Monthly Streamflow Forecast in the Xingu River Basin - Eastern Amazon

Abstract

Empirical rainfall-streamflow models have been increasingly used in recent decades, due to the unavailability of input data from conceptual models, and the reliability, speed, and less complexity of these models. In this context, the Principal Component Regression technique (PCR) to simulate the monthly average streamflow of eight river stations in the Xingu River Basin (XRB) was applied, which belongs to the large Amazon basin. The XRB has the Belo Monte hydroelectric plant, as well as important areas of environmental preservation. The degree of relationship between rainfall and streamflow, input and response of the rainfall-streamflow model based on PCR is shows in the research and the respective degrees of lag with predictive efficiency. The PCR showed good results in the simulation of the monthly streamflow in all the selected stations, characterizing well the dynamics of the time series with excellent results in the dry periods (May to October) and a tendency to slightly underestimate in the rainy periods (November to April). These results, using the rainfall observed for streamflow simulation in the XRB, allow us to conclude that a good climate forecasting system for seasonal rainfall can infer an important predictive degree for streamflow up to three months in advance.

Keywords:
climate dynamics; rainfall-streamflow model; principal component regression