Impacts of snow impurities on sea ice transmittance 

Get Complete Project Material File(s) Now! »

Albedo measurements

Albedo measurements were performed with a custom-built radiometer (Solalb, developed at IGE following Picard et al. (2016b)). Light was collected using a cosine collector and guided through an optical fiber to a spectrometer (Maya 2000 PRO, Ocean Optics). Irradiance was measured at wavelengths ranging from 300 to 1100 nm, with a 3 nm resolution. More details about the cosine collector can be found in Picard et al. (2016b). The cosine was fixed at the end of a 3-meter aluminum pole which rested on a tripod 70 cm above the surface. At the other end, the operator manually controlled the arm and triggered the spectrometer. The horizontality was ensured by the operator within less than 0.3◦ using an electronic inclinometer mounted next to the cosine collector. Albedo determinations required two sets of measurements for reflected and incident light. Measurements of upwelling and a downwelling irradiance were made sequentially using the same cosine collector with the pole being manually rotated by 180◦. 10 spectra are automatically acquired for each measurements (upwelling and downwelling). No absolute or relative calibration was needed, but measurements had to be made under steady incident light conditions during the 30 s of the process, which seldom strictly prevailed during the Arctic spring. The setup therefore included a reference photodiode to measure light fluctuations at all times for subsequent correction. For both upwelling and downwelling irradiance measurements, the integration times was automatically adjusted in order to optimize the signal to noise ratio. A single operator could manage the entire process including albedo measurements along linear transects.

Snow physical properties

Here, snow physical properties refer to temperature, snow grain shape and geometric size, SSA and snow density. We first identified the main stratigraphic layers by visual inspection. For each layer, the average snow grain size and shape were determined using a hand lens. Snow temperature was measured at several depths from the bottom of the cover to up to 10 cm beneath the surface. Freeboard was reported when negative (when sea level was above the interface between snow and sea ice). The vertical profile of snow density was measured using a 100 cm3, 3 cm high box cutter. The collected snow sample was weighted using an electronic scale. According to Conger and McClung (2009), this method allows snow density measurements with an uncertainty of 11%. The main uncertainties concern the real volume extracted by the cutter depends on the type of snow. The density of superimposed ice layers was also measured when it was possible. Finally, vertical profiles of SSA were determined from the snow IR reflectance using the DUFISSS instrument (Gallet et al., 2009). Briefly, DUFISSS measures the albedo of a cylindrical snow sample 63 m in diameter and 25 mm thick at 1310 nm with an integrating sphere. The SSA is deduced from the albedo using a polynomial relationship. The correction concerning the determination of SSA of wet snow introduced by Gallet et al. (2014b) was not applied in this study because it did not induce significant changes on albedo simulations at the end. Uncertainty in SSA determinations is 10% under good conditions (Gallet et al., 2009). Melting can occur if the sample is not handled fast enough, which leads to a lowered SSA value. That is a recurrent issue we had to deal with after melt onset. Special care was taken to keep every sampling tools as cold as possible, for example by placing instruments in bottom snow layers when the surface was melting.

Sampling Protocol

Data presented in this study were collected either in snowpits or along transects. Snowpits : Albedo was measured first since it requires a pristine area. A minimum of 3 measurements were made depending on sky conditions and light variations. All of them were performed facing the sun to avoid any shadow from the operator and the equipment. All stratigraphic measurements were carried out along a one meter long shaded trench. Our objective was to conduct all samplings at the same place in order to fully characterize physical and optical properties of the snow at each station. One or two snowpits (requiring three hours of work each) were sampled each sampling day. Fewer snowpits were sampled in 2016 (10 versus 35 in 2015) because the snowpack was already ripe (i.e isothermal at 0◦C and melting throughout) before sampling operations. Snowpit locations were randomly chosen around the ice camp. Particular efforts were made to sample the widest range of snowpack depth possible in order to catch spatial variability. Transects : Albedo was also measured every 5 m along transects (from 100 m to 150 m long) in order to catch the spatial variability. All the equipment was placed on a sled to make the transport of equipment easier between each measurement station.

