel-2 burn severity (dNBR), (b) predicted dNBR using all predictors (ECOSTRESS, weather, topography), (c) predicted dNBR using weather and topography only, (d) predicted dNBR using ECOSTRESS only, (e) predicted dNBR using topography only. R-squared of the predicted versus observed dNBR is shown in bottom right corner of each plot. Figure created using QGIS 3.18 (https://qgis.org/) software with the Bing Maps Satellite Imagery as a basemap (included with QGIS 3.18 software).

Back to article page