Developing a flood hazard analysis framework in the Cuvelai Basin, Namibia, using a flood model, remote sensing, and GIS.

Worldwide, more than 40% of all natural hazards, and about half of all deaths, are the result of flood disasters. In the Cuvelai River Basin (CRB), northern Namibia, flood disasters have increased dramatically over the past half-century, along with associated economic losses and fatalities. The increase in hydro-meteorological hazards such as floods are mainly attributed to intense urbanisation, changing land-use patterns and a changing climate. These hazards are exacerbated in semi-arid and data-sparse (SADS) regions such as the CBR, because of declining and/or non-existent hydro-meteorological infrastructure. In addition there is a lack of long-term continuous records that is needed to enhance the implementation of traditional flood risk management strategies, whether structural or non-structural, to mitigate hydro- meteorological hazards. This thesis developed a systematic framework that has quantified the uncertainties associated with the hydrological cycle that preclude the development of flood risk management strategies. The framework is based on free-data and open-data and software that is available online. It used remotely sensed data validated against ground-based observational data, where available. The particular components of the hydrological cycle that are assessed are: precipitation, surface runoff (discharge), surface water extent and surface water movement pathways (drainage networks). Hydrologic modelling was used to model the water fluxes in order to derive basin as well as flood characteristics of the study area. The framework can be used as a benchmark for the development of flood risk management policies that will enable SADS regions to mitigate the severe effect hydro-meteorological disasters in the Anthropocene. The flood hazard analysis framework (FHAF) developed for this study consists of two steps: (a) preliminary analysis and (b) hazard estimation. The preliminary analysis enable the development of a hydro-meteorological (floods and droughts) archive using different data sources as well as identifying where more analyses are needed to reduce uncertainty while hazard estimation provide the frequency and magnitude of the hazard. As a result of the growing concern about flood risk, identifying the extreme precipitation events that cause hydro-meteorological disasters is essential. Hence, the preliminary analysis step of FHAF developed a database (a). An up-to-date and broad analysis of the trends of hydro-meteorological events within the CRB was performed. The derived events were also validated against data from other sources. The risk estimation step involved components of the hydrologic cycle that are crucial in determining flood risk and that play an important role in enhancing uncertainty. Precipitation is one of these crucial components, to estimate and validate, especially in the trans-boundary SADS CRB. Four commonly used operational satellite-based rainfall estimation (SBRE) products were rigorously validated and inter-compared on monthly, seasonal, and annual timescales. Rainfall data from gauged stations were compared against SBREs as well as simulated data from a Regional Climate Model (RegCM4) model. Point-to-nearest-pixel and pixel-to-pixel methods were used to validate gauge data against high spatial resolution (0.25o) SBREs data for a period from 2008 to 2014. Validation was performed on a monthly, seasonally, and annual basis as well as taking the long-term mean, whilst error statistics were used to determine the accuracy of the SBREs when compared to the observed rainfall gauge values. Results indicated good statistical relationships between the ground-based gauge stations for some SBREs. Results will help to understand, and ultimately expand, our understanding of the climatologies within this SADS region and will also provide valuable information on the error structures of SBRE products that might be ingested into hydrologic models for water resource management. The results also help to quantify the improvements (bias correction) that are needed for these SBRE products to be useful for water resource and risk management applications. The second component, surface water pathways (drainage networks), is imperative to determine flood inundation extent, which relate to hazards. Also, accurate delineation of drainage networks is crucial for hydrological modelling and hydraulic modelling, and the comprehension of fluvial processes. Channels from topographic maps (blue lines) were compared to those from hydrologically corrected and uncorrected light detection and ranging (LiDAR) DEMs (digital elevation models), heads-up digitised channels from high-resolution digital aerial orthophotographs, field-mapped channels and auxiliary data. The maximum gradient deterministic eight (D8) GIS algorithm was applied to the corrected and uncorrected LiDAR DEMs using two network extraction methods: area threshold support and curvature/drop analysis. Results will aid national mapping agencies in SADS regions to modernise their national hydrography datasets and to account for changing land surface conditions that can affect channel spatial arrangements over time. The third component, deals with the amount, frequency, and magnitude of surface water runoff (discharge). Sustainable management of water resources as well as mitigating hydro- meteorological natural hazards such as flooding and drought requires the precise understanding of the spatio-temporal distribution of water especially in SADS regions where data from various global datasets are used to compensate. Results suggested that input data be ingested in hydrological models especially if the data are to be used especially for water resources estimations and for understanding flood-producing processes. The last component, surface water extent (flood inundation), was also estimated in this study. The mapping of spatial inundation patterns during flood events is important for environmental management and disaster monitoring. This study detected and compared the spatial extent of flood inundation at the peak of three major flood events (2008, 2009, and 2011) in the CRB. The study follows a multi-spectral and multi-sensor approach to identify the flood inundation for each flood event at peak modelled discharge. Results indicated that the quantification of flooding spatial extent can help to provide valuable information to FHAFs and hence potentially improve hydrologic prediction and flood management strategies in ungauged catchments. Furthermore, given the globally availability of satellite- based precipitation and river discharges, this proof-of-concept study can have substantial implications on flood monitoring and forecasting in ungauged basins throughout the globe