Spatial Variability of Falling Snow

One of the level-one requirements of the National Aeronautics and Space Administration’s Global Precipitation Measurement (GPM) mission is the detection of falling snow within the footprint of GPM Dual-frequency Precipitation Radar (DPR) and instantaneous field of view (IFOV) of GPM Microwave imager (GMI) on board GPM core observatory. The DPR footprint is nearly circular with a diameter of 5 km, while the IFOV of GMI is elliptic and has a range of maximum dimension of 32 km at 10.65 GHz and 7 km at 89 GHz. The non-uniform beam filling within the footprint and IFOV is one of the sources of uncertainty of DPR and GMI-based precipitation estimate. The GPM ground validation is committed to quantify these uncertainties (i.e. spatial variability) utilizing the ground based in-situ and remote sensing sensors. While the spatial variability in rain has been investigated using field-campaign based rain gauge and disdrometer network, there is no study on spatial variability of falling snow. This study is the first attempt to quantify the spatial variability of falling snow by using a network of ten Pluvio weighing bucket gauges in Marquette, Michigan. The gauges were separated by distances ranging from 2.6 km to 27.8 km and were within 17.5 km of the nearest operational radar (KMQT). One of the gauges was collocated with laser and camera-based optical disdrometers, a micro rain radar, and operational gauges. The Pluvio gauge had an excellent agreement with a tipping bucket gauge during rain and a very good agreement with a manual gauge during snow when event rainfall and snowfall totals were compared. The Pluvio gauges are designed to measure the snow water equivalent and reports every 10-second or a minute depending on the setting, but were not ideal for short term (less than 10 minutes) accumulations. During the study period, the gauges occasionally malfunctioned by failing to report the falling snow for 30-minute to an hour, but eventually outputs the accumulation accurately. For the purpose of this study, the gauge data is edited utilizing collocated disdrometers and nearby gauges. This labor-intensive effort was necessary. The gauge records were accumulated for 11-, 21-, and 31-minute periods centered at the observed minute. This requires the oversampling and is necessary to have a sufficient sample size. Correlations between the paired gauge records were calculated using Pearson’s correlation coefficient. The spatial variability was then determined using a three- parameter exponential function where the correlation at zero distance was assumed to be 0.99. The correlation at a given distance is the input for the exponential fit, while the correlation distance and shape parameter are the outputs. The longer the correlation distance, the more uniform the field (e.g. snow water equivalent). The study examined 16 events from the winter of 2017-18 where it was found that the lake-effect snow events appear to have lower correlations, which translates to higher variability, than the snow storms from synoptic systems.