Scaling modeled potential residential loss from a storm surge

Natural and human-made hazards are recurrent phenomenon in human history. In recent years the number of natural hazards, especially coastal and hydro-meteorological hazards, is increasing worldwide (UN, 2004). Analogous to the rising global trend of hydro-meteorological hazards, the U.S. has experienced an increase in the number of severe hurricanes and tropical storms (Ayala-Carcedo, 2004). Since the 1980s, the coastal population of the US has also increased tremendously (Crosett et al., 2004). The risk from hurricane occurrence combined with increasing coastal population contributes to a large sum of financial loss due to property damage. This leads to the question - what is the potential loss from a category V hurricane in Jacksonville, FL or Norfolk, VA? At what geographic scale of analysis should a model to compute potential loss be used? Are model results from two different scales comparable? Although several studies and models have been developed to estimate potential losses from hurricane winds, few studies have explored the potential loss of property from a hurricane storm surge. Further, the loss estimation literature does not offer guidance on the "appropriate" geographic scales of analysis. The theoretical purpose of this research was to probe for a scale-modeled loss relationship, and if present to empirically define this relationship. A spatial loss estimation model was developed in this study to investigate the following research questions: (1) Is estimated loss monotonically (either always increasing or always decreasing) related to the scale of observation (e.g. census block, block group, etc.)? (2) Does the scale relationship vary geographically (e.g. east coast vs. west coast)? and (3) What is the geographic distribution of potential residential dollar loss due to a category 'n' (e.g. category V) hurricane induced storm surge? The model was implemented at five scales (parcel, block, block group, census tract and county) for 22 coastal counties in Florida. The model outcome revealed a negative scale-loss relationship such that at coarser resolution the estimated loss is always under-predicted. Not surprisingly, modeled loss at the block level was the best representation of modeled loss at the parcel scale. Though negative in nature, the slope of the relationship is not constant for all counties. The scale impact on modeled loss is large for counties with a high population density along the coastline, such that the modeled loss at the county scale is approximately 50% of the modeled loss at the parcel scale. The scale-loss relationship is either increasing or indeterminate for (1) rural counties, (2) counties with low population density along the coastline and (3) counties with high population densities inland. The scale--loss relationship does not vary systematically geographically. Florida counties along the west coast do not experience greater scale impacts than counties along the east coast.