Snow height evolution

Table 5.2.: Comparison of different snow height determinations. The probed snow depth is com-pared to radar-determined snow height d_snow either by using the mean wave speed in dry snow ¯v = 0.237 m/s or the mean wave speed determined at each snow pit individually for each measurement day,vpit.

Date probed

Using the P-visualization (Fig. 5.3, red ellipses), the development of the snow height determined from GPR is clearly displayed. Backscatter above the snow surface with an approximately similar amplitude to the surface reflection occurred only during strong precipitation events (Fig. 5.3a and c; Fig. A.1, 13.02.09, 20.02.09). The radargrams of both measurements influenced by the precipitation show horizontally non-persistent incoherent reflections above the snow surface. As the measurements were conducted in the time domain, the effect is comparable to weather radar applications, i.e. moving snow flakes generate backscatter during illumination by the radar beam. In comparison, other automatic snow-height sensors also receive a more noisy signal during snow-precipitation events (e.g. ultrasonic sensors, Bavay et al., 2009). In our data set the surface reflection is in all cases clearly detectable. No further data calibration is necessary.

The radar-determined snow height using an average wave speed in dry snow, ¯v = 0.237 m/ns (Heilig et al., in press) varies for all measurements in dry snow conditions in winter 2009 by less than 8% in comparison to the probed snow depth above the radar box (Tab. 5.2). In a 2.0 m high snowpack the miscalculated snow height is therefore ±16 cm. This error is, on the one hand, due to inhomogeneous vertical snowpack conditions, which result in variations in the wave speed between adjacent layers and, on the other hand, due to uncertainties in snow probing.

It is very likely that the box (Fig. 5.1) was compressed by the snow masses and therefore had no plain surface as discovered after digging out the antennas in April. A difference in snow height of up to 10 cm above the box could occur due to the compression in the middle of the

box. While calculating a separate mean wave-speed value for each radar measurement, the difference to the probed snow depth increased up to 15% for the measurement of the 13.02.09 and less than 10% for all other measurements. A contributing factor to these specific variations is that the spatial variability of the snowpack conditions above the antennas and within the test site was larger than expected. This results in a remarkable variability in the characteristic of specific snow layers as layer thicknesses or layer locations. Even two measurements on adjacent days without melt processes but with low precipitation rates (Fig. 5.4B) and just about 3 m next to each other, result in different calculated average wave speed values (Tab. 5.2, 19.02.09, 20.02.09). The accuracy of the snow height determination using v¯ for the conversion of the two-way travel time of the radar data, could have been distinctly higher, if the uncertainty in snow probing was lower. Our estimate is based on a maximum uncertainty of 10 cm in snow probing. Disregarding this error, the accuracy of the radar-measured snow height utilizing v¯ is adequately accurate and slightly above the accuracy range of ultrasonic snow sensors (±3 cm, Egli and Jonas, 2009). In comparison to the measurements of Gubler and Hiller (1984), who determined the snow water equivalent with an inaccuracy of less than 5% in comparison to manually conducted measurements for dry snow densities between 200–400 kg/m³, the here presented data is at about 7% slightly above this range, if the mismeasurement of the probed snow depth is not corrected. Taking this uncertainty into account the accuracy increases to differences of less than 5%, which is likely the accuracy limit, due to uncertainties in manual probing and allocating reflections to height (Gubler and Hiller, 1984). Recent work of Marshall et al. (2005) presents FMCW radar measurements with an uncertainty in SWE and snow depth of about 10%. For other climatic regions, such as e.g. maritime areas with distinctly higher average densities or on the opposite more continental areas with distinctly lower mean density values, respective v¯have to be determined. Of course the utilization of height-dependent wave speeds v(h) are desirable, but seldomly easily obtained. Using a mean wave speed for a certain climatic region and time of the year is a justified approximation. If similar variabilities in density records of dry snow conditions to the values presented by Heilig et al. (in press) with a coefficient of variability of CV = 6% are observable, we assume that the accuracy range of snow height determination utilizing upward-looking GPR is adequately accurate for other climatic regions as well. The SWE in dry snow conditions can be determined in combination with an external snow height measurement directly. If the snow height is known a mean density for the radar measured snowpack above the antennas can be calculated and therefore the SWE be determined.

5.3 Results and Discussion

Figure 5.4.: A: values of the recorded reflection amplitude of the addressed snow layers and the snow surface. B: meteorological parameters manually recorded at the test site (GRUEN) in comparison to the measured snow height (HS) and temperature (TEMP) at the AWS. C: comparison of the three differently determined snow heights above the radar box: the probed HS (HS probe), HS calculated withv¯(HS v_mean) and HS calculated utilizing the density of the corresponding snow pits (HS v_pit).

The black horizontal line segments indicate the position of the internal layers calcu-lated with ¯v. D: the strain rates S^∗ (eq.5.5) of the respective layers (left ordinate)

Figure 5.5.: Calculated effective reflectivity values R (blue diamonds) and measured densities (red lines) of the snow pits measured nearby the radar measurements.