New climate models reveal faster and larger increases in Arctic precipitation than previously projected

Arctic precipitation is widely expected to increase during the twenty-first century, driven by greater evaporation from expanding open water as sea ice declines, a warmer atmosphere with higher moisture-holding capacity, and enhanced poleward moisture transport. A shift from snow- to rain-dominated precipitation has been observed in parts of the Arctic and projected to expand, but uncertainties remain regarding regional extent and seasonal timing. With CMIP6 models showing improved simulations of sea-ice state and trends, historical snow cover, and global precipitation intensities relative to CMIP5, the study asks whether CMIP6 projects larger and faster Arctic precipitation increases and an earlier transition to rainfall dominance than CMIP5, and identifies the drivers behind any differences.

Previous studies estimate a 30–60% increase in Arctic precipitation by 2100 and attribute changes to increased evaporation due to sea-ice loss, higher air temperatures boosting atmospheric moisture capacity, and enhanced poleward moisture transport. A transition toward rain-dominated precipitation has been documented and linked to impacts on ice sheets, sea level, river discharge, sea ice, permafrost, and ecosystems. CMIP6 offers improved representations of Arctic sea ice, historical snow cover, and precipitation extremes compared to CMIP5, suggesting potential improvements in hydrological cycle projections. Observational products have limitations: GPCP may underestimate high-latitude precipitation and differ from reanalyses, while ERA5 has been shown to perform best among current reanalyses for Arctic precipitation.

The study analyzes total precipitation, snowfall, and rainfall (rainfall derived as total precipitation minus snowfall), open water area (inverse of sea-ice concentration), surface air temperature, and vertically integrated moisture flux (VIMF) from CMIP5 and CMIP6 models. Historical periods: 1960–2005 (CMIP5) and 1960–2014 (CMIP6). Future scenarios: RCP8.5 (2006–2100, CMIP5) and SSP5-8.5 (2015–2100, CMIP6), designed to have comparable end-of-century radiative forcing (8.5 W m⁻²). One ensemble member per model was used, including only models with both historical and scenario runs. The Arctic region is defined as 70–90°N. End-of-century changes are computed as 2091–2100 averages relative to 2005–2014 (for CMIP5 the latter uses the first 10 years of RCP8.5). The decade of transition to rainfall dominance is identified when the snowfall-to-precipitation ratio falls below 0.5 in 10-year binned windows. Historical evaluation uses ERA5 reanalysis and GPCP data for 1979–2005; spatial climatologies and time series are compared, including snowfall ratio trends. Sensitivity analyses quantify percentage change in precipitation per degree of Arctic and global warming by month. Inter-model relationships are assessed between changes in rainfall/snowfall and changes in temperature, open water, and VIMF, with correlations and lines of best fit. VIMF is computed by vertically integrating the product of meridional wind and specific humidity from surface (assumed 1000 hPa) to 400 hPa using the trapezoidal rule. Statistical significance of differences between end- and start-of-century is evaluated with Student’s t-test. The relationship between global warming levels (1.5 °C, 2 °C, 3 °C above 1850–1900) and snowfall ratio is obtained by regressing model-specific global temperature anomalies (1960–2100) against snowfall ratio and using slope/intercept to estimate ratios at specified warming thresholds.