Data processing

Upwelling and downwelling irradiance raw acquisitions require several processing steps before the albedo can be obtained. During the field campaigns, spectra were visually checked at the end of the sampling day. Unrealistic data, based on qualitative criteria, were rejected. The first step of processing was to remove the systematic offset in both acquisitions caused by dark current and stray light effects. This offset was approximated for each acquisition as the mean signal at low wavelength (between 200 nm and 260 nm), because there is no incoming photon in this wavelength range. Dark current was assumed to be constant over the entire wavelength range. Then, spectra were divided by their corresponding integration time. Our cosine collectors have been previously characterized on an optical bench in order to assess their exact angular response (Picard et al., 2016b). This angular response was then used to correct the upwelling irradiance measurements depending on the sun zenith angle (SZA) during the acquisition. We excluded any acquisition for which the reference photodiode signal varied by more than 2% between the upwelling and down-welling irradiance measurements. Below 2%, spectra were rescaled using the reference photodiode signal assuming that changes in incident light were equivalent over the entire wavelength range. After all these steps, albedo was calculated as the ratio of upward to downward irradiance. Each upward and downward spectrum is the result of the averaging of a set 10 spectra. Albedo spectra were finally smoothed using a low-pass filter. For each measurement site, it was checked that all spectra correctly overlapped before being averaged. For the 2015 dataset, the average standard de-viation of all integrated albedos (over the 400-1000 nm wavelength range) measured at each snowpit is 0.3% with a maximum of 1%. Thus, in most cases, it is reasonable to assume that the precision on albedo measurements is below 1%.

Radiative transfer modeling

Albedo numerical simulations were performed using the Two-stream Analytical Radiative Trans-fEr In Snow (TARTES) model (Libois et al., 2013). Briefly, TARTES uses the delta Edington ap-proximation (Jimenez-Aquinoa and Varela, 2005) in a layered plane parallel snowpack. Each layer is characterized by an average SSA and density. TARTES solves the radiative transfer equation at all depths. For our analysis, only albedo will be presented. Calculations were made using the ice refractive index presented by Picard et al. (2016a). Issues regarding the occurrence of impurities in snow are not addressed in this study since we focus on results at wavelengths (NIR) where impu-rity effects can be neglected (Warren and Wiscombe, 1980) in comparison with snow SSA effects (Bohren and Barkstrom, 1974). The underlying sea ice is not modeled, only its albedo (measured on the field) is specified at the bottom of the snowpacks. Albedo depends on solar zenith angle and cloud cover, but a fully diffuse radiation is equivalent to a direct radiation with a SZA of ∼50◦ (Warren, 1982). In our case, SZAs are between 47◦ and 57◦, therefore simulations were performed considering a diffuse radiation (SZA of 53◦ in TARTES). Doing so, the maximal error on albedo is ∼0.01 at 1000 nm. The use of TARTES allows the calculation of albedo on a wide wavelength range which makes possible the assessment of broadband albedo and total energy absorbed by the sea ice-ocean system A, in W . Both were calculated as follows : R 3000 αs(λ)I(λ)dλ α = 300 3000 I(λ)dλ (2.1).
A = Z R 300 3000 300 (1 − αs(λ))I(λ)dλ (2.2).
where αs is the spectral albedo calculated with TARTES over the 300-3000 nm wavelength range and I(λ) is the spectral solar irradiance in W m−2 s−1. The solar irradiance spectra was calculated with SBDART, it is representative of solar irradiance observed in Qikiqtarjuaq on June 1st at 12 :00 under typical atmospheric conditions of Arctic spring on snow covered areas. The date of June 1st was chosen as the median of albedo measurements dates. The corresponding total wavelength integrated irradiance for this date is 784 W m−2 and it increased from 740 to 800 W m−2 along the sample period mainly through the decrease of the solar zenithal angle (from 47.79◦ to 43.66◦). Only one solar spectrum was used since the aim of the study was not to investigate absolute radiation and energy budget, but rather broadland albedo which only depends on the spectra variations of the radiation, not the absolute value.

