Defining Rain vs. No Rain: An Evaluation of Thresholds for Delineating Rainfall Regions in Object-Based Methods for Precipitation Verification

Authors: Stephanie Zick*, Virginia Tech
Topics: Climatology and Meteorology, Spatial Analysis & Modeling, Hazards, Risks, and Disasters
Keywords: precipitation, rainfall, flooding, hydrology, spatial analysis
Session Type: Paper
Day: 4/11/2018
Start / End Time: 3:20 PM / 5:00 PM
Room: Grand Couteau, Sheraton, 5th Floor
Presentation File: No File Uploaded


Due to the nonlinear nature of atmospheric convection, determining an appropriate threshold for delineating a binary precipitation field is a nontrivial subject. In some applications, an ideal method is to delimit a region linked to the fundamental dynamical and physical mechanisms that produce and organize precipitation. For example, many previous studies have employed a strict 10 mm hr-1 (~0.4 inches hr-1) threshold as the approximate empirically-based separation between stratiform and convective precipitation. However, a lower or higher threshold may be more meaningful to hydrologists and/or emergency managers. Since object-based methods for precipitation verification are sensitive to the threshold value, it is important to evaluate the sensitivity of model forecast skill to the threshold that is specified. In this study, I consider short range (< 12 hour) hourly forecast precipitation from the High Resolution Rapid Refresh model for the Ellicott City, MD, flooding event in July 2016. Model forecast precipitation is compared against the National Centers for Environmental Prediction (NCEP) Stage IV precipitation analysis. For both datasets, I delineate a binary precipitation field using thresholds of 0.01, 0.1, 0.25, 0.5, 1, 2, and 5 inches. Next, I calculate spatial metrics of area, solidity, elongation, and fragmentation, as well as centroid position errors for matched objects, based on the forecast and observed fields. Comparisons of the performance of the HRRR forecast at each threshold are presented and contextualized for the Ellicott City case study to demonstrate the capabilities and limitations of high resolution rainfall forecasts for extreme rainfall events.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login