The impact of high latitude climate warming on Arctic snow cover and its insulating properties has key implications for the surface and soil energy balance. Few studies have investigated specific trends in Arctic snowpack properties because there is a lack of long-term in situ observations and current detailed snow models fail to represent the main traits of Arctic snowpacks. This results in high uncertainty in modeling snow feedbacks on ground thermal regime due to induced changes in snow insulation. To better simulate Arctic snow structure and snow thermal properties, we implemented new parameterizations of several snow physical processes—including the effect of Arctic low vegetation and wind on snowpack—in the Crocus detailed snowpack model. Significant improvements compared to standard Crocus snow simulations and ERA-Interim (ERAi) reanalysis snow outputs were observed for a large set of in-situ snow data over Siberia and North America. Arctic Crocus simulations produced improved Arctic snow density profiles over the initial Crocus version, leading to a soil surface temperature bias of −0.5 K with RMSE of 2.5 K. We performed Crocus simulations over the past 39 years (1979–2018) for circumpolar taiga (open forest) and pan-Arctic areas at a resolution of 0.5°, driven by ERAi meteorological data. Snowpack properties over that period feature significant increase in spring snow bulk density (mainly in May and June), a downward trend in snow cover duration and an upward trend in wet snow (mainly in spring and fall). The pan-Arctic maximum snow water equivalent shows a decrease of −0.33 cm dec⁻¹. With the ERAi air temperature trend of +0.84 K dec⁻¹ featuring Arctic winter warming, these snow property changes have led to an upward trend in soil surface temperature (Tss) at a rate of +0.41 K dec⁻¹ in winter. We show that the implemented snowpack property changes increased the Tss trend by 36% compared to the standard simulation. Winter induced changes in Tss led to a significant increase of 16% (+4 cm dec⁻¹) in the estimated active layer thickness (ALT) over the past 39 years. An increase in ALT could have a significant impact on permafrost evolution, Arctic erosion and hydrology.

Satellite observations of snow‐covered regions in the microwave range have the potential to retrieve essential climate variables such as snow height. This requires a precise understanding of how microwave scattering is linked to snow microstructural properties (density, grain size, grain shape and arrangement). This link has so far relied on empirical adjustments of the theories, precluding the development of robust retrieval algorithms. Here we solve this problem by introducing a new microstructural parameter able to consistently predict scattering. This “microwave grain size” is demonstrated to be proportional to the measurable optical grain size and to a new factor describing the chord length dispersion in the microstructure, a geometrical property known as polydispersity. By assuming that the polydispersity depends on the snow grain type only, we retrieve its value for rounded and faceted grains by optimization of microwave satellite observations in 18 Antarctic sites, and for depth hoar in 86 Canadian sites using ground‐based observations. The value for the convex grains (0.6) compares favorably to the polydispersity calculated from 3D micro‐computed tomography images for alpine grains, while values for depth hoar show wider variations (1.2–1.9) and are larger in Canada than in the Alps. Nevertheless, using one value for each grain type, the microwave observations in Antarctica and in Canada can be simulated from in‐situ measurements with good accuracy with a fully physical model. These findings improve snow scattering modeling, enabling future more accurate uses of satellite observations in snow hydrological and meteorological applications.

Changes in mass, extent, duration, and physical properties of snow are key elements for studying associated climate change feedbacks in northern regions. In this study, we analyzed snowpack physical properties along a ‘mega’ transect from 47°N to 83°N (4,000 km) in northeastern Canada, which includes marked transitions between ecozones from boreal forest to subarctic and arctic ecosystems. Our unique dataset of 391 detailed snowpits acquired over the last 20 years, complemented with snow data from weather stations, shows that snowpack properties such as snow water equivalent, snow depth, density, grain size and basal depth hoar fraction (DHF) are strongly linked to vegetation type. Based on these results, we propose an updated classification of snow types in three classes: boreal forest snow (47–58°N), tundra snow (58–74°N) and polar desert snow (74–83°N), which is more appropriate to the study area than the general north hemisphere classification commonly used. We also show that shrub presence along the transect contributes to a significant increase in DHF development which contributes most strongly to the thermal insulation properties of the snowpack. Overall, our analysis suggests that snow–vegetation interactions have a positive feedback effect on warming at northern latitudes.