Remote Sensing Satellite Data Preparation for Simulating and Forecasting River Discharge

Simulating and forecasting river discharge using satellite information is one of the most economical ways of measuring discharge, especially in remote areas where in situ gauges are too expensive to install, maintain, and operate. However, satellite signals are affected by climatic factors, such as clouds, mist, dust storms, and smoke from large forest fires. Hence, a reliable method of preparing and treating space-based signals is necessary. In this paper, a method is introduced to remove inaccurate and out-of-range signals having a low correlation with the in situ discharge measurements, using a combination of data classification and outlier detection procedures. To forecast future river discharge using space-based signals, the signal data set should not contain any gaps. Therefore, we introduced a procedure for the missing signals to be estimated by a model calibrated using the measured discharge. This procedure is illustrated using a case study for the White River, near Boston, MA, USA, that involved using a ddata set of three years' daily signals of the passive microwave information from the Advanced Microwave Scanning Radiometer for Earth Observing System satellite, which are obtained from the difference between thermal emission of wet and dry land surfaces, and the in situ discharge measurements of the river. The best model was selected from 6200 available models using the Pareto front. The optimum model detected 339 samples as outliers and eliminated them from the data set. Subsequently, the signals were calibrated with the in situ information, and the results indicated the superior accuracy in simulating White River discharge using satellite information.