An Automated Method For Detecting Precipitation and Cell Type From Radar Products

Abstract

A variety of methods exist for discriminating convective and stratiform precipitation using radar imagery. Additionally, many methods have been proposed to classify thunderstorm cells based on their appearance on radar. Frequently these algorithms employ reflectivity thresholds and shape criteria that have been demonstrated to work in many situations. Through these analyses a variety of information can be determined or estimated including storm type, morphology, severity, and an expected drop size distribution. In addition to examining reflectivity thresholds, the spatial scale of features observed in radar imagery can be useful to distinguish convective and stratiform precipitation. Through the use of a Discrete Fourier Transform (DFT) and Gaussian band pass filtering, signals of different spatial scales can be isolated. The DFT and the filtering process can also be mimicked through use of the Discrete Wavelet Transform to isolate signals in multiple bands. Typical methods of discerning convective and stratiform precipitation that rely on reflectivity thresholds may fail under conditions where convective precipitation has unusually low reflectivity or stratiform precipitation has unusually high reflectivity. Examining both the power and scale of the signal provides more information and yields greater accuracy in discriminating between stratiform and convective precipitation. Additionally by discriminating based on scale this method identifies areas where convective signal is superimposed atop a stratiform signal indicating convection embedded within an area of stratiform precipitation. Once convective regions have been identified, shape analysis and model-derived quantities provide further information on cell type and characteristics. Additionally, this method identifies structures that are contained within larger structures such as individual cells contained within a larger mesoscale convective system.