Title:Scenario and Sensitivity Analysis for Flooding Vulnerability using Genetic Algorithms

Abstract:We try to answer the question: "can we 'modify' our neighborhoods to make them less vulnerable to flooding?" We minimize flooding vulnerability for a city in the central plain of Luzon, by modeling the city as a biological organism with 'traits', and try to 'breed' a 'champion' city (with a low flooding vulnerability) via a genetic algorithm. The result is a description of the traits the barangays (neighborhoods) should have (the 'design' of the city). As far as we can tell, this kind of modeling has not been attempted before. The different components of flooding vulnerability were investigated, and each was given a weight, which allows us to express vulnerability as a weighted sum; this serves as the fitness function for the genetic algorithm. We also allowed non-linear interactions among related but independent components, viz, poverty and mortality rate, and literacy and radio/TV penetration. The two-table system we used to prioritize the components of vulnerability is prone to subjectivity, a common problem in analyses of vulnerability. Thus, a sensitivity analysis was done, which gave a design with a 24% decrease in vulnerability alongside a 14% percent decrease in cost, a significant improvement over this initial scenario analysis, where the proposed design had a 12% decrease in vulnerability with a one percent increase in cost.