Application of open-access and 3rd party geospatial technology for integrated flood risk management in data sparse regions of developing countries

Floods are one of the most devastating disasters known to man, caused by both natural and anthropogenic factors. The trend of flood events is continuously rising, increasing the exposure of the vulnerable populace in both developed and especially developing regions. Floods occur unexpectedly in some circumstances with little or no warning, and in other cases, aggravate rapidly, thereby leaving little time to plan, respond and recover. As such, hydrological data is needed before, during and after the flooding to ensure effective and integrated flood management. Though hydrological data collection in developed countries has been somewhat well established over long periods, the situation is different in the developing world. Developing regions are plagued with challenges that include inadequate ground monitoring networks attributed to deteriorating infrastructure, organizational deficiencies, lack of technical capacity, location inaccessibility and the huge financial implication of data collection at local and transboundary scales. These limitations, therefore, result in flawed flood management decisions and aggravate exposure of the most vulnerable people. Nigeria, the case study for this thesis, experienced unprecedented flooding in 2012 that led to the displacement of 3,871,53 persons, destruction of infrastructure, disruption of socio-economic activities valued at 16.9 billion US Dollars (1.4% GDP) and sadly the loss of 363 lives. This flood event revealed the weakness in the nation’s flood management system, which has been linked to poor data availability. This flood event motivated this study, which aims to assess these data gaps and explore alternative data sources and approaches, with the hope of improving flood management and decision making upon recurrence. This study adopts an integrated approach that applies open-access geospatial technology to curb data and financial limitations that hinder effective flood management in developing regions, to enhance disaster preparedness, response and recovery where resources are limited. To estimate flood magnitudes and return periods needed for planning purposes, the gaps in hydrological data that contribute to poor estimates and consequently ineffective flood management decisions for the Niger-South River Basin of Nigeria were filled using Radar Altimetry (RA) and Multiple Imputation (MI) approaches. This reduced uncertainty associated with missing data, especially at locations where virtual altimetry stations exist. This study revealed that the size and consistency of the gap within hydrological time series significantly influences the imputation approach to be adopted. Flood estimates derived from data filled using both RA and MI approaches were similar for consecutive gaps (1-3 years) in the time series, while wide (inconsecutive) gaps (> 3 years) caused by gauging station discontinuity and damage benefited the most from the RA infilling approach. The 2012 flood event was also quantified as a 1-in-100year flood, suggesting that if flood management measures had been implemented based on this information, the impact of that event would have been considerably mitigated. Other than gaps within hydrological time series, in other cases hydrological data could be totally unavailable or limited in duration to enable satisfactory estimation of flood magnitudes and return periods, due to finance and logistical limitations in several developing and remote regions. In such cases, Regional Flood Frequency Analysis (RFFA) is recommended, to collate and leverage data from gauging stations in proximity to the area of interest. In this study, RFFA was implemented using the open-access International Centre for Integrated Water Resources Management–Regional Analysis of Frequency Tool (ICI-RAFT), which enables the inclusion of climate variability effect into flood frequency estimation at locations where the assumption of hydrological stationarity is not viable. The Madden-Julian Oscillation was identified as the dominant flood influencing climate mechanism, with its effect increasing with return period. Similar to other studies, climate variability inclusive regional flood estimates were less than those derived from direct techniques at various locations, and higher in others. Also, the maximum historical flood experienced in the region was less than the 1-in-100-year flood event recommended for flood management. The 2012 flood in the Niger-South river basin of Nigeria was recreated in the CAESAR-LISFLOOD hydrodynamic model, combining open-access and third-party Digital Elevation Model (DEM), altimetry, bathymetry, aerial photo and hydrological data. The model was calibrated/validated in three sub-domains against in situ water level, overflight photos, Synthetic Aperture Radar (SAR) (TerraSAR-X, Radarsat2, CosmoSkyMed) and optical (MODIS) satellite images where available, to access model performance for a range of geomorphological and data variability. Improved data availability within constricted river channel areas resulted in better inundation extent and water level reconstruction, with the F-statistic reducing from 0.808 to 0.187 downstream into the vegetation dominating delta where data unavailability is pronounced. Overflight photos helped improve the model to reality capture ratio in the vegetation dominated delta and highlighted the deficiencies in SAR data for delineating flooding in the delta. Furthermore, the 2012 flood was within the confine of a 1-in-100-year flood for the sub-domain with maximum data availability, suggesting that in retrospect the 2012 flood event could have been managed effectively if flood management plans were implemented based on a 1-in-100-year flood. During flooding, fast-paced response is required. However, logistical challenges can hinder access to remote areas to collect the necessary data needed to inform real-time decisions. Thus, this adopts an integrated approach that combines crowd-sourcing and MODIS flood maps for near-real-time monitoring during the peak flood season of 2015. The results highlighted the merits and demerits of both approaches, and demonstrate the need for an integrated approach that leverages the strength of both methods to enhance flood capture at macro and micro scales. Crowd-sourcing also provided an option for demographic and risk perception data collection, which was evaluated against a government risk perception map and revealed the weaknesses in the government flood models caused by sparse/coarse data application and model uncertainty. The C4.5 decision tree algorithm was applied to integrate multiple open-access geospatial data to improve SAR image flood detection efficiency and the outputs were further applied in flood model validation. This approach resulted in F-Statistic improvement from 0.187 to 0.365 and reduced the CAESAR-LISFLOOD model overall bias from 3.432 to 0.699. Coarse data resolution, vegetation density, obsolete/non-existent river bathymetry, wetlands, ponds, uncontrolled dredging and illegal sand mining, were identified as the factors that contribute to flood model and map uncertainties in the delta region, hence the low accuracy depicted, despite the improvements that were achieved. Managing floods requires the coordination of efforts before, during and after flooding to ensure optimal mitigation in the event of an occurrence. In this study, and integrated flood modelling and mapping approach is undertaken, combining multiple open-access data using freely available tools to curb the effects of data and resources deficiency on hydrological, hydrodynamic and inundation mapping processes and outcomes in developing countries. This approach if adopted and implemented on a large-scale would improve flood preparedness, response and recovery in data sparse regions and ensure floods are managed sustainably with limited resources.