Computer programs applied to disaster simulation are widely used and widespread today, taking as input various data types, from specific to the application area to DTMs (Digital Terrain Models). This need for data input and, in particular, data related to relief is very relevant in prediction procedures for forecasting catastrophes, such as the failure of tailings mining deposit dams. Therefore, it is of fundamental importance to know and quantify the quality of this input data in question, in order to effectively serve this application. To this end, in this research, several tests were carried out, using as a reference for best results those obtained with the introduction of DTM from a LIDAR (Light Detection and Ranging) flight survey, this data being used as one of the primary and input into HEC-RAS (Hydrologic Engineering Center-River Analysis System). Subsequently, these same data had their spatial resolution degraded, that is, the pixel size increased, generating models with worse quality for new inputs and obtaining new simulation results of mining dam failures. The test area of the B1 Dam, located in Brumadinho-MG, Brazil, was used as a laboratory, where one of the biggest mining disasters in the world recently occurred and, for which, there are real data from the area affected by the dam collapse. The results obtained demonstrated that the use of an DTM with a spatial resolution of at least 2.5 meters or better, with DTM class A cartographic quality compatible with the most recent Brazilian standard, would guarantee reliable results.