General evolution and meteorological conditions

Surface conditions changed drastically during both sampling campaigns as depicted in Fi-gure 2.2. From the first day of surface melting, it took approximately one month for the snowpack to melt entirely. As previously observed by Perovich et al. (2002) and by Nicolaus et al. (2010), as the melting season progressed sea ice surface became darker and spatial variability increased. The time evolution of albedo at 500 nm and 1000 nm are presented in Figure 2.3 and, similarly to Perovich et al. (2002) and Nicolaus et al. (2010), this evolution clearly shows four main stages confirming visual observations in the field. These phases are defined below.
Phase I : Cold, dry snow (from the first sampling day on May 13 to 24 in 2015). Sea ice was covered by a dry winter snowpack that had not experienced any melting event. Air temperature increased during this phase but remained below 0◦C (Figure 2.4). A significant snowfall event associated with strong winds occurred before the first sampling day in 2015 (May 8 and 9), building a fresh snow layer at least 10 cm-thick. Temperature in snow was first colder at the surface, or at least at mid-depth, (-6.2◦C) than at the bottom-most layer where temperatures remained fairly steady between -5◦C and -4.5◦C in the day time. The subsequent increase in air temperature reversed the temperature gradient within the snow during this first phase.
Phase II : Surface melting (May 25 to June 11 in 2015 ; from the first sampling day on May 19 to June 9 in 2016). This phase started with the first surface melting event which coincided with the first positive air temperature in 2015 (Figure 2.4). Coarse rounded grains and wet grains appeared and albedo decreased in the infrared (Figure 2.3). Air temperature fluctuated around 0◦C and several snowfalls were observed both years (Figure 2.4, snowfalls specified only for 2015). Moreover, the weather was cloudier than during the previous phase and heavy fogs were more common in the early morning. These meteorological conditions persisted in the next phases. Overall, snow temperatures gradually increased until the 0◦C isothermal state was reached.
Phase III : Ripe snowpack (June 12 to the last sampling day on June 16 in 2015 ; June 10 to June 17 in 2016). At this stage, the snowpack was at the melting temperature and comprised entirely of rounded polycrystals. This phase is characterized by a decrease in albedo over the visible range for the first time of the season (Figure 2.3 and 2.5). Snowpack thickness decreased very quickly until melt-out (4 days in 2015, 7 days in 2016) .
Phase IV : Melt pond formation (June 18 to the last sampling day on June 26, 2016). Snowpacks gave way to a mixture of bare ice and melt ponds. The transition between snow cover and bare ice was progressive, because the ice surface was granular and looked similar to the large wet grains observed in the ultimate stages of snow melt. As previously observed, sea ice was first rapidly flooded by extended shallow ponds before they drained and got their final shape. During our last sampling day in 2016, June 25, a cooling event associated with snowfall temporally froze the ponds (Figure 2.2F) and increased albedo (Figure 2.3).

READ  Are Environmental Transitions More Prone to Biological Invasions?

