**2. Data**

This study is based on the 5-day mean surface albedo data of the Satellite Application Facility for Climate Monitoring (CM SAF, funded by EUMETSAT) CLouds, Albedo and RAdiation second release Surface ALbedo (CLARA-A2 SAL) data record [37,38], which is constructed using Advanced Very High Resolution Radiometer (AVHRR) data. The albedo is defined as the broadband shortwave directional-hemispherical reflectance, i.e., the black-sky albedo. The retrieved albedo corresponds to the wavelength range 0.25–2.5 μm and the observations are averaged to a 0.25◦ × 0.25◦ grid, which is also the resolution of the final product. The albedo values are given in the range 0–100%. At the time of the analysis this was the longest available homogeneous data record of surface albedo.

The basis of the derivation of the 5-day mean albedo product used here is similar to CLARA-A1 SAL [39]. The albedo values for a five day period are first determined by observation and then averaged. After cloud masking the satellite data, the effect of topography and inclined slope and related shading on location and reflectance of the satellite data is corrected. The land pixels are then corrected for atmospheric effects on the radiation. The atmospheric correction utilizes a dynamic aerosol optical depth (AOD) time series [40] as input. The AOD time series was constructed using the total ozone mapping spectrometer (TOMS) and ozone monitoring instrument (OMI) aerosol index data [40]. The scattering properties of the surface are described by bidirectional reflectance distribution functions (BRDF) for different land-use types. The land-use classes are derived from four different land-use classifications, using always the classification which has been constructed from data that is temporally closest to the observation in question. Finally, the 0.6 and 0.8 micrometer (AVHRR channels 1 and 2) albedos are converted into broadband albedo. The reflectance characteristics of snow surfaces vary between different snow types [41]. Therefore the albedo of snow- and ice-covered areas is derived by averaging the broadband bidirectional reflectance values of the AVHRR overpasses into pentad and monthly means. The albedo of open water, such as oceans, is constructed using solar zenith angle and wind speed. The existence of sea ice is verified using the Ocean and Sea Ice Satellite Application Facility (OSI SAF) sea ice extent data [42].

The data record has been validated against in situ data and compared with the Moderate Resolution Imaging Spectroradiometer (MODIS) MCD43C3 edition 5 data set [43,44]. The mean relative retrieval error of CLARA-A2 SAL is −0.6%, the mean root mean square error (RMSE) is 0.075 and the decadal relative stability (over Greenland Summit) is 8.5%. Larger differences between the in situ measurements and the satellite-based albedo value are mostly related to the heterogeneity of the land surface within CLARA-A2 SAL pixels [45]. A comparison between CLARA-A2 SAL and MODIS MCD43C3 showed that the two products are in good agreement. The relative difference between the two products is typically between −10% and 10%, with the global mean CLARA-A2 SAL surface albedo being 2–3% higher than the MCD43C3 global mean surface albedo for some periods. One has to take into account that the SAL product includes a topographic correction in mountainous areas, whereas the MODIS product does not [43], and that mountains typically cause underestimation of albedo due to shadowing [37]. The water areas are excluded from the comparison as the MODIS product is not defined over water areas or sea ice.

Our study utilizes the global version of the CLARA-A2 SAL products and covers the land areas between latitudes 40◦N and 80◦N, and the years from 1982 to 2015. Using a 5-day mean albedo limits the role of possible individual low-quality albedo values with large retrieval errors (due to observation geometry, cloud contamination or geolocation error). Furthermore, using sigmoid fitting (described in Section 3) for the analysis limits the effect of possible erroneous individual mean albedo values.

The influence of climatic parameters on melt season and albedo was studied using ERA-Interim reanalysis data [46] for 14 day period before the previously defined date for the onset of melt. The parameters extracted from the data were air temperature (2 m), wind speed (10 m above ground), accumulated precipitation, amount of snow fall (giving also the accumulated rain) and the number of days on which the maximum temperature during that period was above 0 ◦C, −4 ◦C and −10 ◦C. These parameters were chosen due to their possible effect on snow reflectance, metamorphism and albedo. The air temperature affects the amount of liquid water and heat flux within the snow pack. Wind can affect the surface albedo by affecting the mechanical breaking of the surface crystals, by producing wind related surface structures such as ripples and ablation and accumulation areas. It can also affect the amount of vegetation visible above the snow surface and fraction of bare ground by removing snow partly or altogether from some areas. It also typically affects the amount of snow on trees. In the case of evergreen trees, this can have a significant effect on surface albedo. Precipitation can affect the surface albedo through adding fresh snow crystals on the surface and on vegetation and by affecting the snow depth. In the case of rain-on-snow, this can bring heat into the snow pack thus affecting the melt processes. The three temperature thresholds were chosen based on the relationship

between air temperature, snow metamorphism and albedo. The 0 ◦C was chosen since it is the melting point for snow in normal conditions. The −4 ◦C was used to take into account the fact that snow metamorphism starts already at sub-zero temperatures. In the wide variety of snow albedo models, the simplest parameterizations presume a steady albedo for colder temperatures, and then a linear decline in snow surface albedo for air temperatures from −5 ◦C to 0 ◦C. At −4 ◦C the heating of the sun can already affect the snow surface crystals; −10 ◦C was chosen to represent a temperature at which the air temperature does not considerably affect the snow surface crystals, so if the maximum temperature of the day stays colder than this it can be presumed there is no change in the snow surface due to the temperature. The data was originally in 6-hourly temporal resolution (for snow fall 12 hourly) from which it was further processed to daily values. The resolution of the ERA-Interim data was 0.25◦.

The role of land use in the trends in melt season albedo and timing is assessed using data from GlobCover2009 [47]. The data was coarsened to the same resolution as the melt season data (0.25◦) by choosing the most common land-use class within the melt season grid cell. Figure 1 shows the GlobCover data at CLARA-A2 SAL resolution. The GlobCover land-use classes present in the study area are listed in Table 1.


**Table 1.** GlobCover2009 classes found to be most common within one surface albedo (SAL) resolution unit in the study area and the number of occurrences of each class as the most common land-use class in one resolution unit of melt season data.

**Figure 1.** GlobCover2009 land-use classes coarsened to melt season data resolution.
