Tarpanelli, Angelica; Santi, Emanuele; Tourian, Mohammad J.; Filippucci, Paolo; Amarnath, Giriraj; Brocca, Luca, 2019, Daily River Discharge Estimates by Merging Satellite Optical Sensors and Radar Altimetry Through Artificial Neural Network, IEEE transactions on geoscience and remote sensing (2019). doi_10.1109/TGRS.2018.2854625,
URL: http://www.cnr.it/prodotto/i/393404

Thanks to the large number of satellites, the multimission approach is becoming a viable method to integrate measurements and intensify the number of samples in space and time for monitoring the earth system. In this paper, we merged data from different satellite missions, optical sensors, and altimetry, for estimating daily river discharge through the application of the artificial neural network (ANN) technique. ANN was selected among other retrieval techniques because it offers an easy but effective way of combining input data from different sources into the same retrieval algorithm. The network is trained in a calibration period and validated in an independent period against in situ observations of river discharge for two gauging sites_ Lokoja along the Niger River and Pontelagoscuro along the Po River. For optical sensors, we found that the temporal resolution is more important than the spatial resolution for obtaining accurate discharge estimates. Our results show that Landsat fails in the estimation of extreme events by missing most of the peak values due to its long revisit time (14-16 days). Better performances are obtained from Moderate Resolution Imaging Spectroradiometer (MODIS) and Medium Resolution Imaging Spectrometer. Radar altimetry provides results in between MODIS-TERRA and MODIS-AQUA at Lokoja, whereas it outperforms all single optical sensors at Pontelagoscuro. The multimission approach, involving optical sensors and altimetry, is found the most reliable tool to estimate river discharge with a relative root-mean-square error of 0.12% and 0.27% and Nash-Sutcliffe coefficient of 0.98 and 0.83 for the Niger and Po rivers, respectively.

Data from https://intranet.cnr.it/people/