**5. Discussion**

The magnitude of changes in albedo prior to melt season seem to be linked strongly to vegetation characteristics, whereas the changes in melt season timing seem to be more strongly linked to climatic factors. The influence of vegetation on surface albedo prior to melt can be understood by considering the difference between the albedo of vegetation and the albedo of snow. Even a small increase in the vegetation cover can alter the surface albedo by several absolute albedo percentage units [49,50], whereas changes in climatic factors prior to melt affect the snow depth and/or the surface crystal structure, and have smaller (but still significant) impact on the surface albedo. Changes in the number of days with snow on the trees are an obvious exception to this, putting emphasis on the influence of climatic parameters on surface albedo. Increased precipitation in vegetated areas would mean increased albedo while, in open areas, such as tundra, more precipitation would more strongly affect the depth and surface crystal structure of the snow pack, with a weaker effect on the surface albedo.

The start date of melt is linked to the air temperature. Increasing temperatures would result in earlier melt onset. However, in this study the focus was on the temperature prior to melt. That means that the time period in question changes from year to year. An increase in air temperatures prior to melt onset would mean that at the start of melt the snow pack is already warmer and the surface crystals may have been affected by melt and sublimation. Conversely, lower air temperatures, such as in the area around the borders of Russia, Mongolia and China, would result in a colder snowpack, and, in the case of a thin snow cover, resulting from low rates of precipitation, colder ground. This would mean slower melt. Both the end date of melt and the length of the melt season are affected by the air temperature, snow depth, snowpack characteristics, and the ground temperature. These are all regulated by climatic factors. For changes in vegetation to cause significant changes in the melt season timing, such as observed in this study, the vegetation would need to change considerably, for example from bare ground to tree cover. At larger spatial scales, such changes take more than 34 years.

The relationship between vegetation and climate is not straightforward. Existing studies of changes in vegetation, permafrost and impurities in snow [26,51–55] reveal similar spatial patterns of consistent change as do the melt season parameters, but the effect is not the same in all areas. None of the aforementioned factors alone are able to explain the changes over the whole study area. In fact, they can have opposite effects on the albedo in different regions.

The inconsistency in the effects of vegetation on winter time albedo can be partly explained by the different response of vegetation to the changing climate. According to Xu et al. [26], in Eurasia the normalized difference vegetation index (NDVI) is positively correlated with warming temperature, whereas in North America the effect varies in different regions and patterns of greening and browning in North America are fragmented [26]. It is noteworthy that while the NDVI data describe growing season conditions and are thus not directly translatable to winter conditions, they should be considered an indicator of changes that may also be visible in winter, such as growth in shrub size and coverage. The observed changes in vegetation cannot be directly translated to changes in surface albedo. According to Myers-Smith et al. [56] shrub growth is responsive to different drivers in different regions. Sturm et al. [50] found that if shrubs protrude above the snow and cover 10% of the surface, the albedo will decrease by 30%. The spatial coverage of continuous permafrost shows similar spatial patterns with many of the trends for the melt season parameters. For example, at the southern edge of Central Siberian Plain and in Labrador Peninsula continuous permafrost ends in the same area as the trends for melt season change. The difference in trends for melt season parameters can be due to changes in permafrost coverage causing changes in land use, such as vegetation or formation of melt ponds, or by different response of vegetation and snow cover to climatic changes in areas with and without permafrost.
