Quality Assessment of Space-Borne Active and Passive Microwave Snowfall
Products Over the Continental United States
Abstract
Surface snowfall rate estimates from the Global Precipitation
Measurement (GPM) mission’s core satellite sensors and CloudSat radar
are compared to those from the Multi-Radar Multi-Sensor (MRMS) radar
composite product generated over the continental United States (CONUS).
The considered algorithms include: Dual-Frequency Precipitation Radar
(DPR) product and its single frequency counterparts (Ka- and Ku-only);
the combined DPR and multifrequency microwave imager (CORRA) product;
the CloudSat SnowProfile product (2C-SNOW-PROFILE); two passive
microwave products i.e. the Goddard PROFiling algorithm (GPROF) and the
Snow retrievaL ALgorithm fOr gMi (SLALOM). The spaceborne and
ground-based snowfall products are collocated spatially and temporally
and compared at the spatial resolution of spaceborne instruments over
the period spanning from January 2016 to March 2020 (4 winters).
Detection capabilities of the sensors is assessed in terms of the most
commonly used forecast metrices (Probability of Detection, False Alarm
Ratio, etc.) whereas precision of the products is quantified by the mean
error (ME) and root-mean-square-error (RMSE). 2C-SNOW product agrees
with MRMS by far better than any other product. Passive microwave
algorithms tend to detect more precipitation events than the DPR and
CORRA retrievals, but they also trigger more false alarms. Due to
limited sensitivity, DPR detects only approx. 30% of the snow events.
All the retrievals underestimate snowfall rates, for the detected
snowstorms they produce approximately only a half of the precipitation
reported by MRMS. Large discrepancies (RMSE from 0.7 to 2.5 mm/h)
between spaceborne and ground-based snowfall rate estimates is the
result of limitations of both systems and complex ice scattering
properties. The MRMS product is based on a power law relation and it has
difficulties in detecting precipitation at far ranges; the DPR system is
affected by low sensitivity while the GPM Microwave Imager (GMI)
measurements are affected by the confounding effect of the background
surface emissivity for snow-covered surfaces and of the emission of
supercooled liquid droplet layers.