An Area‐Based Approach for Estimating Extreme Precipitation Probability

Abstract

Accurate estimates of heavy rainfall probabilities reduce loss of life, property, and infrastructure failure resulting from flooding. NOAA's Atlas-14 provides point-based precipitation exceedance probability estimates for a range of durations and recurrence intervals. While it has been used as an engineering reference, Atlas-14 does not provide direct estimates of areal rainfall totals which provide a better predictor of flooding that leads to infrastructure failure, and more relevant input for storm water or hydrologic modeling. This study produces heavy precipitation exceedance probability estimates based on basin-level precipitation totals. We adapted a Generalized Extreme Value distribution to estimate Intensity-Duration-Frequency curves from annual maximum totals. The method exploits a high-resolution precipitation data set and uses a bootstrapping approach to borrow spatially across homogeneous regions, substituting space in lieu of long-time series. We compared area-based estimates of 1-, 2-, and 4-day annual maximum total probabilities against point-based estimates at rain gauges within watersheds impacted by five recent extraordinary precipitation and flooding events. We found considerable differences between point-based and area-based estimates. It suggests that caveats are needed when using pointed-based estimates to represent areal estimates as model inputs for the purpose of storm water management and flood risk assessment.

Regionalization of Hydroclimate Variables

“…We adopted a hierarchical clustering method, Regionalization with Dynamically Constrained Agglomerative Clustering and Partitioning (REDCAP; Guo 2008), to identify clusters based on underlying parameters for the probability distributions used to calculate SPI (gamma), SPEI (Pearson-Type III), and 1-to 4-day precipitation (gamma). REDCAP is a widely used clustering algorithm with previous hydroclimate applications (Gao et al 2018;Yang et al 2020). REDCAP offers a flexible algorithm that overcomes limitations inherent in conventional clustering methods that do not consider spatial information, one of the key factors that shape climate regions.…”

“…REDCAP offers a flexible algorithm that overcomes limitations inherent in conventional clustering methods that do not consider spatial information, one of the key factors that shape climate regions. It uses the average linkage clustering method but directly enforces spatial contiguity during the clustering procedure forming spatially contiguous regions within which various analyses --such as climate model evaluation and resampling of rare or extreme events --could be conducted (Gao et al 2015;Gao et al 2018). By accommodating different variables derived from raw temperature and precipitation and various distance measurements (Table 1), REDCAP allows us to investigate multiple facets of hydroclimate and distill information from raw data.…”

“…Regionalization also can help improve existing methods for drought forecasts or for deriving future scenarios (Mishra and Singh 2011). Creating rational climate regions is equally relevant in analyzing heavy precipitation and flooding, as well as in storm-water and disaster and risk management (DeGaetano 2001;Du et al 2014;Gao et al 2018;Irwin et al 2017). It is the basis for understanding heavy precipitation regions and is essential to stochastic storm transposition, wherein historic storms are superimposed on different regions as plausible heavy rain events.…”

“…It is the basis for understanding heavy precipitation regions and is essential to stochastic storm transposition, wherein historic storms are superimposed on different regions as plausible heavy rain events. Resampling the intensity, duration, and spatial extent of a past event provides a complementary method for rainfall and flood frequency analysis (Yin et al 2016;Wright et al 2020) but demands coherency in the statistical properties used to construct intensity-duration-frequency (IDF) curves and other measures of probability and risk due to heavy precipitation (Gao et al 2018). Hence, investigators have looked at spatial coherence in precipitation intensity, including diurnal rainfall patterns (Mooney et al 2017), annual precipitation extremes (Yang et al 2010), and variability in mountainous regions (Sugg and Konrad 2020).…”

A comparison of extreme precipitation event frequency and magnitude using a high‐resolution rain gage network and NOAA Atlas 14 across Delaware

“…This tool contains both precipitation frequency estimates, as well as associated confidence intervals. Atlas 14 uses surface stations, primarily from the National Weather Service's Cooperative Observer Program (COOP), and statistical methodologies to provide point-based precipitation exceedance probability estimates for a number of durations and potential recurrence intervals (Gao et al, 2018). Unfortunately, COOP stations present an array of biases that are difficult to identify (Daly et al, 2007), are not homogeneously distributed across the United States, and contain varying periods of record.…”