Modelling vegetation trends using biophysical and demographic datasets in the savanna of Burkina Faso

Abstract

Savanna represents an important vegetation biome in West Africa, providing food and services to people, and habitat for large amount of vegetation and animal species. However, this biome knows a rapid degradation of its vegetation cover driven by anthropogenic and climatic stressors. Monitoring and modelling vegetation change are relevant to safeguarding forest and combat land degradation. This study explored the use of biophysical and demographic datasets to model vegetation trends in the savanna of Burkina Faso. For that, vegetation trends were detected from 2001 to 2020 with the Mann-Kendall’s trend test. Random Forest (RF), Super Vector Machine (SVM) and Artificial Neural Network (ANN) algorithms were used to model vegetation trends driven by biophysical and demographic variables. The result revealed that non-significant vegetation trends were prominent (73%) in the study area during 2001-2020, while greening and degradation trends characterised 13% and 14% of the pixels, respectively. RF was found superior to SVM and ANN in the modelling of vegetation trends categories with overall accuracy (Kappa index) above 0.80 (0.70). The study provided sound information that can support the development of efficient strategies to combat land degradation.