The increasing demand for goods and services by the increasing human population exerts immense pressure on ecosystem and leads to clearing of natural habitats and results in a decline of the capacity to provide ecosystem services (ESs). Understanding the change in ESs and their response to landscape pattern changes provides important insight for designing management interventions. This study was aimed at analyzing the spatio-temporal changes in ESs from 1972 to 2047 and their linkage with landscape characteristics in the Beressa watershed. The land use/land cover (LULC) layers of 1972, 1987, 2002, and 2017 were mapped from Landsat images using the maximum likelihood classification (MLC) algorithm, and projected to 2032 and 2047 by Cellular Automata and Markov Chain (CA-Markov) model. Landscape structural change was then analyzed using FRAGSTATS tool. The Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) model was employed to evaluate the state of the ESs indicators. Global Moran's I and Getis-ord Gi* spatial statistics were employed to identify the ESs spatial clustering and hotspots. The impact of individual landscape metrics and topographic variables was analyzed using Partial Least Squares Regression (PLSR) model. A post-classification comparison indicated that, natural vegetation, grassland, barren land and waterbody declined by 89.9%, 67.9%, 67.8 and 15.9%, respectively during 1972-2017. Conversely, plantation increased by 692.1% followed by human settlement (138%) and farmland (21.8%). A similar trend is likely to continue for 2032 and 2047, with a slight decline of plantation in 2047. Analysis of landscape metrics revealed that, number of patches (NP) for plantation, barren land, settlement and grassland increased by 171.4%, 69.7%, 65.8% and 28.6%, respectively. In contrast, natural vegetation, farmland and waterbody declined by 53.1%, 46.3% and 33.9%, respectively during 1972-2017. Future predictions showed a declining trend in the number of patches for all LULC types. An increasing trend in the largest patch index (LPI) and mean patch size (MPS) for farmland, plantation, and settlement categories was observed across all years, suggesting intensification of human activities in the landscape. Consequently, natural habitats have declined and become fragmented. These changes resulted in the significant loss of ESs provisions at varied rate. The mean carbon storage (CS), habitat quality (HQ) and crop production (CP) decreased by 23.3%, 20% and 14%, respectively, while the mean of sediment export (SE) and water yield (WY) increased by 45% and 27.4%, respectively during 1972-2017. Analysis of the projections for the next three decades (2017-2047) suggested a decrease in CS and HQ by 30.8% and 6.3%, and an increase in SE, WY and CP by 69.2%, 13.3% and 2.3%, respectively. The result of spatial clustering indicated hotspot areas declined, while coldspot areas increased during the rnstudy period. The spatio-temporal variations in ESs appeared to be determined by human activities, especially expansion of artificial plantation, human settlements, agricultural land. The most important landscape structure variables accounted for ESs change were percentage of landscape (PLAND), MPS and LPI of anthropogenic habitats, suggesting increased habitat modifications. Restoring degraded natural habitats is crucial to maintain ES supply in the study area.