Bias Correction of Modeled Water Surface Elevations to Improve Exposure Predictions in the Eastern Carolinas

Floods cause large-scale economic stress and loss of human life; with climate change, more geographies are experiencing increases in flooding, putting more people at risk. Hydrodynamic models have been used to simulate floods at large scales to predict flooding and mitigate damages. However, model errors often range from +/- 0.11-0.33m, even for the best models. This may lead to lack of preparedness or reduce public trust in model outputs. In this study, we apply kriging to bias correct modeled water surface elevations from the Super Fast INundation of CoastS (SFINCS). We compare the change in predicted error by the SFINCS model and corresponding high water marks, finding a decrease in the mean average error in modeled WSE by 0.361m. We then leverage address-level NFIP claims and policies data at buildings in North Carolina to measure the improvement in exposure prediction. We find that the domain sensitivity, specificity, precision, and F1 score all improved in the corrected model. Applying kriging to hydrodynamic models can improve the accuracy of exposure predictions, resulting in better preparedness, more reliable exposure assessments, and risk mitigation for communities vulnerable to flooding.