Rainfall Frequency Analysis Based on Long‐Term High‐Resolution Radar Rainfall Fields: Spatial Heterogeneities and Temporal Nonstationarities

Abstract

AbstractRainfall frequency analysis methods are developed and implemented based on high‐resolution radar rainfall data sets, with the Baltimore metropolitan area serving as the principal study region. Analyses focus on spatial heterogeneities and time trends in sub‐daily rainfall extremes. The 22‐year radar rainfall data set for the Baltimore study region combines reflectivity‐based rainfall fields during the period from 2000 to 2011 and polarimetric rainfall fields for the period from 2012 to 2021. Rainfall frequency analyses are based on non‐stationary formulations of peaks‐over‐threshold and annual peak methods. Increasing trends in short‐duration rainfall extremes are inferred from both peaks‐over‐threshold and annual peak analyses for the period from 2000 to 2021. There are pronounced spatial gradients in short‐duration rainfall extremes over the study region, with peak values of rainfall between Baltimore City and Chesapeake Bay. Spatial gradients in 100‐year, 1 hr rainfall over 20 km length scale are comparable to time trends over 20 years. Rainfall analyses address the broad challenge of assessing changing properties of short‐duration rainfall in urban regions. Analyses of high‐resolution rainfall fields show that sub‐daily rainfall extremes are only weakly related to daily extremes, pointing to difficulties in inferring climatological properties of sub‐daily rainfall from daily rainfall analyses. Changing measurement properties are a key challenge for application of radar rainfall data sets to detection of time trends. Mean field bias correction of radar rainfall fields using rain gauge observations is an important tool for improving radar rainfall fields and provides a useful tool for addressing problems associated with changing radar measurement properties.