H52D-07
Estimation of surface water storage in the Congo Basin

Friday, 18 December 2015: 11:50
3011 (Moscone West)
Fiachra O'Loughlin1, Jeffrey C Neal1, Guy Schumann2, Edward Beighley3 and Paul D Bates4, (1)University of Bristol, Bristol, BS8, United Kingdom, (2)Remote Sensing Solutions, Inc., Pasadena, CA, United States, (3)Northeastern University, Department of Civil and Environmental Engineering, Boston, MA, United States, (4)University of Bristol, School of Geography, Bristol, United Kingdom
Abstract:
For many large river basins, especially in Africa, the lack of access to in-situ measurements, and the large areas involved, make modelling of water storage and runoff difficult. However, remote sensing datasets are useful alternative sources of information, which overcome these issues. In this study, we focus on the Congo Basin and, in particular, the cuvette central. Despite being the second largest river basin on earth and containing a large percentage of the world’s tropical wetlands and forest, little is known about this basin’s hydrology.

Combining discharge estimates from in-situ measurements and outputs from a hydrological model, we build the first large-scale hydrodynamic model for this region to estimate the volume of water stored in the corresponding floodplains and to investigate how important these floodplains are to the behaviour of the overall system. This hydrodynamic model covers an area over 1.6 million square kilometres and 13 thousand kilometres of rivers and is calibrated to water surface heights at 33 virtual gauging stations obtained from ESA’s Envisat satellite.

Our results show that the use of different sources of discharge estimations and calibration via Envisat observations can produce accurate water levels and downstream discharges. Our model produced un-biased (bias =-0.08 m), sub-metre Root Mean Square Error (RMSE =0.862 m) with a Nash-Sutcliffe efficiency greater than 80% (NSE =0.81). The spatial-temporal variations in our simulated inundated areas are consistent with the pattern obtained from satellites. Overall, we find a high correlation coefficient (R =0.88) between our modelled inundated areas and those estimated from satellites.