Page 311 - 《高原气象》2026年第3期
P. 311
3 期 谭澳博等:基于多源观测与陆面模式模拟的雨雪拆分研究 919
Rain-Snow Partitioning Based on Multi-Source Observations
and Land Surface Model Simulations
TAN Aobo, WEI Zhongwang, DAI Yongjiu
(School of Atmospheric Sciences, Sun Yat-sen University, Zhuhai 519082, Guangdong, China)
Abstract: Accurate discrimination of precipitation phase in numerical models is crucial for simulating snow ac‐
cumulation on land surfaces and its subsequent evolution. To evaluate the performance of mainstream precipita‐
tion-phase partitioning parameterization schemes used in land surface models for snow simulation, this study
conducted a systematic assessment. The evaluation was performed at both station and regional scales. At the sta‐
tion scale, observations from the Meteorological Information Comprehensive Analysis and Process System (MI‐
CAPS) were used to statistically evaluate rain-snow partitioning ratios. At the regional scale, the fifth-generation
global atmospheric reanalysis dataset ERA5 (0. 25°×0. 25°) was used to drive the Common Land Model version
2024 (CoLM2024) to simulate land surface processes over North America during 2005 -2019, and the model
outputs were compared with independent observations for regional evaluation. Five commonly used precipitation-
phase input schemes were considered in this study: air-temperature ramp, air-temperature threshold, wet-bulb-
temperature ramp, hydrometeor-temperature ramp, and direct input of reanalysis snowfall forcing. The main con‐
clusions are as follows: (1) Station-based statistics indicate clear differences among the schemes; the wet-bulb-
temperature ramp and hydrometeor-temperature ramp schemes are more consistent with observations in predict‐
ing snowfall probability, whereas the root mean square error (RMSE) of the air-temperature threshold scheme is
approximately twice that of the hydrometeor-temperature ramp scheme.(2) Land surface model simulations
show that all schemes generally underestimate surface snowpack; in regions where the grid-cell mean exceeds
1 mm, the annual mean snow amount is underestimated by more than 20%. Owing to its stable snowfall partition‐
ing behavior, the wet-bulb-temperature ramp scheme performs best at the regional scale, effectively alleviating
snow underestimation in the western United States while maintaining reasonable snow simulation performance in
the eastern United States.(3) Based on the analysis of snow simulation errors, model output errors were found
to be significantly positively correlated with both snow amount and elevation, with a correlation coefficient (R)
of 0. 86. This confirms the limitations of the model in simulating snowpack in cold high-elevation regions and
provides useful guidance for improving precipitation-phase parameterization schemes.
Key words: rain-snow partitioning; parameterization; land surface model; snow water equivalent

