Learning inter-annual flood loss risk models from historical flood insurance claims.

Flooding is a natural hazard that causes substantial loss of lives and livelihoods worldwide. Developing predictive models for flood-induced financial losses is crucial for applications such as insurance underwriting. This research uses the National Flood Insurance Program (NFIP) dataset between 2000 and 2020 to evaluate the predictive skill of past data in predicting near-future flood loss risk. Our approach applies neural networks (Conditional Generative Adversarial Networks), decision trees (Extreme Gradient Boosting), and kernel-based regressors (Gaussian Processes) to estimate pointwise losses. It aggregates them over intervals using a bias-corrected Burr-Pareto distribution to predict risk. The regression models help identify the most informative predictors and highlight crucial factors influencing flood-related financial losses. Applying our approach to quantify the county-level coastal flood loss risk in eight US Southern states results in an R2=0.807, substantially outperforming related work using stage-damage curves. More detailed testing on 11 counties with significant claims in the NFIP dataset reveals that Extreme Gradient Boosting yields the most favorable results, and bias correction significantly improves the similarity between the predicted and reference claim amount distributions. Our experiments also show that, despite the already experienced climate change, the difference in future short-term risk predictions of flood-loss amounts between historical shifting or expanding training data windows is insignificant.

Climate controls on the terrestrial water balance: Influence of aridity on the basin characteristics parameter in the Budyko framework.

The terrestrial water balance can be represented by the ratio of evapotranspiration to precipitation, which is expressed as a function of the aridity index (ϕ) and the basin characteristics parameter (n) in the Budyko framework. Traditionally n is assumed to be a constant for a catchment, independent to the climatic variables and altered only by changes in land cover and human activities. Another conceptual framework, Climate Change Impact Hypotheses (CCUW), makes similar assumption of constant catchment efficiency for evapotranspiration. In this study, using Variation Infiltration Capacity (VIC) model experiments, we show that the basin characteristics parameter and catchment efficiency are influenced by aridity index, in contrast with the traditional assumption. We also examine the analytical derivation of a functional form of Budyko equation and show that the assumption of n being independent of the climate variables is not valid. Hydrologic simulations with VIC show that the influence of seasonal change in vegetation (in the form of Leaf Area Index) on n is negligible compared to that of aridity, but the intra-seasonal rainfall variability does have impacts. We demonstrate these with a case study on impact of 1.5 °C and 2 °C global warming scenarios on the terrestrial water cycle in the Ganga river basin, one of the large river basins of South Asia with multiple sub-basins. Our findings imply that, with these assumptions, classical conceptual frameworks cannot fully explain the hydrometeorological impacts of climate change. These results highlight the importance of model evaluation and assessment of model assumptions before regional impact assessment studies.