Soil erosion is a global environmental problem, which affects the sustainability and productivity of agriculture, the siltation of reservoirs, and the degradation of water quality. One of the main causes of soil erosion is the change of land use/cover (LULC) in a region. The impact of land use on soil erosion has been studied in many parts of Africa, but no published papers have been found to address this problem in Mozambique. From 1998 to 2018, significant land use change was observed in the Alto Molócuè watershed, having consequences on soil erosion and degradation of water quality, through sedimentation of the Alto Molócuè river (the main source of water for most of the population living on the namesake village), but no studies have been done to address this problem on the study area. The objectives of this study are to map the average annual soil loss and its spatial distribution in the watershed, assess land use changes using satellite images referring to the years 1998, 2008, and 2018, and predict future (10- year) erosion scenarios based on the predicted land use map for the year 2028. Landsat images from the years 1998, 2008, and 2018 were downloaded from Earth Explorer and processed in ArcGIS. Using the Kappa (K) coefficient and world land cover data set of 2013, it was possible to assess the accuracy of the image classification. The Land change modeler (LCM) tool from TerrSet was used to predict a future land use map for the year 2028 using Markov chain analysis and the Support Vector Machine (SVM) model. We used the Revised Universal Soil Loss Equation (RUSLE) to calculate and map past and future annual soil loss for the years 1998, 2008, 2018, and 2028. The accuracy of land use classification shows values of 87.9%, 80.4 %, and 85.2% with Kappa coefficients of 0.8, 0.85, and 0.97 for 1998, 2008, and 2018, respectively. Land use change detection showed that in the period (1998 to 2018) the decrease in shrubland and forest (4% and 11%), and the increase in urban areas and agricultural (3% and 11%) contributed to an increase in the mean erosion rate by16.07 t ha-1y-1(64.36%). The predicted land use map for 2028 shows that agricultural land and urban area will increase from 11% and 3% to 18% and 6% respectively, compared to the year 2018. On the other hand, forest and shrubland area will decrease from 43.8% and 33.8% in 2018 to 42% and 24% in 2028, respectively, with an average soil loss value of 29.64 t ha-1y-1, corresponding to an increase of 4.67 t ha-1y-1 (15%) compared to the year 2018. Remote sensing, LCM, and RULSE proved to be very important resources for assessing past and future soil erosion scenarios in the watershed. The findings from this research can be used by policymakers as a starting point to design land use and erosion control planning strategies in the watershed. The research approach can be adapted to address the same problems in other parts of Mozambique and Africa.

Keywords: GIS, Remote sensing, RUSLE, LULC, Land use modeling