The use of hydrological information to improve flood management-integrated hydrological modelling of the Zambezi River basin

摘要

The recent high profile flooding events – that have occurred in many parts of the world – have drawn attention to the need for new and improved methods for water resources assessment, water management and the modelling of large-scale flooding events. In the case of the Zambezi Basin, a review of the 2000 and 2001 floods identified the need for tools to enable hydrologists to assess and predict daily stream flow and identify the areas that are likely to be affected by flooding. As a way to address the problem, a methodology was set up to derive catchment soil moisture statistics from Earth Observation (EO) data and to study the improvements brought about by an assimilation of this information into hydrological models for improving reservoir management in a data scarce environment. Rainfall data were obtained from the FEWSNet Web site and computed by the National Oceanic and Atmospheric Administration Climatic Prediction Center (NOAA/CPC). These datasets were processed and used to monitor rainfall variability and subsequently fed into a hydrological model to predict the daily flows for the Zambezi River Basin. The hydrological model used was the Geospatial Stream Flow Model (GeoSFM), developed by the United States Geological Survey (USGS). GeoSFM is a spatially semi-distributed physically-based hydrological model, parameterised using spatially distributed topographic data, soil characteristics and land cover data sets available globally from both Remote Sensing and in situ sources. The Satellite rainfall data were validated against data from twenty (20) rainfall gauges located on the Lower Zambezi. However, at several rain gauge stations (especially those with complex topography, which tended to experience high rainfall spatial variability), there was no direct correlation between the satellite estimates and the ground data as recorded in daily time steps. The model was calibrated for seven gauging stations. The calibrated model performed quite well at seven selected locations (R2=0.66 to 0.90, CE=0.51 to 0.88, RSR=0.35 to 0.69, PBIAS=−4.5 to 7.5). The observed data were obtained from the National Water Agencies of the riparian countries. After GeoSFM calibration, the model generated an integration of the flows into a reservoir and hydropower model to optimise the operation of Kariba and Cahora Bassa dams. The Kariba and Cahora Bassa dams were selected because this study considers these two dams as the major infrastructures for controlling and alleviating floods in the Zambezi River Basin. Other dams (such as the Kafue and Itezhi-Thezi) were recognised in terms of their importance but including them was beyond the scope of this study because of financial and time constraints. The licence of the reservoir model was limited to one year for the same reason. The reservoir model used was the MIKE BASIN, a professional engineering software package and quasi-steady-state mass balance modelling tool for integrated river basin and management, developed by the Denmark Hydraulic Institute (DHI) in 2003. The model was parameterised by the geometry of the reservoir basin (level, area, volume relationships) and by the discharge-level (Q-h) relationship of the dam spillways. The integrated modelling system simulated the daily flow variation for all Zambezi River sub-basins between 1998 and 2008 and validated between 2009 and 2011. The resulting streamflows have been expressed in terms of hydrograph comparisons between simulated and observed flow values at the four gauging stations located downstream of Cahora Bassa dam. The integrated model performed well, between observed and forecast streamflows, at four selected gauging stations (R2=0.53 to 0.90, CE=0.50 to 0.80, RSR=0.49 to 0.69, PBIAS=−2.10 to 4.8). From the results of integrated modelling, it was observed that both Kariba and Cahora Bassa are currently being operated based on the maximum rule curve and both remain focused on maximising hydropower production and ensuring dam safety rather than other potential influences by the Zambezi River (such as flood control downstream – where the communities are located – and environmental issues). In addition, the flood mapping analysis demonstrated that the Cahora Bassa dam plays an important part in flood mitigation downstream of the dams. In the absence of optimisation of flow releases from both the Kariba and Cahora Bassa dams, in additional to the contribution of any other tributaries located downstream of the dams, the impact of flooding can be severe. As such, this study has developed new approaches for flood monitoring downstream of the Zambezi Basin, through the application of an integrated modelling system. The modelling system consists of: predicting daily streamflow (using the calibrated GeoSFM), then feeding the predicted streamflow into MIKE BASIN (for checking the operating rules) and to optimise the releases. Therefore, before releases are made, the flood maps can be used as a decision-making tool to both assess the impact of each level of release downstream and to identify the communities likely to be affected by the flood – this ensures that the necessary warnings can be issued before flooding occurs. Finally an integrated flood management tool was proposed – to host the results produced by the integrated system – which would then be accessible for assessment by the different users. These results were expressed in terms of water level (m). Four discharge-level (Q-h) relationships were developed for converting the simulated flow into water level at four selected sites downstream of Cahora Bassa dam – namely: Cahora Bassa dam site, Tete (E-320), Caia (E-291) and Marromeu (E-285). However, the uncertainties in these predictions suggested that improved monitoring systems may be achieved if data access at appropriate scale and quality was improved.