Distinguishing between rain and snow is critical, yet challenging – crowdsourced data is an invaluable source in improving real-time weather forecasts.
DRI scientists Guo Yu, Meghan Collins, Monica Arienzo, and Anne Heggli co-authored a new study that examines how Mountain Rain or Snow is helping improve weather forecasting models. The citizen science project collects observations of precipitation from community volunteers across the country to further scientific understanding of the environmental variables impacting where precipitation falls as either rain or snow. This information is critical for informing avalanche forecasting, road safety, and water resources management.
Read the full story from the University of Vermont below:
By Hannah Fischer
A recent study published in the journal, Research Letters, leverages crowdsourced data to better represent patterns of precipitation phase—e.g., rain, snow, and mixed precipitation—across the US. In total, volunteer observers submitted nearly 40,000 precipitation phase reports using a simple smartphone app from mountain ranges such as the Green Mountains of Vermont, the Colorado Rocky Mountains, and the Sierra Nevada of California and Nevada.
The study’s authors used these observations to validate the rain and snow estimates from large-scale reanalysis products created by several US research agencies. These products use computer models and weather data to classify precipitation as rain, snow, or mixed, allowing for direct comparisons with the crowdsourced data. The study, Crowdsourced Data Reveal Shortcomings in Precipitation Phase Products for Rain and Snow Partitioning, provides insight into the strengths and weaknesses of these widely used precipitation phase products by comparing them to visual observations of precipitation across different ecoregions and elevation zones.
Correctly estimating precipitation phases is critical, as snow and impact transportation safety, air travel, infrastructure operations, and water resources management. Being able to accurately distinguish between solid and liquid phases improves real-time weather, flood, and streamflow forecasting.
However, distinguishing between rain and snow is challenging and few weather stations report the type of precipitation that is falling at any given time. Therefore, forecasters and scientists often have to use computer models like the ones evaluated in the study, many of which rely solely on near-surface meteorology data like air temperature and relative humidity. Methods using these data to predict whether it is raining or snowing struggle at temperatures near freezing, making it harder to forecast and respond to winter weather events.
Participatory science provides an alternative, effective way to monitor rain and snow patterns, and improve the capabilities of reanalysis products. The newly published research is led by the University of Vermont Water Resources Institute (WRI) Director of Research, Keith Jennings, and Guo Yu, an Assistant Research Professor at the Desert Research Institute (DRI), in partnership with other researchers at DRI, the Cooperative Institute for Research in the Atmosphere, Colorado State University, NOAA/Global Systems Laboratory, and the University of Nevada.
The team compared crowdsourced data from the Mountain Rain or Snow (MRoS) project with results from three precipitation phase products: the Global Precipitation Measurement (GPM) mission Integrated Multi‐Satellite Retrievals for GPM (IMERG), the Modern‐Era Retrospective Analysis for Research and Applications (MERRA‐2), and the North American Land Data Assimilation System (NLDAS‐2).
The MRoS project engages volunteer observers throughout the United States to submit real-time observations of precipitation phases. The data is collected via a mobile app with a simple interface that prompts observers to report the precipitation phase as rain, snow, or mixed. Each observation is automatically geotagged and time-stamped. Participants are notified via text message alerts of incoming storm events, increasing the number of consistent observations.
The study covered various elevations (0–4000 meters) between January 2020 and July 2023, with 39,680 MRoS observations. The visual observations were sent to an online database, where they underwent data preprocessing, meteorological variable estimation, and quality control (QC).
When validating the products against the crowdsourced MRoS data, the study team found accuracy discrepancies within the precipitation products. IMERG and MERRA-2 generally outperformed NLDAS-2. NLDAS-2 frequently misclassified rain as snow, exhibiting a bias towards rainfall. However, the reanalysis products all performed poorly when temperatures were close to freezing (around 0°C or 32°F).
When detecting rainfall at subfreezing temperatures or snowfall at warmer air temperatures, all three products struggled, as a result of the overlapping meteorological conditions associated with rain, snow, and mixed precipitation. These results highlight a gap in current modeling methods, particularly for temperature-sensitive events such as rain-on-snow and mixed precipitation.
This study demonstrates the value of crowdsourced data in mountain regions. Participatory scientists can fill gaps in satellite and model-based weather predictions, especially for regions that have complex weather patterns. Volunteer observers using a smartphone app are helping validate and improve reanalysis products.
If you want to volunteer your weather observing skills, sign up for Mountain Rain or Snow today.
