No abstract

Quantitative knowledge of river discharge measurements is essential for understanding coastal and estuarine dynamics and salinity variations. However, direct measurements are scarce for a large portion of rivers in Brazil. In this study, five simple models based on remote sensing and local rainfall datasets (MERGE) from the Ribeira de Iguape catchment are used to estimate the Valo Grande Channel (VGC) discharge on the southeastern coast of Brazil. These models use linear, quadratic, exponential and two different multiple linear regression methods. The predicted VGC discharge time series resulting from each model is compared with the estimated time series based on in situ data from the Water and Electric Energy Department (DAEE in Portuguese). The estimated time series presented reasonable results, with skills varying from 0.84 to 0.92 and Nash–Sutcliffe efficiency (NSE) indices varying from 0.54 to 0.75, with the highest values corresponding to the multiple linear regression models. This methodology allowed us to reproduce longer time series at a daily frequency, as well as monthly averages between 2000 and 2020.

Quantitative knowledge of river discharge measurements is essential for understanding coastal and estuarine dynamics and salinity variations. However, direct measurements are scarce for a large portion of rivers in Brazil. In this study, ve simple models based on remote sensing and local rainfall datasets (MERGE) from the Ribeira de Iguape catchment are used to estimate the Valo Grande Channel (VGC) discharge on the southeastern coast of Brazil. These models use linear, quadratic, exponential and two different multiple linear regression methods. The predicted VGC discharge time series resulting from each model is compared with the estimated time series based on in situ data from the Water and Electric Energy Department (DAEE in Portuguese). The estimated time series presented reasonable results, with skills varying from 0.84 to 0.92 and Nash-Sutcliffe e ciency (NSE) indices varying from 0.54 to 0.75, with the highest values corresponding to the multiple linear regression models. This methodology allowed us to reproduce longer time series at a daily frequency, as well as monthly averages between 2000 and 2020.

No abstract

No abstract