How Changes in Relative Humidity in the Polar Boundary Layer impact Arctic Amplification in Climate Models

The Arctic is warming three to four times faster than the global average due to multiple feedback processes – a phenomenon known as Arctic Amplification. Cloud feedbacks, in particular, represent one of the largest sources of uncertainty in projections of this amplified warming. Relative humidity (RH) is critical to these cloud feedbacks through its influence on cloud formation and radiation balance, yet changes in Arctic RH under a warming climate remain poorly understood.

Using 27 CMIP6 Coupled Model Intercomparison Project (CMIP6) models, this study investigates Arctic RH changes and their drivers by comparing historical conditions (1985-2015) with future projections under SSP5-8.5 (2070-2100). The multi-model mean reveals a robust vertical dipole pattern in surface-temperature-normalized RH changes across the Arctic. Near the surface (1000-925 hPa), RH decreases by up to 2 % K⁻¹ in winter, while mid-tropospheric RH (950-750 hPa) increases. This counterintuitive pattern – surface drying despite increased open ocean from sea-ice loss – is particularly pronounced during autumn and winter. The dipole signal is strongest over regions experiencing substantial sea ice loss, but remains visible at reduced amplitude over persistent ice regions, indicating both local (sea-ice driven) and broader (stability-driven) components to the RH response.

The multi-model mean, however, emerges from markedly different individual model responses. DIPOLE models reproduce the characteristic dipole pattern with drying near the surface and moistening around 1 km above the surface; DECREASE models show drying in both layers; INCREASE models show moistening at both levels. While DIPOLE and DECREASE models both exhibit a dipole pattern over ice-loss regions, INCREASE models do not, suggesting fundamental differences in model physics that are also evident in present-day RH distributions. Cloud liquid and ice water changes do not follow the dipole pattern but instead show increases across all groups, with inter-group differences in magnitude and vertical extent. Cloud liquid water increases peak near 925 hPa in all groups but are strongest over ice-loss regions in DECREASE and DIPOLE models, while DIPOLE models show strong cloud ice increases throughout the lower troposphere (surface–700 hPa), INCREASE and DECREASE models exhibit two distinct maxima at 850 and 500 hPa.

The primary driver of the dipole pattern is the transition from a predominantly stable atmosphere over sea ice (with an RH maximum near the surface) to a well-mixed atmosphere over open ocean (with an RH maximum at cloud base). This physical mechanism suggests that the DIPOLE models have a more realistic representation of moisture in the Arctic boundary layer and its response to sea-ice loss. If further analysis can rule out the behaviour of the INCREASE and DECREASE models, we expect that this will allow us to better constrain Arctic cloud feedbacks.