The phase of precipitation (solid or liquid) reaching the ground is largely governed by atmospheric temperature and humidity. Climate warming has shifted precipitation from snowfall to rainfall, reducing the snow fraction of total precipitation. This snow-to-rain shift has led to increased rainfall and decreased snow cover and snowpack depth across pan-Arctic permafrost regions. The reduction in snow depth decreases snowpack thermal insulation, allowing more heat to escape from the soil during the cold season, which results in a cooling effect on soil temperature. This cooling partly counteracts the warming of cold-season soil temperatures caused by rising atmospheric temperatures.

Additionally, reduced snow coverage and shortened snow seasons have dual effects. On the one hand, less snow reduces surface albedo, increasing solar radiation absorption, which warms the soil and enhances ecosystem respiration. On the other hand, it increases outgoing longwave radiation, cooling the soil. The overall impact of reduced snowfall on soil temperature and the carbon cycle remains uncertain, as these processes work in opposite directions, potentially leading to either positive or negative feedback.

Despite the significance of snow conditions in shaping terrestrial hydrology and ecosystem biogeochemical cycles, the specific effects of snow-to-rain shifts on permafrost carbon feedback remain unclear. To address this gap, we employed the Energy Exascale Earth System Model (E3SM) land model (ELM) to explore how snow-to-rain shifts impact CO2 emissions from pan-Arctic permafrost ecosystems. We first analyzed the sensitivity of ELM-simulated snow water equivalent (SWE), active layer thickness (ALT), warm-season CO2 uptake, and cold-season CO2 emissions to climate forcing and precipitation-phase partitioning methods (PPMs). Next, we compared ELM simulations of SWE and CO2 emissions with observational datasets. Finally, we projected trends in snow-to-rain shifts and cold-season carbon emissions under the RCP 8.5 scenario.

Our simulations showed agreements with observed SWE, ALT, and CO2 emissions over tundra ecosystems. We also found a strong positive correlation between changes in snowfall fraction and cold-season cumulative net CO2 flux, with Pearson coefficients of 0.89 for tundra and 0.79 for taiga areas. These results suggest that current snow-to-rain shifts may slow cold-season CO2 emissions from permafrost tundra and taiga areas, implying a potential negative feedback to climate warming. However, under future climate scenarios, the snowfall fraction during the cold season is projected to drop to around 10% of total precipitation by the end of the 21st century under RCP 8.5. At that point, the positive feedback from reduced snow cover may outweigh the negative feedback from decreased snow thermal insulation, substantially altering permafrost water-carbon-climate dynamics.