Modeling human displacement in the 2022 Pakistan floods: Current gaps and opportunities.

The 2022 Pakistan floods have been unprecedented in their extent. They affected around 33 million people, caused about 15 billion USD in damages, and took the lives of more than 1,800 persons, dominantly in the southern parts of the country. Effective disaster response requires fast assessments of likely impacts from hazardous weather to inform decision-makers and guide relief efforts for early action. Displacement modeling is a key technique towards these goals. However, displacement modeling which accounts for socio-economic components and uncertainties is methodologically challenging, and quantitative evidence largely remains limited and fragmented. Much work is needed to resolve these. This study aims at providing a case study for disaster displacement modeling by using the open-source impact assessment platform CLIMADA to investigate the extent by which flood-related hazards can be used to quantify displacement numbers in a data-limited region. Here, we estimate displacement from the 2022 Pakistan floods in Sindh province as a case study. We combine data on flood depth, exposed population, and provide impact functions that relate vulnerability of people likely to be displaced. We further use published numbers of affected people as target data for our model. The centerpiece of our analysis is the choice of impact functions. We test different forms of impact functions as well as assumptions about critical flood depths to proxy the number of displaced people, first using ex-ante assumptions, and then a numerically optimized version. With ex-ante assumptions, our model predicts a range of 1.94 to 5.65 million of displaced people in Sindh province, as compared to a total number of 6.76 million as reported by government sources. When we apply numerically optimized impact functions, the results closely resemble those obtained using the ex-ante assumptions, indicating that the current methods underestimate the extent of displacement. Additionally, we have evaluated the relationship between local vulnerability and the level of urbanization, and our findings reveal a negative correlation. We use this model to explain different displacement estimates for the 2022 floods across Pakistan and thereby contribute a case study to the growing field of displacement models, and towards the development of more refined ones. It highlights opportunities as well as limitations, and is a quantitative contribution to an existing discussion on how much disaster-related displacement can be modelled, and in how far assumptions can be generalized. These insights also support a better understanding of displacement and migration from future climate risks.