On the relationship between snow grain morphology and in-situ near infrared calibrated reflectance photographs

Abstract

Seasonal and permanent snow cover a significant portion of our planet, and its impact on climate is significant. Through specific thermophysical properties, snow controls radiative and turbulent fluxes between the ground and the atmosphere, but many aspects of the energy balance are poorly understood due to lingering uncertainties regarding snow properties, such as grain size in particular. Rapid and accurate measurement method has yet to be developed given the reality of field and laboratory logistical constraints, and the sensitivity of snow to any sort of manipulation.

In this paper, we investigate the relationship between snow grain morphology parameters measured from visible (traditional) snow grain photography and optical diameter estimated from Near-InfraRed (NIR) reflectance photographs of snow walls. A total of 51 snowpits were analyzed during our International Polar Year field campaign across a 1000 km South-to-North transect over Eastern Canada. We compared the NIR measurements with the theoretical snow albedo model of Kokhanovsky and Zege (2004). Results show the large difference between the snow specific surface area (SSA) of snow grains derived from snow albedo model and the geometrical (visual) diameter. From three different snow grain classes which can be distinguished from traditional photography, linkages can be made with shape factors required in the optical model in order to retrieve optical grain size from NIR photography.

Introduction

Snow, one of the most important components of the cryosphere, covers up to 50% of Earth's landmasses during the winter season. Given its thermophysical properties, snow controls both radiative and turbulent exchanges between the ground and atmosphere, playing a crucial role on how the cryosphere reacts to climate change. Of particular relevance, net shortwave radiation is highly influenced by the presence or absence of snow on the ground. The high albedo of snow significantly reduces absorbed downwelling shortwave radiation particularly in the near infrared portion of the spectrum (Li et al., 2001), exerting a significant control on both melt timing and magnitude. Particularly, radiative transfer of shortwave energy is dominated by grain size and shape (e.g. Warren, 1982, Zhou and Li, 2002). Furthermore, it is also well known that snow morphology affects thermal conductivity and diffusivity (e.g. Mellor, 1977, Sturm et al., 1997, Langlois and Barber, 2007). However, previous work highlighted the lack of knowledge with regards to snow morphology/metamorphism and improved observations are needed to correctly parameterize the effect of snow on global surface energy balance (Massom et al., 2001, Eiken, 2003, Matzl, 2006, Picard et al., 2009, Dominé et al., 2008).

Hence, in order to properly evaluate the impact of snow on the changing cryosphere at global scales, it becomes necessary to improve the retrieval of snow grain size in current models and remote sensing signals given the lack of field measurements arising from the sampling constraints of such variable. Some of the literature suggests that, ‘grain size’ is poorly defined and measured with repeatability problems (e.g. Dominé et al., 2006). Hence recent work focused on defining grain structure and morphology rather than ‘size’. Since the morphology is extremely variable and can change in the matter of hours (e.g. Colbeck, 1989, Arons and Colbeck, 1995, Dominé et al., 2008, Langlois et al., 2008), validation of such models with accurate field measurements has yet to be done. Of particular relevance, most global snow mass balance algorithm make use of passive microwave radiative transfer principles, where large uncertainties are related to the poor definition of snow grain size profiles (e.g. Grenfell and Warren, 1999, Mätzler and Wiesmann, 1999, Roy et al., 2004, Foster et al., 2005).

