UNCERTAINTY PROPAGATION OF MANNING'S COEFFICIENT IN 2D NUMERICAL MODELS OF HYDRAULIC FLOW GENERATED BY HYPOTHETICAL DAM FAILURE
Free surface, land use and cover; failures of civil structures.
The propagation of uncertainty in hypothetical dam failure studies is associated with factors such as the determination of hydraulic flow parameters. The attribution of roughness coefficient values from the comparison of satellite images with tabulated intervals has a personal character, related to the experience of the analyst. It also occurs the attribution of the same roughness coefficient value for large areas, justified by the long stretches, quality of topographic data and computational expense of the simulations. These factors generate inaccuracies in the models' responses, such as the area inundated by the wave, crucial information required in Emergency Action Plans. Thus, the objective of this work was to analyze, in a probabilistic way, the effect of the roughness coefficient variation on the runoff generated by a hypothetical dam failure, characterizing the uncertainties from the inundation spot area, computational time, velocity and depth in cross sections along the runoff. The HEC-RAS 2D software was used to model the wave propagation generated by the hypothetical breach of the Três Marias dam with the Foz do Rio das Velhas (Brazil) as the stop criterion. The Manning values for the roughness coefficient were distributed in a normal and uniform manner, from the Monte Carlo method, respecting the Land Use and Occupancy data. It is observed that the results for area, velocity and depth did not follow normal distribution. The application of statistical test showed that there was no significant difference in the values of velocity and depth related to the two distributions used, but when compared to the deterministic scenario, the normal distribution presents greater variations of percentiles for the parameters under analysis.