Investigating Uncertainty in Coastal Flood Risk Assessment in Small Island Developing States

Abstract

Small Island Developing States (SIDS) are increasingly under threat of coastal flooding, which challenges the safety of their societies and vulnerable economies. The emergency of this issue, exacerbated by climate change, has alarmed international organisations and national governments that have been demanding for robust risk assessments to guide the development of resilient adaptation strategies. In SIDS, the paucity of local data, required to perform such kind of coastal risk analyses, hinders the application of highly detailed models that therefore need to rely on inaccurate and publicly available data, thus introducing uncertainty in the assessment. This thesis aims to investigate the uncertainty in input data and its impact on coastal flood damage estimates. This study examines prominent uncertainty sources in the coastal flood risk modeling chain, namely: the stochastic variability of (i) significant wave height and (ii) storm surge water level, the quality of (iii) bathymetry data and (iv) digital elevation models and (v) the choice of depth-damage function. To account for risk temporal changes, two other inputs are included, specifically (vi) different sea level rise projections and (vii) socioeconomic developments. A methodology is developed to test the afore-mentioned inputs through global sensitivity analysis, using an ensemble of hydrodynamic models (XBeach and SFINCS) coupled with an impact model (Delft-FIAT). The impacts of these sources on the flood damage estimates are evaluated in a case study on the islands of São Tomé and Príncipe. Model results indicate, for the current time horizon, depth-damage functions and digital elevation models as the inputs with the most significant contribution to the overall damage estimation uncertainty, yielding a variation in the output prediction of a factor 16 and 10, respectively. As future climate and socioeconomic development uncertainties are introduced in the system, sea level rise projection becomes, followed by digital elevation models and depth-damage functions, the most relevant input for the year 2100. Neglecting economic growth in the risk analysis leads to an extremely high underestimation of damages. However, given the constrained intrinsic uncertainty for the projected societal trends, its sensitivity on the risk output is limited. The scarcity of accurate input data proves to have an enormous impact on risk assessments in Small Island Developing States, leading to considerable prediction error and affecting the model outcome uncertainty. New emerging data collection techniques, such as unmanned aerial vehicles, could augment the trustworthiness of risk assessments by providing more accurate datasets for bathymetry and topography. Furthermore, research efforts could be directed towards developing knowledge on the physics of damages and their implementation in a risk modeling scheme. The uncertainty framework presented could be applied in projects with the aim to support risk communication to stakeholders by portraying the implications of the various inputs used and assumptions made, but also to guide the allocation of limited economic resources towards the acquisition of the input data that matters the most in terms of reliability of damage estimates.