Evaluation of Precipitation Phase Algorithms in Southern New England

Precipitation is one of the two main variables that govern the climate and global water cycle. Its global distribution, anomaly (departure from norm), and extremes play a fundamental role in nature. Due to the ocean’s global surface coverage (> 70%) and lack of ground-based observations in a vast majority of the land, satellite-based precipitation products alongside global and regional weather forecasting models are the main source for mapping global precipitation. Yet, precipitation phase needs to be identified prior to determining the precipitation amount. NASA’s multi-satellite precipitation product, IMERG, has been operated and funded under the Global Precipitation Measurement (GPM) mission and is the single most used precipitation product globally. IMERG employs a probabilistic approach to precipitation phase relying on the European weather forecast center’s Reanalysis (ERA5) wet-bulb temperature estimates. NOAA’s Multi-Radar Multi-Sensor (MRMS), a gauge-adjusted weather radar product over the United States, has been adopted as the ground validation product by the GPM program. MRMS employs a deterministic approach to precipitation phase relying on High Resolution Rapid Refresh (HRRR) air and wet-bulb temperatures. HRRR itself has both deterministic and probabilistic precipitation phase algorithms.

This study aims to evaluate IMERG and MRMS algorithms using a three-winter-long field campaign data sets in Connecticut. The field campaigns included an All-In-One (AIO) weather station, laser-optical PARSIVEL disdrometer, camera-based Precipitation Imaging Package (PIP) video disdrometer, and vertically pointing Micro-Rain-Radar (MRR) among other instruments. MRMS and IMERG algorithms utilizing the AIO temperatures are used as references. MRMS and IMERG algorithms utilizing HRRR, ERA5, and NASA’s global modeling and assimilation office (MERRA-2) reanalysis have been evaluated. The MRMS and IMERG phase algorithms are evaluated against several observationally based phase algorithms, including that of PARSIVEL fall velocity, PIP bulk and equivalent density, PIP and MRR Doppler velocity. A good agreement between the PIP derived deterministic Doppler velocity, PARSIVEL phase algorithms, PIP bulk density creates a reliable reference for the transitional events. The AIO-based IMERG algorithm performed better than any other probabilistic and deterministic algorithms. The relatively coarse spatial scales of HRRR, ERA5, and MERRA-2 outputs lower wet-bulb temperature resulting in higher snow probability. Several transition events showed significant differences in wet-bulb temperatures between the two sites that are 3.3 km apart during the second winter. This reflects the significance of the spatial resolution of models when considering wet-bulb temperature. HRRR deterministic and probabilistic phase algorithms, on the other hand, performed better than air and wet-bulb temperature-based algorithms including the AIO-MRMS algorithm.