资源环境科技发展态势分析平台

Predicting the phase of precipitation is fundamental to water supply and hazard forecasting. Phase prediction methods (PPMs) are used to predict snow fraction, or the ratio of snowfall to total precipitation. Common temperature-based regression (Dai method) and threshold at freezing (0 degrees C) PPMs had comparable accuracy to a humidity-based PPM (T-RH method) using 6 and 24h observations. Using a daily climate data set from 1980 to 2015, the T-RH method estimates 14% and 6% greater precipitation-weighted snow fraction than the 0 degrees C and Dai methods, respectively. The T-RH method predicts four times less area with declining snow fraction than the Dai method (2.1% and 8.1% of the study domain, respectively) from 1980 to 2015, with the largest differences in the Cascade and Sierra Nevada mountains and southwestern U.S. Future Representative Concentration Pathway (RCP) 8.5 projections suggest warming temperatures of 4.2 degrees C and declining relative humidity of 1% over the 21st century. The T-RH method predicts a smaller reduction in snow fraction than temperature-only PPMs by 2100, consistent with lower humidity buffering declines in snow fraction caused by regional warming.

Plain Language Summary The phase of precipitation as rain or snow when it lands on the ground can cause different flood risks and alter water supplies for people and ecosystems. Most of our water supply forecasting models use simple temperature-only phase prediction methods, or PPMs, despite the fact that atmospheric conditions like humidity and pressure can also affect precipitation phase. In this study, we compared simple temperature-only PPMs, with another PPM that included temperature and relative humidity. Relative humidity well-below 100% acts to cool the falling snowflake through evaporation, similar to how sweating cools people through evaporative heat loss. We show that the different PPMs have relatively similar accuracy compared to observations. This gives confidence that we can hindcast back in time and forecast forward in time using the different PPMs and gridded climate datasets. We show that both hindcasts and forecasts are sensitive to the PPM chosen, with the PPM using relative humidity generally leading to more snowfall than the two temperature-only PPMs. These results suggest that future declines in relative humidity may be sufficient to buffer declines in snowfall as temperatures warm over the 21st century.