Estimation of river discharge from spaceborne observations : assessment of different models

Abstract

Surface water storage and fluxes in rivers are currently poorly observed at the global scale, although they represent major components of the water cycle and heavily influence human life. Stream gauges can provide timely measurements which are crucial for flood prediction and water resource management. However, heterogeneous distribution of in situ networks together with rising worldwide concern on river systems, prompt new techniques for realizing river dynamics and discharge.

Satellite remote sensing of rivers is a rapidly emerging subdiscipline within hydrology. Satellite altimeters like ENVISAT and JASON can provide information on water surface elevation. Satellite imagery missions such as LANDSAT and MODIS can be used to estimate surface water extent, e.g., river width. Nonetheless, these techniques provide only one of the multiple measurements required to accurately estimate discharge from space. Furthermore, all of these measurements require in situ calibration to provide accurate discharge estimates. Therefore, a satellite mission which observes simultaneous, high-resolution measurements of parameters to estimate discharge without in situ calibration, is demanding. The future Surface Water and Ocean Topography (SWOT) mission is dedicated to provide terrestrial and ocean water surface elevations globally with high spatial resolution.

Given SWOT observables namely the river surface elevation, width and slope, periodic estimation of river discharge would be possible. There have been several models to estimate river discharge with respective to SWOT measurements. In this thesis, we analyze nine different discharge models including experimental models and the hydraulic based models. We used water level time series from satellite altimetry H, effective river width W obtained from satellite imagery and slope S to evaluate the models along four defined reaches of two rivers namely Niger and Po with different morphological and climatic conditions.

To assess the discharge models, we applied the Gauss-Helmert adjustment model for eight of these models to estimate unknown parameters of each model. The unknown parameters of each model depend on the characteristics of the defined reach along the rivers. The reason for the implementation of such algorithm is dealing with erroneous quantities on both sides of the discharge model or the observation equation in adjustment theory terminology. We tested and validated our results over the two mentioned rivers. Among the experimental models, namely models 2 to 5, model 3 with NSE of 0.92 for the river Niger and 0.44 for the river Po outperforms the other models. For the other models which have been developed by hydrologists, model 7 outperforms other models with NSE of 0.97 for the river Niger and 0.53 for the river Po.

In addition to the discharge estimation, average base zero flow height-which will not among the observable quantities of SWOT mission–is estimated. This quantity which leads to river depth estimation is evaluated by three models including model 1, 6 and 7 where the convergence was possible during the iterative scheme of the Gauss-Helmert model. Validation of the results is done for the river Po where we have surveyed cross-sectional data. The first model shows the average error of ±0.5 m between estimated discharge and surveyed value, although the standard deviation is respectively large in comparison to the other two models.