Incorporação de filtro numérico de separação de escoamento na assimilação de dados para previsão de vazões utilizando modelagem hidrológica

Abstract

One of the key steps in streamflow forecasting using hydrologic models is data assimilation to adjust the model state variables and make the initial condition the closest possible to observed reality. One way to perform data assimilation is the use of empirical methods, utilizing the observed data to directly fix the variables of the hydrological model. An example of a model that uses this type of empirical technique is MGB-IPH, recently widely used for streamflow forecasting in Brazilian case studies. However, some problems may occur using this method of assimilation when the observed telemetry data mismatch the total volume of water in the basin, which may be reflected by a loss of quality in the forecasts. Thus, research about incorporating the use of baseflow separation filters is tested in this study to improve the empirical data assimilation, using the MGB - IPH model as a case study. The work is based on the hypothesis that it is possible to enhance the hydrograph recession predictions by incorporating an improved control of the amount of water that contributes to the baseflow at a location using digital flow separation filters. To test the hypothesis, some hindcasting experiments were issued in the Sao Francisco River basin in the HPP Tres Marias region for three rainy seasons between 2010 and 2013. Visually the results of the forecasts show that the problem noted in the data assimilation was removed using the baseflow filter. This benefit is also verified through the calculation of four performance measures, where most of the results pointed to improvements in the forecasts. In conclusion, the tested technique enables better performance in streamflow forecasting using the MGB - IPH model, and the use of numerical filters in data assimilation is presented as promising for other similar applications of similar data assimilation techniques