Precipitation threshold for urban flood warning - an analysis using the satellite-based flooded area and radar-gauge composite rainfall data

Abstract

A unique empirical approach of estimating the urban flood warning threshold is presented. First, the rainfall depth-duration relationship was estimated based on radar-gauge merged rainfall data with 1 km–10 min space-time resolution for 319 highly urbanized watersheds of the study area, for the extreme rainfall event occurred on the September 11, 2017 in Busan, Korea. The rainfall depth-duration relationship at the watersheds were further categorized by the nine different domains of the rainfall temporal variability and the watershed flooded area proportion that is estimated from the Sentinel-1 Synthetic Aperture Radar data analysis. The minimum possible slope of the depth-duration relationship in each of the domain was determined as the flood warning rainfall intensity threshold for this urban area. The results revealed that 55 mm/hr of rainfall intensity was the universal threshold. When this threshold was applied to issue the flood warning, the number of missed warning was minimized at the expense of the increased frequencies of false warnings. The highest identified threshold was 73 mm/hr. It was also found that the temporally variable rainfall is more likely to cause floods than the one with less temporal variability. The identified rainfall threshold values were validated against the 2010 and 2011 extreme events occurred in Seoul, Korea. The validation revealed that the balance between the accuracy and the reliability of the flood warning system highly depends on the choice of the tolerable flooded area proportions. The study reveals the great potential of operational urban flood warning system composed of a weather radar and a network of ground rainfall gauges that can be simply established using satellite-based flooded area data, which may be especially useful for developing countries.