Use of satellite remote sensing to understand land surface hydrology - human systems interactions over data sparse and unresolved hydrology regions: the Zambezi River Basin case study.

The exchange of water between hillslopes, river channels and floodplain can be quite complex and the difficulty in capturing the mechanisms behind it is exacerbated by the impact of human activities such as reservoir operations and irrigation schemes. Over the last decade, large-scale water management models have emerged toward improving the representation of human activities into earth system models and understanding hydrology-atmosphere dynamics both over scales that extend beyond typical operational water management models and also over data sparse regions. This large scale representation of water management has been shown to successfully represent the associated major spatial and temporal redistributions of water resources. While satellite observations of lakes and reservoirs have become more widely available, it remains difficult to validate and evaluate such large scale integrated simulations with such observations, because it is difficult to dissociate errors from hydrologic simulations with errors in the water management model.

The Zambezi River Basin has a complex hydrology. Major reservoirs have been built for hydropower operations leading to an inherent management of seasonal floods to some extent. According to satellite observations, the reservoir storage is maintained throughout the year and is not significantly dropping to store the seasonal floods. On the other hand, flow observations suggest that seasonal floods are maintained downstream of the dams supporting the associated complex ecosystem. Water management models have been built over selected tributaries toward investigating water-energy-food dynamics. However the models are mostly informed with observed inflow due to large errors in current hydrological representation of the region. The lack of observed inflow into reservoirs across the basin has also limited the development of a basin wide water management model to better understand the actual seasonal flood regimes.

We hypothesize that the large errors in hydrologic simulations can be attributed to the lack of representation of inundated floodplains upstream of the large reservoirs. We use satellite borne lake surface level variations which we process into volume variations toward calibrating the large-scale water management model. The simple reservoir storage management model and its ability to meet the imposed reservoir operations when forced by hydrologic simulations, shed light on the errors in hydrologic simulations despite the lack of observed inflow data. The approach results in a range of water balance errors which is used to evaluate and calibrate a 2-D inundation model and especially the timing and volume of floods that feed into the water management model. We then evaluate the overall calibrated integrated water model and changes in land-atmosphere fluxes.