Framing the Forest: A Comparative Analysis of Google Earth Engine Classifiers for Accurate Species Extraction

Forests are critical ecosystems that provide multiple ecological services; therefore, accurate classification of their species is crucial for effective forest management and conservation. In this study, we compared and evaluated the performance of several Google Earth Engine (GEE) classifiers for extracting main forest species in Portugal. For this purpose, a time series of Sentinel-2 images was investigated for 2018 using a two-stage supervised classification workflow, which involved Random Forest, Naive Bayes, Classification and Regression Trees, Gradient Boosting Trees, and Support Vector Machine classifiers. In the first stage, Sentinel-2 images were analyzed to provide a binary land cover map from where the forest layer was extracted. The layer obtained was then used for masking the input features to the second stage. The latter stage discriminated six forest species by analyzing several input feature configurations. These included multiple spectral indices selected from the time series, as well as topographic factors. The performance of each classifier in species discrimination was assessed by comparing accuracy metrics obtained from the selected input configurations. The highest overall accuracy of the map was estimated at 68.75%, with a kappa of 0.62, for the Support Vector Machine classifier. The study's findings highlight the importance of algorithm selection using the GEE platform for accurate forest species classification in similar landscape characteristics, informing researchers, conservation practitioners, and forest managers for improved management and conservation strategies.