Historical evaluation: CMIP5 and CMIP6 simulate annual-mean Arctic (70–90°N) precipitation consistent with ERA5 (~0.94 ± 0.03 mm day⁻¹) and realistic spatial patterns. GPCP shows lower magnitudes and weak interannual correlation with ERA5 (r = 0.33), suggesting a low bias. ERA5 snowfall ratio shows a small but significant decline (≈−0.06) over 1979–2005, captured by both CMIP ensembles; ERA5 generally lies within CMIP6 spread more than CMIP5, indicating improved precipitation partitioning in CMIP6. End-of-century changes (SSP5-8.5/RCP8.5): Total precipitation increases in all seasons, dominated by rainfall increases. CMIP6 projects larger increases than CMIP5, especially in autumn and winter. By 2100 vs 2000, rainfall increase percentages (CMIP6 vs CMIP5): winter 422% vs 260%; spring 261% vs 141%; summer 71% vs 51%; autumn 268% vs 192%. This yields additional rainfall in CMIP6 relative to CMIP5 by ~0.3 mm day⁻¹ (27.3 mm per season) in autumn and ~0.2 mm day⁻¹ (18.2 mm per season) in spring and winter. From 2020–2100, autumn rainfall increases by 0.9 mm day⁻¹ (81.9 mm per season) in CMIP6 vs 0.7 mm day⁻¹ (63.7 mm per season) in CMIP5 (+24%). Winter, spring, and summer rainfall increases are 39%, 36%, and 14% greater, respectively, in CMIP6. Snowfall decreases more strongly in CMIP6 in summer (−16%) and autumn (−38%), while winter snowfall still increases across much of the Arctic, more so in CMIP6. Uncertainty and extremes: Inter-model spread in precipitation changes is larger in CMIP6, with higher upper-tail projections (e.g., autumn 95th percentile rainfall 1.41 vs 1.26 mm day⁻¹ in CMIP6 vs CMIP5; autumn upper limit ≈1.7 vs 1.4 mm day⁻¹). Spatial patterns: Greater autumn rainfall increases and snowfall decreases in CMIP6, with significant rainfall increases up to 0.6 mm day⁻¹ around the Greenland and Barents Seas and larger winter increases across the Arctic Ocean and peripheral seas. Winter snowfall increases are larger in CMIP6, notably in Siberia and the Canadian Arctic Archipelago. Drivers: CMIP6 shows greater precipitation sensitivity per degree of warming, especially in autumn and early winter, and larger Arctic amplification. Rainfall increase (and snowfall decrease) magnitudes correlate significantly with Arctic warming across models in all seasons (r ≈ 0.6–0.89), with larger CMIP6 winter warming (multi-model mean ~15 °C by 2100 vs 13 °C in CMIP5; maxima 23 °C vs 18 °C in individual models). Open water increases are larger in CMIP6 (winter open water ≈9 vs 5.5 million km² by 2100), and changes in open water correlate with rainfall increases (r ≈ 0.53–0.92) and with snowfall decreases in summer and autumn (≈−0.99 and −0.76). VIMF increases in all seasons are larger in CMIP6 and positively correlate with rainfall increases and with snowfall decreases in summer and autumn. For given global warming levels (1.5 °C, 2 °C, 3 °C), CMIP6 yields more rainfall than CMIP5 in all seasons, indicating faster hydrological intensification beyond that explained by greater warming alone. Transition to rain dominance: CMIP6 projects an earlier transition from snow- to rain-dominated precipitation, particularly in autumn, with most of the Arctic Ocean, Siberia, and the Canadian Archipelago transitioning 1–2 decades earlier than in CMIP5. Snowfall ratio declines faster in CMIP6 (from ~0.7 to ~0.3 by 2100 vs ~0.4 in CMIP5). Winter and spring remain largely snow-dominated on average, though the snowfall ratio declines more rapidly in CMIP6 (winter from ~0.85 to ~0.75 vs ~0.8). Under 1.5–2 °C global warming, Greenland and Norwegian Seas transition to rainfall-dominated annually (more robustly in CMIP6), while Pacific-side seas mostly remain snow-dominated; at 3 °C warming, most regions except the Pacific side become rainfall dominated annually, but winter remains snow-dominated in most regions.

The findings demonstrate that CMIP6 projects larger and faster Arctic precipitation increases than CMIP5 and an earlier shift to a rain-dominated regime, particularly in autumn. These changes are driven by stronger Arctic amplification, greater sea-ice loss and open water expansion, enhanced poleward moisture transport, and higher sensitivity of precipitation to warming in CMIP6. The accelerated transition has significant implications for Arctic climate and ecosystems, including reduced snow cover duration, amplified warming via albedo feedbacks, increased winter CO2 and methane fluxes from soils and thawing permafrost, altered soil moisture and groundwater with flood risk implications, and more frequent rain-on-snow events that can devastate reindeer, caribou, and muskox populations. Hydrological impacts may include changes in river discharge and freshwater input that influence sea-ice melt, ocean stratification, circulation, and primary productivity via altered light transmission through sea ice. For Greenland, increased snowfall in the interior may partially offset melt, but greater rainfall along the southern and coastal margins may destabilize these regions and accelerate sea-level contributions. The results suggest that impacts previously expected at ~2 °C of global warming could occur under 1.5 °C, underscoring the urgency of mitigation.

CMIP6 models indicate amplified hydrological changes in the Arctic relative to CMIP5, with larger and faster precipitation increases and an earlier, more extensive transition to rain-dominated conditions—especially in autumn and in several regions even under 1.5–2 °C global warming. These shifts are linked to stronger Arctic warming, greater sea-ice loss and open water, enhanced moisture transport, and higher precipitation sensitivity to warming. The earlier regime shift carries wide-ranging climatic, ecological, and socio-economic consequences and implies that more stringent mitigation policies are required, as changes once associated with 2 °C warming may materialize at 1.5 °C.

Observational constraints on Arctic precipitation are uncertain: GPCP likely underestimates precipitation and poorly captures interannual variability relative to ERA5 and other reanalyses. Validation relies on ERA5 reanalysis rather than direct observations. Only one ensemble member per model was analyzed. CMIP6 exhibits larger inter-model spread in projected precipitation, temperature, open water, and VIMF, indicating greater uncertainty. Analyses focus on the Arctic region defined as 70–90°N. Projections are primarily evaluated under high-end forcing scenarios (RCP8.5/SSP5-8.5), though relative relationships with temperature are assessed to be scenario-insensitive.