**3. Results**

The data set includes vertical profile measurements of reflectance, SSA, macrophotography-based traditional grain size, and density from various snow types with varying grain size and density in a taiga snowpack. The snow height of 0 cm was set to the ground surface. The height of the snow and air temperature for spring 2017 are presented in Figure 4. An overview of all data used in the study is presented in Figure 5.

## *3.1. Snowpack Properties*

The height of snow increased before 21 March and melting had already started in 3 April, as shown in Figure 4. The air temperature was close to zero or positive (in ◦C) on 16 March, 21 March, and 3 April. Layer properties had temporal variations; however, there were some lasting similarities. The hard crust layer at approximately 30 cm height (red in Figure 5c) was observed in all snow pits. The traditional grain size was larger below than above that layer, which is typical, because grain growth is largest at the bottom due to the higher temperature gradient. Between 30–40 cm height of snow, the grain shape changed from round (pink in Figure 5c) to faceted (light blue in Figure 5c), and grain size increased simultaneously. In mid-winter, the bottom layers consisted of smaller grains than in the layers above those. Typically, traditional grain size was larger in faceted crystal and depth hoar layers than layers with rounded grains, as expected based on the temperature gradient. Density was lowest in the top 5–15 cm of the snowpack. Maximum density was found in the 5–15 cm bottom layer of the snowpack, and in the crust layer around 30–40 cm height. Based on those observations, the reflectance and SSA measurements were made at differing air temperatures and snow conditions with variating grain size, grain type, and density.

**Figure 4.** The height of snow (blue) and air temperature (green) measured at the Intensive Observation Area (IOA). Vertical black lines indicate the measurements made with QST.

**Figure 5.** Reflectance from QST, specific surface area (SSA) from IceCube, grain type, macrophotography-based traditional grain size, and density profiles of snow are plotted in columns (**<sup>a</sup>**–**<sup>e</sup>**) for all dates. Original SSA data are marked with circles and interpolated values with dots in column (**b**). SSA for 21 March 2017 is plotted from the first measurement of three for every height. Grain type abbreviations are described in [4].

#### *3.2. Repeatability of SSA Measurements*

Repeatability of the SSA measurements with IceCube was tested to confirm the reliability of observations for the following analysis. First, the measurement method was tested by repeating measurements with the same sample; then, sampling was tested by sampling snow to different densities. The repeatability of the IceCube measurement method was tested on 21 March 2017. Every sample from the profile was measured three times with IceCube. The mean standard deviation of the SSA was 0.34 m<sup>2</sup> kg−1, which means a relative standard deviation (RSD) of 2.0%. RSD is calculated by Equation (3)

$$RSD = 100\,\sigma/\mu \tag{3}$$

where *σ* is the standard deviation and μ is the average. The repeatability of IceCube sampling was also tested on 21 March 2017. Precipitation particles in the surface layer and faceted crystals in the bottom of the snowpack were both sampled three times. The samples were taken next to each other so that snow was as homogenous as possible for all the samples. The samples were packed to different densities, measured with IceCube and weighed for density calculation. The average standard deviation was 0.65 and RSD was 2.85% (Table 1). Since the total RSD was low (<5%), SSA is relied on as a truthful value in the subsequent analysis.

**Table 1.** The SSA from three samples of precipitation particles at the surface and faceted crystals in the bottom of the snowpack. Standard deviations (STD) and relative standard deviations (RSD) from Equation (3) are calculated for both surface and bottom.


#### *3.3. Comparison of QST Reflectance Profiles*

Reflectance profiles were measured with QST from both the snowpack and the sampled snow in 21 March and 3 April 2017. Both profiles were made next to each other at the same snow pit. The reflectance profiles were compared with three wavelengths (1160, 1260, and 1310 nm), which are used later in the study.

The comparison results showed a strong correlation between the reflectance profiles directly from the snowpack and the sampled snow with a correlation coefficient of 0.92–0.94 (Table 2), as was expected. Bias varied from −0.020 to 0.016 and RMSE varied from 0.022 to 0.040 (Table 2). According to the results, IceCube sampling had only a small effect on the reflective properties of snow (microstructure) at 1160, 1260, and 1310 nm wavelengths.

**Table 2.** Bias, root-mean-square error (RMSE) and correlation coefficient ( *R*) for QST reflectances from the snowpack profile and the sampled snow.


#### *3.4. Empirical Relationship between SSA and Reflectance*

Previously shown relationships between snow microstructural parameters and albedo or hemispherical reflectance are described in the Section 1. Similarly, we hypothesized that reflectance measured with QST could have an empirical connection to the SSA. The empirical relationship between SSA and reflectance was studied with 1160 and 1260 nm reflectance-dependent coefficient Q (see Section 2.4). Examples of the measured reflectance are presented in Figure 6, where the upper part of an ice absorption feature was located close to 1160 nm and bottom of it close to 1260 nm. The correlation between SSA and Q was high with correlation coefficients between 0.85 and 0.98 (Table 3). Linear fits are presented for measurements from the snowpack and the sampled snow in Figure 7. The fit is better for the QST measurements from the snowpack (Figure 7a). Single outliers existed for observations in 21 March from sampled snow, which are visible in the scatter plot of Figure 7b. There is approximately 0.2 bias in the linear fits between QST measurements from the snowpack and from the sampled snow. The results prove that an empirical relationship between SSA and reflectance exists, although measurement of light, new snow and fragile depth hoar is challenging.

**Figure 6.** Example of reflectances measured on 7 March 2017 at heights of 73 cm (blue), 64 cm (cyan), 52 cm (magenta), and 34 cm (red). Vertical lines indicate wavelengths 1160 nm and 1260 nm.

**Table 3.** Correlation coefficient (*R*) between SSA and Q for the QST measurements from the snowpack profile and profile of the sampled snow.


**Figure 7.** Scatter plot of Q with Equation (2) (*y*-axis) and SSA (*x*-axis) from the QST measurements made (**a**) directly from the snowpack profile and (**b**) the measurements from the sampled snow. Lines fitted to the points are marked.
