Soil temperature beneath snow cover
Author: Tyler Herrington
Figure 1. Relationship between the averaged three month (Dec-Jan-Feb) offset between near surface soil temperatures and air temperatures as a function of snow depth, based on satellite observations, and for eight different reanalysis products.
Changes in permafrost are among the most important impacts of climate change in the Arctic. Melting permafrost is not only expected to release large amounts of stored carbon, but freezing and thawing of soil also has a large impact on its hydraulic properties and stability, leading to further changes in the hydrological cycle and threatening infrastructure. To predict these changes using climate models, it is necessary to understand how soil temperature, and hence soil freezing, is represented in climate models, and if that representation matches reality.
Snow cover is an important control on soil temperatures in cold regions, insulating soils from atmospheric processes in winter. This can mean that when snow is present the soil is often substantially warmer than the air above; the difference can be as much as 20 degrees centigrade at high latitudes (Zhang, 2005).
Vegetation, snow cover and soil properties can vary drastically over just a few kilometers. A significant challenge in the C1W modelling initiative is accounting for the geographic variability of these factors and their impact on the surface climate. Field measurements in the Canadian arctic are sparse and only represent local conditions, so that they can not easily be extrapolated. Satellite data, ideally constrained or calibrated by field measurements, can mitigate this problem to some extent, but satellite data are limited to what the satellite can see at the surface. Physical models are needed to extrapolate the information to what the satellite cannot see. For this purpose, so-called reanalysis products are used, which integrate (or assimilate) field observations and data from satellites into a physical model, which then estimates a range of physical parameters (including soil temperature) where they are not directly observed. There are at least seven reanalysis products that provide global estimates of snow cover and can be used to estimate soil temperature. Each reanalysis product is constrained by a different set of observational data, and the models that they are coupled to also vary between products. As a result, different products will produce different estimates of snow cover, and soil temperature.
Fig. 1 shows (for the three months of Dec-Jan-Feb) the average temperature offset between the soil temperature and air temperature under different snow cover depths from satellite observations (Observed), for eight different reanalysis products. The relationship shows that temperature offsets are much larger when snow is present, though above a snow depth of about 0.3 m - 0.4 m the effect begins to diminish. Some products are generally able to capture the observed relationship of snow insulation, but others greatly underestimate the insulation of snow.
Due to variations in snow cover depth from different reanalysis products, the strength of the coupling between snow cover and soil temperatures also varies. In many cases, reanalysis products do not sufficiently represent the insulation effects of soil by snow cover, which can contribute to substantial errors in modelled soil temperature estimates at high latitudes.
In previous work it has been found that the soil temperatures in many of the reanalysis products are too cold, and these results may explain some of this error. These findings may also have implications for climate models that are used to generate future projections, since the underlying physical representations are very similar; if the insulating effect of snow is underestimated, the impact of receding snow cover on permafrost would also be underestimated. Further work will explore whether products which underestimate the observed snow insulation strength are more likely to exhibit cold biases in soil temperature.