Deterministic and Probabilistic Flood risk assessment

Flood modeling, as an important part of coastal hazard assessment, is highly influenced by topography dataset and specifically ground elevation. Lower resolution Digital Elevation Models (DEMs) are usually used because of their availability and less computational burden. However, inherent errors in these DEMs propagate into flood risk analysis through spatial modeling. This study aims to explore the DEM resolution effects on coastal flood risk assessments. For this purpose, deterministic and probabilistic approaches are employed. Flood inundation modeling is carried out using hydrologically connected bathtub method. Given the high resolution Light Detection And Ranging (LiDAR) DEM, different resolution maps are obtained used resampling techniques and incorporated into an error analysis framework along with USGS national elevation dataset (NED) DEMs. The probabilistic framework is developed by simulating the spatial variability of elevation errors compared to LiDAR DEM through a Monte Carlo based method called sequential Gaussian simulation. The proposed methodology is applied to the lower Manhattan in New York City. By integrating the flood model into the developed framework, this approach results in flood inundation probability at each grid cells. In this study, using the concept of accuracy-efficiency tradeoffs, a framework for selecting a suitable spatial resolution for probabilistic flood risk assessment has been suggested. The results show that by exercising a range of options presented in this paper, a broader insight into mapping resolution can be made for making better flood assessment, evacuation zones, and mitigation plans depending upon the data availability in a region for flood preparedness.

洪水模拟作为海岸灾害评估的重要组成部分，其精度高度受地形数据集，尤其是地面高程数据的影响。由于低分辨率数字高程模型（Digital Elevation Models, DEMs）获取门槛较低且计算负荷更小，当前实践中常被采用。然而，此类DEM固有的误差会通过空间建模过程传导至洪水风险分析环节，进而对评估结果造成影响。本研究旨在探究DEM分辨率对海岸洪水风险评估的影响效应。为此，本研究采用了确定性与概率性两类分析方法。洪水淹没模拟采用水文连通型浴缸模型法（hydrologically connected bathtub method）开展。依托高分辨率激光雷达（Light Detection And Ranging, LiDAR）DEM，本研究通过重采样技术生成不同分辨率的高程图，并将其与美国地质调查局（United States Geological Survey, USGS）国家高程数据集（National Elevation Dataset, NED）DEM一同纳入误差分析框架。概率性分析框架通过基于蒙特卡洛（Monte Carlo）的序贯高斯模拟（sequential Gaussian simulation）方法，模拟高程误差相对于LiDAR DEM的空间变异特性而构建。所提出的研究方法应用于纽约市曼哈顿下城区区域。通过将洪水模型整合至所构建的分析框架中，本方法可输出每个网格单元的洪水淹没概率。本研究基于精度-效率权衡（accuracy-efficiency tradeoffs）理念，提出了一套面向概率性洪水风险评估的适宜空间分辨率选择框架。研究结果表明，通过本文所提出的一系列可选方案，可结合区域洪水备灾的可用数据情况，为优化洪水评估、划定疏散区域以及制定减灾规划提供更全面的分辨率选择思路，从而助力更科学的洪水备灾决策。