Habitat fragmentation is a threat to conservation of biodiversity hotspots in the Morogoro region, Tanzania. However, on-going research on fragmentation has not kept pace with temporal lapses and how individual species respond to habitat transformation and heterogeneity. This study sought to model spatial and temporal fragmentation patterns. Cloud free multi-temporal Landsat imagery with similar spectral resolution were acquired in the same season in 1975, 1995 and 2012. The images were used to characterize the biophysical landscape characteristics and a range of metrics used to quantify the magnitude of fragmentation. Patches and classes in the landscape were assessed using Fragstats, a spatial statistics program useful in computing landscape metrics. Results show that patch number was higher in dense forest and woodland than in less dense forest and grassland in 1975, 1995 and 2012 while the interspersion Juxtaposition Index (IJI) ranged between 0 (for clumped patches) and 100 (for grassland). In 1975 and 1995, the grassland habitat had the highest IJI while in 2012 less dense forest had the highest IJI. The Games-Howell test showed a significant fragmentation trend in less dense forests class (p≤0.05). Generally, the study indicates a high fragmentation pattern in the vulnerable tropical eastern arc mountain region of East Africa. This finding demonstrates the value of remotely sensed data in understanding the impact of anthropogenic processes on natural landscape transformation. Furthermore, the study provides a basis for informed conservation policy design and implementation in the region.