Snow grain morphology is controlled by snow metamorphism (e.g. Colbeck, 1982, Schneebeli and Sokratov, 2004). Snow grains change size and structure through different metamorphism mechanisms given dry or wet conditions. In dry conditions, temperature gradient metamorphism rises from the temperature difference between snow grains in the vertical direction whereas the warmer grains act as the source of mass in the vapor phase and the colder as a sink (e.g. Colbeck, 1983, Gubler, 1985). Large elongated grains (prisms) are usually found under these particular conditions, forming a ‘hoar’ layer typically found at the bottom, near the ground. Equilibrium metamorphism is also found in dry conditions where the bottom grains are at equilibrium with water vapor at a higher density than the upper grain. The high Specific Surface Area of snow (SSA), i.e. the ratio of surface area to volume, provides a lot of energy to induce microscale heat and mass transfer (e.g. Bader et al., 1939, Colbeck, 1982) changing the structure of the snow grain through a decrease in SSA (Cabanes et al., 2002). In wet conditions, snow grain metamorphism will be driven by the three wet snow regimes namely pendular, funicular and saturated regimes. In the pendular regime, heat transport is possible and is the main mechanism behind bond growth and structural coarsening (Blackford, 2007). Snow grains tend to be separated from each other in the funicular regime, and the smaller grains are melted from heat flow (colder melting temperature) while larger snow grains are expected to grow due to the adhesion of water to the cold ice crystals.

For many years, scientists have struggled to measure snow grain morphology parameters such as axis length, volume, area, and more recently optical diameter (d_opt). The optical grain diameter corresponds to the diameter of non-contracting spheres with the same surface area and the same ice volume as the snowpack under consideration, and thus with the same SSA (Grenfell and Warren, 1999). The fact that SSA can be directly converted to optical grain size was discussed by Mitchell (2002). The optical grain size can be estimated from Kokhanovsky and Zege (2004) such that: $d_{\text{opt}}=6\frac{V}{S}$, where V and S are respectively the average volume and surface of grains. For monodispersed spheres, d_opt is equal to their diameter. But, for example, the effective size of long needles is quite small, because the optical diameter is close to the diameter of the needles; for disks of large diameter, the optical grain size could also be small, depending on their thickness (Mätzler, 1997). Stellar snow crystals can have a maximum extent of 1 cm, but their thickness is only 20 to 40 µm. Given the sensitivity of snow grain's morphology to change in local temperature and wetness, field measurements usually contain large errors and laboratory experiments are long and natural conditions difficult to duplicate (Hoff et al., 1998, Dominé et al., 2001). Snow grain optical diameter is very hard to measure on the field and requires thorough analysis through gas adsorption techniques (Dominé et al., 2006) or stereological measurements (Matzl, 2006). Previous work did established a significant relationship between infrared reflectance (R) and optical diameter, however the methods need further validation as they were developed in specific conditions and samples. Micro-photographs of snow samples and individual grains have been widely used in numerous studies and results shown that the derived geometrical diameters (axis average, derived from projected surface) can provide fair estimates on metamorphism timing and magnitude (e.g. Colbeck, 1983, Painter et al., 2007, Langlois et al., 2008), but no significant information on morphology can be retrieved.

Promising results were recently obtained using infrared photography (Matzl and Schneebeli, 2006) and spectroscopy (Painter et al., 2007) to determine snow grain optical diameter and SSA. Using a near infrared converted digital camera to acquire reflectance values of snow walls and stereological SSA measurements, SSA was obtained with R² of 0.91 (Matzl, 2006) from measured reflectance values. This method is quick and accurate where a full profile can be obtained in a matter of minutes rather than hours with typical snow micro-photographs. Furthermore, NIR photography also provides detailed information on snow macrostructure (layering information) such as wind and/or compaction layers, ice crusts which are critical in snow avalanche studies. Linkages between NIR reflectance parameter and the different snow grain size parameters need to be further explored.

Recent modeling works done by Kokhanovsky and Zege (2004) did provide accurate relationships between infrared reflectance and snow grain optical diameter (Picard et al., 2009). It was shown that a simple analytical model can be employed to calculate the infrared reflectance given different shape factors, but validation with field measurements is still required. Hence, the specific objectives of this paper are to a) establish a quick, accurate and calibrated NIR snow photography method, b) to derive snow grain optical diameter using the method in a) coupled with the modeling approach of Kokhanovsky and Zege (2004), c) to use traditional micro-photographs of snow grains in order to provide snow grain shape information required in the modeling approach, and d) to discuss the problem of shape analysis by comparing optical diameter values derived from b) with averaged geometrical diameter measured from c).