Despite the critical role of snow conditions in terrestrial hydrology and ecosystem biogeochemical cycling, the specific impact of snow-to-rain shifts on permafrost carbon-climate feedback is still unclear. To address this knowledge gap, we utilize the Energy Exascale Earth System Model (E3SM) land model (ELM) to investigate the impact of snow-to-rain shifts on CO2 emissions from pan-Arctic permafrost ecosystems. Specifically, we first analyzed the sensitivity of ELM-simulated snow water equivalent (SWE), active layer thickness (ALT), warm-season net CO2 uptake, and cold-season net CO2 emissions to climate forcing and precipitation-phase partitioning methods (PPMs). We then evaluated ELM-simulated SWE and CO2 emissions against observationally-constrained datasets. Furthermore, we predicted trends of snow-to-rain shifting and cold-season carbon emissions over pan-Arctic ecosystems under the Representative Concentration Pathway (RCP) 8.5 scenario. The simulation results demonstrated good agreements with observation-derived SWE, ALT, and CO2 emissions over tundra ecosystems. The results also reveal positively correlated relationships between changes in ELM-simulated snowfall fraction and the corresponding changes in cold-season cumulative net CO2 flux with Pearson coefficient R as 0.89 and 0.79 for the tundra and taiga areas, respectively. This suggests that current snow-to-rain shifts will likely slow down cold-season CO2 emissions from permafrost tundra and taiga areas, indicating potentially negative feedback between snow-to-rain shifts and climate warming. However, under future climate change scenarios, the predicted cold-season snowfall fraction is expected to be only around 10% of the total precipitation by the end of the 21st century under the RCP 8.5 scenario. By then, snow cover decrease-induced positive feedback may exceed the negative feedback associated with decreasing snow thermal insulation, significantly modulating permafrost water-carbon-climate interactions.

