Complexity Analyses of Sao Francisco River Streamflow: Influence of Dams and Reservoirs

Abstract

This study investigated the potential of methods from information science to detect hydrological alterations caused by human activity. In particular, the influence of the construction of a cascade of dams and reservoirs on the daily streamflow of the São Francisco River in Brazil is investigated by using the sample entropy (SampEn) method and its generalization, the multiscale entropy (MSE). A long daily-streamflow time series at locations upstream and downstream of a cascade of dams, recorded during the period 1929-2015, encompassing the Sobradinho dam (1979) and the Xingó dam (1994), were analyzed. It was found that reservoir operations changed the temporal variability of both the original and deseasonalized streamflow series by decreasing the degree of regularity, as indicated by higher SampEn values. In the MSE analysis, this was held for the small time scales, while larger scales reservoir operations induced a more regular streamflow regime (lower entropy values). The time variation of the streamflow regularity was also analyzed using the time-dependent sample entropy, which confirmed the preceding finding. In both the MSE and time-dependent SampEn analyses, the streamflow recorded at the São Francisco station, which is located upstream of dams and reservoirs, did not exhibit any change in entropy values due to reservoir operation, while the deseasonalized series showed a similar (although less pronounced) behavior as that for the downstream stations, indicating that, in addition to the reservoir operations, some other natural factors could have coinduced, such as a shift toward the lower regularity of the streamflow regime. These results provide the evidence that methods from information science can be useful in assessing hydrological alterations caused by human activities.