Tracking changes in forest composition, structure, and distribution over time is essential for developing effective conservation strategies and sustainable management practices in these ecologically sensitive regions. In this study, our objective was to conduct a diachronic analysis, comparing land cover and vegetation status in the Timekssaouine forest within the Central Plateau region of Morocco over a 20-year period. The aim was to analyze the spatiotemporal evolution of plant formations. We utilized satellite imagery, specifically Landsat images taken during the summer period (July) of 1999 and 2020, to provide a detailed observation of changes over time and space. Additionally, machine learning modeling using random forest (RF) was implemented to further explore the dynamics of change in the forest. The RF models developed achieved reasonable to good predictive performance, with AUC scores of between 0.67 and 0.80. Our findings revealed a concerning regression, with both the diachronic (59% of the forest area) and RF (35%) approaches highlighting extensive regression of the forest particularly in the cork oak formations at 9%, with notable de-densification across density classes between 1999 and 2020, a diachronic study. Dense cork oak and moderately dense strata were particularly affected, experiencing regressions of 455 ha and 1204 ha, respectively, during this period. Conversely, open and sparse strata expanded, primarily sourced from the dense and moderately dense strata, resulting in an overall regression rate of 60 ha/year. The dense cork oak strata were prevalent on steep slopes with deep, slightly acidic soil, while scattered and clear strata were observed in low-lying areas with shallow soils and a pH range from neutral to slightly basic. Autumn precipitation and amplified overgrazing intensity emerged as the pivotal factors influencing the categorization of forest formations in the study forest, impacting tree density levels and posing a significant threat to forest regeneration.