Random Forest Detection of charcoal driven vegetation loss in Ghana's Afram Plains.

This study examines long-term vegetation loss and land cover change in Ghana's Afram Plains, a key forest-savannah transition zone, using integrated remote sensing and machine learning techniques over a 34-year period (1986-2021). While national forest assessments typically attribute deforestation to cocoa farming, illegal logging, and mining, this research identifies charcoal production as the primary cause of vegetation loss in this non-cocoa region. By applying supervised classification, Normalised Difference Vegetation Index (NDVI) analysis, and Random Forest algorithms to Landsat imagery, the study documents a dramatic 75% reduction in vegetation cover from 72,647 hectares (20% of land area) in 1987 to 18,019 hectares (5%) in 2021. The results reveal complex temporal dynamics, with a median annual deforestation rate of 4,728 hectares only partially offset by natural regeneration of 2,837 hectares per year. Notably, the period of the 2017-2018 national energy crisis saw unprecedented deforestation rates of 54,953 hectares per year, directly linking energy policy challenges to accelerated vegetation loss. The methodology achieved exceptional accuracy (Out-of-Bag error rates below 0.04%), ensuring high confidence in these findings. While the Afram Plains currently experiences high deforestation pressure, the landscape also exhibits notable ecosystem resilience through natural regeneration. These findings advance understanding of deforestation dynamics in forest-savannah transition zones and emphasize the urgent need for integrated conservation strategies that regulate charcoal production activities and inform sustainable forest management policies in regions where charcoal production remains a major energy source.