Biomass Prediction Using Machine Learning Techniques In Google Earth Engine: A Case Study Of The Azrou Forest In The Middle Atlas Mountains, Morocco

In the context of climate change, forests are a vital source of ecosystem services for humankind, acting primarily as carbon sinks. The aim of this study is to use the machine learning algorithms available in the Google Earth Engine (GEE) to predict the above-ground biomass of the Azrou forest in the Middle Atlas Mountains of Morocco. After a literature review, the work consisted of characterizing the natural features through Land Use Land Cover analysis (LULC) and forest stand types. The accuracy of the forest stand type classification was assessed at 81.55% using the kappa index. Analysis of vegetation cover time series data, derived from NASA imagery and MODIS, was carried out, focusing on four key indices: NDVI (Normalized Difference Vegetation Index), EVI (Enhanced Vegetation Index), LAI (Leaf Area Index), and FPAR (Fraction of Photo synthetically Active Radiation). The study predicted biomass using the Random Forest machine-learning model, implemented in GEE with JavaScript. NASA/ORNL biomass data for 2010 served as the dependent variable, while LULC, elevation, and the four indices were used (selected summer period) as independent explanatory variables. In addition, forest stand types were integrated to calculate total biomass for specific stand types and for the study area as a whole for the years 2015, 2020 and 2024. In 2024, the predicted biomass is 461,587 tons, compared with 501,172 tons in 2010. The biomass median values by species were 29 tons/ha for pure Atlas cedar (Cedrus atlantica Manetti), 24 tons/ha for pure holm oak (Quercus ilex) and 31 tons/ha for a mixture of Atlas cedar and holm oak. The results highlight challenging conditions for the Azrou forest, with a notable decline in biomass over the study period. These results will serve as a basis for future forestry planning in the context of climate change.