This study contributes with a multi-temporal estimation of soil losses by sheet erosion which represents the most probable path of water flow, sediments and its accumulation, using dynamic modeling. Besides that, the results of this study are a contribution to regional planning for medium and long term prognosis, in an area of population growth, simulating different scenarios. In this frame we chose the database from the upper course of the Rio Descoberto basin to apply the proposed methods. At the upper section of the Rio Descoberto basin, bordering the Federal District and the State of Goiás, as well as at most Brazilian river basins with inappropriate land use and occupation, referring specially to those rural sections with multiple use and chaotic urban expansion areas, one observes environmental impacts related to water erosion and sediment transport. Such conditions caused both silting and an accentuated quality and quantity loss of water availability in the Rio Descoberto reservoir, which is responsible for the supply of 2/3 of water for the Federal District produced by CAESB, the Environmental Sanitation Company of the Federal District. The first phase of this study was the proposition of an alternative topographic factor (the modified Length of Slope – LS – factor) for the Universal Soil Loss Equation (USLE), based on the theory of flow direction and accumulation in a river basin, in order to describe erosion processes associated to the sheet flow over slopes. The modified LS factor allowed a more precise estimation of this parameter at USLE as well as the identification of those sections most susceptible to erosion, using method D (infinite possible flow directions), to determine the direction of the water flow and sediments within the basin. The second phase of the work consisted of modeling the dynamics of changes from land use/occupation at one of the sub-basins from Rio Descoberto, the Ribeirão Rodeador. The dynamic model was based on the historical tendencies of land use/occupation within this sub-basin, as well as on the static variables available, that could be considered as evidence weights to justify the phenomenon analyzed. These tendencies were evaluated from 1984 on, followed by the simulation of pessimistic and optimistic tendency scenarios for the years 2003, 2008, 2013 and 2017. The results of modeling were calibrated and validated not only by simple human perception, but also by quantitative measurements of the similarity degree between the reality observed and the simulations. Finally, in order to generate a prognosis of soil losses within the Rodeador sub-basin, we applied the modified LS factor at USLE to the maps obtained during this phase. These studies and the results obtained show that the dynamic modeling is a potential tool to anticipate probable tendencies to occur in a river basin. Therefore the results could be a relevant instrument for strategic planning and for stakeholders at governmental and non-governmental level.