Discharge Estimates for Ungauged Rivers Flowing over Complex High-Mountainous Regions based Solely on Remote Sensing-Derived Datasets

Abstract

Reliable information about river discharge plays a key role in sustainably managing water resources and better understanding of hydrological systems. Therefore, river discharge estimation using remote sensing techniques is an ongoing research goal, especially in small, headwater catchments which are mostly ungauged due to environmental or financial limitations. Here, a novel method for river discharge estimation based entirely on remote sensing-derived parameters is presented. The model inputs include average river width, estimated from Landsat imagery by using the modified normalized difference water index (MNDWI) approach; average depth and velocity, based on empirical equations with inputs from remote sensing; channel slope from a high resolution shuttle radar topography mission digital elevation model (SRTM DEM); and channel roughness coefficient via further analysis and classification of Landsat images with support of previously published values. The discharge of the Lhasa River was then estimated based on these derived parameters and by using either the Manning equation (Model 1) or Bjerklie equation (Model 2). In general, both of the two models tend to overestimate discharge at moderate and high flows, and underestimate discharge at low flows. The overall performances of both models at the Lhasa gauge were satisfactory: comparisons with the observations yielded Nash–Sutcliffe efficiency coefficient (NSE) and R2 values ≥ 0.886. Both models also performed well at the upper gauge (Tanggya) of the Lhasa River (NSE ≥ 0.950) indicating the transferability of the methodology to river cross-sections with different morphologies, thus demonstrating the potential to quantify streamflow entirely from remote sensing data in poorly-gauged or ungauged rivers on the Tibetan Plateau.

Long-Term Discharge Estimation for the Lower Mississippi River Using Satellite Altimetry and Remote Sensing Images

“…Our method differs from classical state of the art AHG approaches [38,41,43] by the construction of the river bathymetry which uses as many observational data from satellite altimetry and remote sensing images as possible and a hypsometric function which was originally developed for lakes [26]. Our results for 16 cross-sections show that the hypsometry can also be fitted to surface area and satellite altimetry observations over large alluvial rivers to increase the number of water level observations when the river depth is estimated correctly.…”

“…The roughness coefficient is estimated using multiple adjustment factors, a method that has been well established in several studies [38,43,47,67]. Most of the adjustment factors are set to 0, because we select only uniform sections without irregularities such as eroded banks, abrupt changes in CS size, or obstructions based on the DAHITI water occurrence mask.…”

“…s is calculated using a segment of the river centerline around the selected CS that has a length of 20 times the maximum width, since this distance is likely to include at least one meander wavelength [68][69][70]. This method was also used in other related studies, e.g., for the Yangtze River [47], the Lhasa River [43] and two siberian rivers [38]. Except for m, we use constant adjustment factors for all cross-sections.…”

“…The NRMSE is 23% using depth information from topographic maps and 20% after calibration of the parameters of the Manning formula. Kebede et al [43] estimate discharge time series for the Lhasa River using only Landsat and SRTM data. The NRMSE is in a range of 25.7% to 41.4% and the NSE is between 0.886 and 0.956.…”

“…Based on the predicted geometry, the velocity is estimated with the Manning Formula. The required roughness coefficient is estimated similar to other studies using adjustment factors [38,43,47]. The flow gradient is derived from satellite altimetry measurements at multiple stations along the river.…”