PubAg

Main content area

River discharge estimation at daily resolution from satellite altimetry over an entire river basin

Author:
Tourian, M.J., Schwatke, C., Sneeuw, N.
Source:
Journal of hydrology 2017 v.546 pp. 230-247
ISSN:
0022-1694
Subject:
basins, covariance, databases, dynamic models, gauges, hydrologic models, rivers, satellites, stochastic processes, time series analysis, variance, watersheds, Niger
Abstract:
One of the main challenges of hydrological modeling is the poor spatiotemporal coverage of in situ discharge databases which have steadily been declining over the past few decades. It has been demonstrated that water heights over rivers from satellite altimetry can sensibly be used to deal with the growing lack of in situ discharge data. However, the altimetric discharge is often estimated from a single virtual station suffering from coarse temporal resolution, sometimes with data outages, poor modeling and inconsistent sampling. In this study, we propose a method to estimate daily river discharge using altimetric time series of an entire river basin including its tributaries. Here, we implement a linear dynamic model to (1) provide a scheme for data assimilation of multiple altimetric discharge along a river; (2) estimate daily discharge; (3) deal with data outages, and (4) smooth the estimated discharge. The model consists of a stochastic process model that benefits from the cyclostationary behavior of discharge. Our process model comprises the covariance and cross-covariance information of river discharge at different gauges. Combined with altimetric discharge time series, we solve the linear dynamic system using the Kalman filter and smoother providing unbiased discharge with minimum variance. We evaluate our method over the Niger basin, where we generate altimetric discharge using water level time series derived from missions ENVISAT, SARAL/AltiKa, and Jason-2. Validation against in situ discharge shows that our method provides daily river discharge with an average correlation of 0.95, relative RMS error of 12%, relative bias of 10% and NSE coefficient of 0.7. Using a modified NSE-metric, that assesses the non-cyclostationary behavior, we show that our estimated discharge outperforms available legacy mean daily discharge.
Agid:
5609904