Search

Flood susceptibility in the Lower Connecticut River Valley Region attributable to nonclimatic flood risk factors is mapped using a quantitative method using logistic regression. Flood risk factors considered include elevation, slope, curvature (concave, convex, or flat), distance to water, land cover, vegetative density, surficial materials, soil drainage, and impervious surface. Values of factors at point locations were correlated to whether a location was located within or outside of the U.S. Federal Emergency Management Agency 100‐year Special Flood Hazard Area (SFHA). The Lower Connecticut River Valley Region was divided into urban, rural, and coastal subregions to assess the differences in factor contributions to flood susceptibility between different region types; for each region flood risk factors were extracted from 4,000 points, of which an equal number were within or outside of the 100‐year SFHA. Logistic regression coefficients were obtained. It was found that elevationand distance to waterhave the greatest contribution to flood susceptibility in the urban and coastal subregions, whereas distance to water and surficial materialsdominate in the rural subregion. The contribution of land useto flood susceptibility increased by over 200% between the rural and urban regions. Probabilities of flooding were computed using each regional logistic regression equation. Several areas classified as very high risk(80–100%) and high risk(60–80%) were located outside of the SFHA and included several types of infrastructure critical for human health, safety, and education. This study demonstrates the utility of logistic regression as an efficient methodology to map regional flood susceptibility. Flooding is one of the most severe and potentially devastating natural disasters that can occur. Floods can come in many forms, including river, coastal, and flash flooding. Whenever and wherever any of these types of flooding occur, long‐term planning and adaptation, preparedness, and response time are all critical factors in reducing the overall impacts. Awareness of areas that are currently prone and will remain prone to flooding in the future is essential to consider in both short‐term and long‐term planning. Such awareness comes from an understanding of a combination not only of regional climatic factors but also of nonclimate factors that relate to natural, physical, and development characteristics. The current study estimates the risk of flooding throughout the Lower Connecticut River Valley Region (LCRVR) based on site and regional characteristics not related to climate. Several methods were considered to estimate flood risk; the method that was finally selected for this study involves a statistical approach in which a data set having one or more independent variables that produce a binary value of no or yes (0 or 1, respectively) for the dependent variable is analyzed. The independent variables in this case include several nonclimate factors related to flood risk that could potentially affect the region and for which sufficient data were available and are referred to as flood risk factors. Flood risk factors considered include elevation, land slope, land curvature (concave, convex, or flat), distance to water body, land cover, density of vegetation, surface geology, ability of the soil to drain water, and the percent of impervious surface (e.g., pavement). The objective is to link each of the flood risk factors to the dependent variables, which in this case is the occurrence of flooding for a flood event that is estimated to occur on average once in every 100 years. It was found that the overall quality of recent satellite images of the LCRVR during large flood events was not sufficient for the current analysis; therefore, it was decided to use the U.S. Federal Emergency Management Agency 100‐year Special Flood Hazard Area (SFHA) to indicate areas where flood inundation would occur. The advantage of using the SFHA and the selected statistical modeling methodology is that they allow the contribution of each flood risk factor within the SFHA to be estimated and then applied to the entire study region to identify additional areas outside of the SFHA that have high flood risk. The LCRVR was divided into three subregions (urban, rural, and coastal) to accentuate the differences in the contributions of each flood risk factor to flood risk between an urban and a rural area and between inland and coastal areas; for each subregion 4,000 point locations were randomly chosen from which to extract data for each flood risk factor. An equal number of these points were selected in locations that were within and outside of the SFHA for each subregion. Site data for each flood risk factor were extracted and associated with a 1 if the location was within the SFHA and a 0 otherwise. The resulting relations between each flood risk factor and flood occurrence were analyzed so that regression coefficients could be estimated for each factor, the magnitude of which indicates the relative strength of each flood risk factor's influence on flooding in a subregion. It was found that elevationand distance to waterhave the most influence on flood risk in the urban and coastal subregions, whereas distance to water and surface geologydominate in the rural subregion. The contribution of elevation and land usewere also found to increase the most between the rural and urban subregions. The coefficients for each subregion are then used to assign probabilities of flooding to all locations over a grid covering that subregion. The results for each subregion were combined to create an overall flood probability map of the LCRVR. Probabilities were classified very low risk(0–20%), low risk(20–40%), medium risk(40–60%), high risk(60–80%), and very high risk(80–100%). It was observed that several areas classified as very high riskand high riskwere located outside of the SFHA. Several types of infrastructure critical for human health, safety, and education were finally overlaid on the flood risk map to identify those assets that are most vulnerable to the 100‐year flood and may therefore require additional flood risk mitigation. Elevation, distance to water, and surficial materials had the highest contributions to flood susceptibility throughout the study areaThe contribution of elevation and land use to flood susceptibility increased substantially when comparing the urban to the rural subregionVery high and high susceptible areas add over 6% of nonwater and wetland area to the SFHA, including 8% more developed area

Precipitation quantile estimates are used in engineering, agriculture, and a variety of other disciplines. Index flood regional frequency methods pool normalized gauge data in the case of homogeneity among the constituent gauges of the region. Unitless regional quantile estimates are outputted and rescaled at each gauge. Because violation of the homogeneity hypothesis is a major component of quantile estimation error in regional frequency analysis, heterogeneity estimators should be “reasonable proxies” of the error of quantile estimation. In this study, three Monte Carlo heterogeneity statistics tested in Hosking and Wallis (1997) are plotted against Monte Carlo estimates of quantile error for all five‐or‐more‐gauge regionalizations in a 12 gauge network in the Twin Cities region of Minnesota. Upper‐tail quantiles with nonexceedance probabilities of 0.75 and above are examined at time steps ranging from daily to monthly. A linear relationship between heterogeneity and error estimates is found and quantified using Pearson's r score. Two of Hosking and Wallis's (1997) heterogeneity measures, incorporating the coefficient of variation in one case and additionally the skewness in the other, are found to be reasonable proxies for quantile error at the L‐moment ratio values characterizing these data. This result, in addition to confirming the utility of a commonly used coefficient of variation‐based heterogeneity statistic, provides evidence for the utility of a heterogeneity measure that incorporates skewness information. Two commonly used heterogeneity statistics based on linear moment ratios are validatedLinearity with respect to quantile error of estimation is shown for both statisticsA method for evaluating all regionalizations of a gauge data set is presented