USING SENTINEL-2 AND STACKING REGRESSORS FOR FOREST HEIGHT ESTIMATION

The climate change impacts can also be seen in the growing number of wildfires. Consequently, forest management and the updating of forest inventories become more important in the wildfires' avoidance. Measuring the Forest Height (FH) is an important activity in forests monitoring, since the FH can serve as a proxy variable of other parameters, as the aboveground biomass. Normally, FH is mapped through field campaigns or airborne laser scanning missions. However, these approaches do not offer the scalability needed and they are expensive. Therefore, multispectral data from Remote Sensing can be used for producing regional maps of FH. Here it is proposed a regionally calibrated Regression Methodology that uses multispectral data from Sentinel-2 and a Stacking Regressor for mapping the FH in Mediterranean forests. For a total of 17 regions across Portugal, Spain, and California, a R-2 between 43.71% and 72.85% and a RMSE between 0.85m and 4.03m.