ABSTRACT:

Radar altimetry is useful for monitoring water surface dynamics in an era of satellite remote sensing. One of the uses of satellite altimetry is to evaluate/calibrate the hydrodynamic model or to improve river-related variables by optimization/assimilation methods. However, with the discrete nature of the hydrodynamic models, comparing simulated water surface elevation (WSE) with satellite altimetry is challenging due to the difficulty in the exact matching of representative locations by satellite altimetry virtual station (VS) and model grids. We developed an algorithm to assign the VSs to the river network and labeled each VS with a flag according to the characteristics of the allocation to the river network. We presented an altimetry mapping procedure (AltiMaP) to allocate VS locations given in the HydroWeb database (https://hydroweb.theia-land.fr/) to the MERIT Hydro (http://hydro.iis.u-tokyo.ac.jp/~yamadai/MERIT_Hydro/) river network. Each VS was assigned to the nearest river in the MERIT Hydro considering the geometric distance. During the altimetry mapping, flags were added to each VS depending on the method of allocation. The flags were added in cases if the latitude-longitude of VSs is identified as a single-channel river (Flag 10); as a land pixel and allocated to the nearest single-channel river (Flag 20); as a multi-channel river (Flag 30); and as an ocean pixel and allocated to the nearest river (Flag 40). We allocated more than 12000 VSs in which most of the VSs were allocated as Flag 10 which accounts for 71.7% of all VSs. Moreover, most of the Flags 10 and 20 are located up and mid-streams whereas Flags 30 and 40 were in downstream river reaches. The biased VSs which have a large elevation difference between the mean observed WSE with the model elevation were approximately 0.8% of the total VSs and were mostly seen in narrow rivers at high elevations. The error of simulated WSE by a global river hydrodynamic model was reduced by 10.6% by employing the AltiMaP against the allocation VSs by converting longitude and latitude information to the coarse resolution river network. Therefore, the allocation of VSs using the method developed in this study (AltiMaP) to the river network enhanced the comparability of the simulated WSE by the global hydrodynamic model against satellite altimetry. The AltiMaP algorithm and data will be useful for the global hydrological community.

ABSTRACT:

Quantifying continental-scale river discharge is essential for understanding the terrestrial water cycle, which is susceptible to errors due to lack of observations or limitations in hydrodynamic modelling. Data assimilation (DA) methods are increasingly utilized to estimate river discharge combined with the emerging amount of river-related remote sensing data (e.g., water surface elevation, water surface slope, river width, flood extent, etc.). However, direct comparison of simulated water surface elevation (WSE) with the satellite altimetry data remains still challenging (i.e., large bias between simulations and observation, uncertainty in parameters, etc.) and can introduce large errors when assimilating satellite observations to hydrodynamic models. We performed several experiments, namely, direct, anomaly, and normalized value assimilations, to investigate the capability of DA to improve the river discharge with the current limitations of hydrodynamic modelling. The hydrological data assimilation was performed using a physically-based empirical localization method in the Amazon Basin. We used satellite altimetry data from ENVISAT, Jason 1 and Jason 2 for this study. The direct DA was used as the baseline of the assimilations, but it was subjected to errors due to the biases in the simulated WSE. As an alternative to direct DA, we used anomaly DA to overcome the errors due to the biases in the simulated WSE. In addition, we found that the modelled WSE distribution and the observed distribution differed considerably (i.e., amplitude differences, seasonal flow variations, distribution skewness due to limitations of hydrodynamic models, etc.). Therefore, a normalized value DA was performed to realize better discharge estimation. River discharge improved in 24%, 38%, and 62% of the stream gauges in the direct, anomaly, and normalized value assimilations compared to simulations without DA. The normalized value assimilation performed better in estimating river discharge given the current limitations of hydrodynamic models. Most of the gauges within the river reaches with satellite observations accurately estimated the river discharge with Nash-Sutcliffe Efficiency (NSE) > 0.6. The amplitudes of WSE were improved in the normalized DA experiment. Furthermore, in the Amazon Basin, normalized assimilation (median NSE=0.47) can improve river discharge estimation over the open-loop simulation with global hydrodynamic modeling (median NSE=0.13). River discharge estimation by direct DA methods can be improved by 7% of NSE by calibrating river bathymetry. Moreover, the direct DA approach outperforms the other DA methods when the runoff is considerably (50%) biased. The uncertainties in hydrodynamic modelling (i.e., river bottom elevation, river width, simplified floodplain dynamics, rectangular cross-section assumption, etc.) should be improved for better estimation of river discharge by assimilating satellite altimetry. This study will contribute to developing a global river discharge reanalysis product that is consistent spatially and temporally.

ABSTRACT:

This include the global river discharges estimated using physically-based adaptive empirical localization method.