Extreme Precipitation and Flooding: Exposure Characterization and the Association Between Exposure and Mortality in 108 United States Communities, 1987-2005

There is substantial evidence that extreme precipitation and flooding are serious threats to public health and safety. These threats are predicted to increase with climate change. Epidemiological studies investigating the health effects of these events vary in the methods used to characterize exposure. Here, we compare two sources of precipitation data (National Oceanic and Atmospheric Administration (NOAA) station-based and North American Land Data Assimilation Systems (NLDAS-2) Reanalysis data-based) for estimating exposure to extreme precipitation and two sources of flooding data, based on United States Geological Survey (USGS) streamflow gages and the NOAA Storm Events database. We investigate associations between each of the four exposure metrics and short-term risk of four causes of mortality (accidental, respiratory-related, cardiovascular-related, and all-cause) in the United States from 1987 through 2005. Average daily precipitation values from the two precipitation data sources were moderately correlated (Spearman's rho = 0.74); however, values from the two data sources were less correlated when comparing binary metrics of exposure to extreme precipitation days (Jaccard index (J) = 0.35). Binary metrics of daily flood exposure were poorly correlated between the two flood data sources (Spearman's rho = 0.07; J = 0.05). There was little correlation between extreme precipitation exposure and flood exposure in study communities. We did not observe evidence of a positive association between any of the four exposure metrics and risk of any of the four mortality outcomes considered. Our results suggest, due to the observed lack of agreement between different extreme precipitation and flood metrics, that exposure to extreme precipitation may not serve as an effective surrogate for exposures related to flooding. Furthermore, It is possible that extreme precipitation and flood exposures may often be too localized to allow accurate exposure assessment at the community level for epidemiological studies.