Comparing Satellite Estimates of Precipitation Phase to Citizen Science Observations

Abstract:

Despite advances in remote sensing, modeling, and automated in situ monitoring, there is as of yet no substitute for visual observations of precipitation phase. Satellite retrievals, in particular, struggle with differentiating between rain and snow at the land surface, often relying on reanalysis products to make temperature-based predictions. In this work, we present results and analysis from the Tahoe Rain or Snow Citizen Science project, first using observations to better classify rain-snow partitioning and then comparing the output to estimates of precipitation phase from the Global Precipitation Measurement (GPM) mission. From January through May 2020, we collected over 1000 visual reports of rain, snow, and mixed precipitation, primarily from the Lake Tahoe Basin and the Truckee Meadows area near Reno, NV. Each rain-snow-mixed report included a timestamp and latitude-longitude position from the GPS data on the observer’s phone. After the snow season ended, we associated every observation with a predicted air temperature as distributed from several networks of ground-based meteorological stations. Although 0°C is commonly thought of as the air temperature at which snowfall melts into rainfall, previous research has shown that snow is typically the dominant form of precipitation at freezing and slightly warmer temperatures. Based on our predicted air temperatures, we found snow probability exceeded 50% at air temperatures <= 3.8°C. A preliminary analysis revealed GPM performed well at predicting the probability of snow by air temperature, explaining 81.5% of the variance in observations. However, GPM performed considerably worse at predicting the phase of precipitation on a case by case basis. At air temperatures between 2.5°C and 9.5°C, GPM correctly identified precipitation phase less than 50% of the time. Our analysis suggests the GPM output is seasonally non-biased in this region, while getting a majority of phase estimates wrong at air temperatures slightly above freezing. These findings indicate the utility of using citizen science data to validate satellite estimates of precipitation phase and future research should expand this effort to include a more climatically diverse selection of regions.