GPM Ground Validation: Comparative Study of Snow Detection Algorithms

One of the four level-one sceince requirements of the National Aeronautics and Space Administration (NASA) Global Precipitation Measurement (GPM) mission states that the dual-frequency precipitation radar (DPR) and GPM Microwave Imager on board the GPM Core observatory will detect the falling snow at effective resolution of 5 km and 15 km, respectively. For validation, the dual-pol radar based Hydrometeor identification (HID) is a considered as a prime product. The GPM ground validation team routinely process the National Weather Service (NWS) operational radars during GPM overpasses at selected sites. A number of research radars including NASA’s S-band dual-pol radar (NPOL) are also processed and HID is a standard output. The HID algorithm uses dual-pol variables and classifies each radar pixel to eleven different classes including bad data and an unclassified category (Dolan et al. 2013). The National Severe Storm Laboratory Multi-Radar Multi-Sensor (MRMS) product outputs percent of snow at grid resolution of 0.01x0.01 degree on hourly basis across the continental US (Wen et al. 2017). It should be noted that MRMS used for GPM is further processed as a research product to provide both instantaneous precipitation rates and types (liquid/frozen) for comparisons to individual GPM core overpasses and also the snow percentage every half-an-hour after June 2014. The NWS Automated Surface Observation Systems (ASOS) have a present weather sensor and report rain and snow at light, moderate, and heavy intensities at one-minute resolution. The ASOS network has more than 900 stations across the US and its temperature sensor provides additional information on the characteristics of snow.

The NASA Wallops Flight Facility (WFF) is located at the mid-Atlantic coast of the US (37.94˚N, 75.64˚W) and hosts both ASOS and sounding stations. NPOL, located in Newark, Maryland, and the nearest operational radar (KDOX) are 38 km and 99 km away from the WFF main base, respectively and the lowest elevations of NPOL and KDOX are at 530 m and 830 m above the ground with no blockage. WFF also hosts a number of research instruments including the Particle Imaging Package (PIP), Autonomous Parsivel² disdrometer Unit (APU), and vertically pointing K-band Micro Rain Radar (MRR). Both PIP and APU measures the size and fall velocity of individual hydrometeors but the former was developed for rain and the latter was developed for snow. APU is also a present weather sensor and outputs two different weather codes at 10-second resolution. Given the two weather codes output the same precipitation type, which does not change within a minute, one of the APU based weather codes is considered at one-minute resolution. PIP, on the other hand, outputs rain and non-rain rate at one-minute resolution. The ratio of rain versus non-rain rate is taken as percent rain.

MRR was set to its lowest height mode where the vertical profile of reflectivity and Doppler velocity are given at 35 m vertical and one-minute temporal resolution extending from near the surface (105 m) to 1 km above the ground. The vertical air velocity plays a relatively insignificant role in winter precipitation and Doppler velocity is directly related to the fall speed of hydrometeors. MRR is therefore a great asset for visually inspecting and evaluating the precipitation structure from beginning to end of the event within the boundary layer.

During the 2013-14 winter, ten distinct precipitation events were observed at WFF. Two of the events were labeled “cold rain” events where ASOS temperature was within 5-10˚C range and two other events were labeled “cold snow” events where temperature remained less than -5˚C. The remaining events occurred at temperatures above -5˚C where a number of systems started as rain followed by mixed and frozen precipitation and vice versa. A number of systems had frozen precipitation aloft but mixed or liquid precipitation was observed at the ground. Considering the ASOS precipitation code as a reference, a comparative study of precipitation type was conducted among US-wide available HID and MRMS products and WFF specific APU and PIP outputs. It is clear that each sensor or product outputs the precipitation type differently. An important aspect is to document the performance each output at various conditions. The classification of snow in a cold rain event or rain in a cold snow event is of course important feedback to report to the algorithm developer. It is understandable that the radar-based products may fail to report the correct precipitation type at the surface in warm snow events. Indeed, currently the DPR and combined radar-radiometer algorithms determine the precipitation type and rate near, but not necessarily at the surface based on lowest reliable clutter-free gate, which can be 2 km above the ground at farthest edge of the scan.

The study is expected to provide a reference on the accuracy of radar-based precipitation type products. While WFF is a challenging site since it sits on rain-snow line for numerous systems, the mountainous areas can bring additional challenge in determining the precipitation type.

Dolan, B., S. Rutledge, S. Lim, V. Chandrasekar, and M. Thurai, 2013: A robust C-band hydrometeor identification algorithm and application to a long-term polarimetric radar dataset. J. Applied Meteorology Climatology, 52, 2162-2186.

Wen, Y., P. Kirstetter, J. J. Gourley, Y. Hong, A. Behrangi, Z. Flaming, 2017: Evaluation of MRMS snowfall products over the western United States. J. Hydrometeorology (in print).