The development of pan-African food forecasting and the exploration of satellite-based precipitation estimates

Abstract

The main objective of this PhD is to contribute to the development of a pan-African flood forecasting system in order to enhance flood forecasting for the whole of Africa. In view of the dimension and complexity of this goal, this research focused on particular aspects of flood forecasting, considered of key importance to devise a technically-sound flood forecasting system that could potentially be implemented operationally to facilitate international civil protection services. Emphasize was on four focal areas of research 1) explore the current situation on flood forecasting and early warning in Africa, 2) investigate the feasibility of probabilistic, medium-ranged flood forecasting in Africa, 3) explore satellite-based precipitation estimates (SRFE) for hydrological modeling, and 4) test a pan-African flood forecasting system (AFFS). The methods used in this research varied, adjusted to the type of data available and the specific objectives of each study. For the review on the current state of flood forecasting in Africa, research published in scientific literature and on institutional websites were scouted, and supplemented by data collected from a questionnaire. For the second focal point, the methods of the European Flood Awareness System (EFAS) were tested on an equatorial, semi-arid river basin. This included the set-up and optimization of LISFLOOD for the target catchment, the calculation of hindcasts for two large historical flood events using various meteorological input data as well as a statistical evaluation using various skill scores fitted for probabilistic forecasting. The ability of SRFE to replicate hydrologically-relevant rainfall characteristics was assessed through ground truthing at three spatially aggregated levels using a range of standard statistical methods and visual inspection. The usefulness of SRFE for hydrological modelling, as well as the necessity of bias correction and SRFE-specific calibration, was assessed by forcing LISFLOOD with various SRFE over four test basins holding various climatic and geomorphologic characteristics. The outcome was compared against discharge measurements using a novel performance indicator. For the final study, LISFLOOD was set up, calibrated and validated for the whole of Africa. Ensemble hydrological predictions were reproduced for a whole year and flood signals were derived through the application of critical thresholds. The predictive capability was investigated by comparing the results against ground measurements and reports of various flood archives. Among the main outcome of this PhD a network of more than 50 African institutions were identified, which are currently working on flood management issues and expressed further need for medium-ranged flood forecasts and could hence be potential users of AFFS. Additionally, SRFE proved as valuable source for hydrological modeling, provided that the right SRFE is chosen considering the given climatic and geomorphologic conditions as well as the type of hydrometeorological application. Further, the methods developed for EFAS proved to be transferable to the African continent providing that the hydrological model is adjusted to fit different hydro-climatic conditions and meteorological measurements are replaced by SRFE. Lastly, from a technical perspective, AFFS shows a large potential as an operational pan-African flood forecasting system as it detected around 70 % of reported flood events correctly.