*3.1. Fit Statistics*

The relevance of including both air temperature and forest structure as predictor variables can be appreciated when looking at the coefficients of determination (R2), which are presented in Table 1. Although snow cover (*SC*) alone explained 55–81% of the variance (depending on model complexity and albedo broad band), the addition of both air temperature and forest structure (volume or biomass) further increased the percentage of variance explained to 85–88%.

**Table 1.** Coefficients of determination (R2) for all candidate models. Number of observations = 4,524,377. "VIS" = Visible broadband (0.3–0.7 μm); "NIR" = Near-infrared broadband (0.7–5 μm); "SW" = Shortwave broadband (0.3–5 μm); "BS" = Black-sky (directional hemispherical); "WS" = White-sky (bidirectional hemispherical); "ef" = Endmember fraction; "SC" = NDSI snow cover fraction; "T" = Air temperature (2 m; ◦C); "V" = Stand volume (m<sup>3</sup> ha−1); and "B" = Stand aboveground biomass (t ha−1).


In general, for any given model permutation, the fits for the VIS broad band explained more of the variance than the fits for SW and NIR broad bands. Furthermore, the fits for black-sky albedo explained more of the variance than the fits for the white-sky albedo (Table 1). The models fit at the effective spatial resolution did not lead to as large R<sup>2</sup> improvements as those reported elsewhere [81], although it should be noted that the study of ref. [81] was restricted to forests where there is a much larger variation in vegetation structural controls of the surface albedo.