Snow stratigraphy and physical properties

Only physical properties sampled in 2015 are presented here because they cover the main first three phases, unlike in 2016.
Phase I : Cold, dry snow. The observation of 15 snowpits during this phase revealed a dominant stratigraphy composed of three or four main layers. Snow grain types (and main layers) are presented in Figure 2.5, vertical profiles of SSA and density are presented in Figure 2.6 with average values in Table 2.1. The bottom-most layer (layer I), in contact with the underlying sea ice, was indurated depth hoar formed from a wind slab (Domine et al., 2016; Sturm et al., 2008), as evidenced by the presence of depth hoar crystal embedded in a matrix of small rounded grains, and confirmed by its high density of 372 pm 51 kg m−3. Its SSA was 8.9 ± 4.4, typical of depth hoar, whether indurated or not (Domine et al., 2016). Generally the upper part of the snowpack consisted of a layer of indurated faceted grains (layer II) with average SSA of 12.1 ± 1.8 m2 kg−1 and average density of 409 ± 40 kg m−3, topped by a wind slab layer (layer III) made of rounded grains characterized by significantly higher SSA values, 33.4 ± 2.6 m2 kg−1 and lower density 276 ± 38 kg m−3. Occasionally a layer of dentritic crystals or fragmented particles could be observed at the surface (layer IVa). The highest SSA values were recorded in this layer, 49.3 ± 5.9 m2 kg−1 on average (see dark red areas at the surface in Figure 2.6a). Moreover, sublimation crystals (Gallet et al., 2014a) sometimes formed at the surface of the snowpacks. Figure 2.6 also shows a significant dichotomy in both profiles with layers I and II characterized by lower SSA and higher density than layer III. Moreover SSA in layer III gradually decreased over time. Overall, snow depth ranged from 15 cm to 54 cm. Snow dunes were studied on May 19, 22, 23, 29 and June 4. They corresponded to deeper snowpacks, and were composed of layers I and II only. Furthermore, layer II could be divided into two distinct layers of indurated faceted crystals which showed highest densities values, up to 500 kg m−3, topped by a wind slab. All this information is shown on vertical profiles in Figure 2.5. Smaller features like sastrugi (Figure 2.2a) and barchan dunes were currently observed along the sea ice before melt onset. Freeboard was always positive during phase I.
Phase II : Surface melting. First melting was observed on May 26, one centimeter below the surface and coincided with a low SSA layer at that depth (see Figure 2.6). Overall, surface mel-ting was characterized by the formation of a layer of rounded polycrystals (layer Va) of low SSA (10.6 ± 4.1 m2 kg−1 ). Additionally, as melting conditions persisted this layer got thicker and its SSA kept on decreasing to a minimum of 2.6 m2 kg−1 on June 13 (phase III). The alternation of negative and positive temperature during night and daytime subjected the surface of the snowpack to a diurnal cycle. During daytime, at the surface, bonds between snow grains melted leading to the observation of wet clustered rounded grain which partially (at least near the surface) froze during the following night forming again dry rounded polycrystals and often a thin layer of melt-freeze crust at the surface. Several snowfalls deposited a new fresh snow layer covering layer Va (Figure 2.5), which then quickly metamorphised. Fresh snow tended to accumulate in depressions instead of on top of dunes. Melting and subsequent refreezing increased cohesion between snow grains which totally stopped erosion of snow by wind and therefore its transportation. As the weather became cloudier, a thin layer of surface hoar or needle crystals deposited during the night were regularly observed at the beginning of the day before it rapidly melted. The underlying snow layers I and II, unaffected by surface melting, remained nearly unchanged (with SSAs of 10.6 ± 4.1 m2 kg−1 and 13.8 ± 6.9 m2 kg−1 and densities of 370 ± 26 kg m−3 418 ± 51 kg m−3 for layers I and II respectively). SSA of layer III (24.7 ± 4.3 m2 kg−1 kept on decreasing during phase II (Figure 2.6) until it had completely transformed into wet grains (phase III). Ice layers within the snowpack were first observed on May 29 and became more and more common, to the point that they were present everywhere at the end of phase II and several of them could be found in the same snow column. Two main processes of formation were observed : first, melt-freeze crust formed from the melted surface layers that were buried under new snow and then consequently froze within the snowpack.

Table of contents :

1 Introduction g´en´erale 
1.1 L’oc´ean Arctique et sa banquise
1.2 Le manteau neigeux, ´el´ement essentiel des r´egions polaires
1.3 Banquise arctique en mutation
1.4 Objectifs et organisation de la th`ese
2 Metamorphism of Arctic marine snow during the melt season. Impact on albedo 
2.1 R´esum´e
2.2 Abstract
2.3 Introduction
2.4 Materials and methods
2.4.1 Study area
2.4.2 Albedo measurements
2.4.3 Snow physical properties
2.4.4 Sampling Protocol
2.4.5 Data processing
2.4.6 Radiative transfer modeling
2.5 Results
2.5.1 General evolution and meteorological conditions
2.5.2 Snow stratigraphy and physical properties
2.5.3 Spectral Albedo
2.5.4 Albedo Modeling
2.6 Discussion
2.6.1 Snowpack formation
2.6.2 Albedo and surface evolution
2.6.3 Albedo modeling, limitations and suggestions
2.7 Conclusion
3 Impacts of snow impurities on sea ice transmittance 
3.1 Abstract
3.2 R´esum´e
3.3 Introduction
3.4 Materials and methods
3.4.1 Study area
3.4.2 Stratigraphy and physical properties of snow
3.4.3 Optical measurements
3.4.4 Transmittance measurements
3.4.5 SOLEXS measurements
3.4.6 Albedo measurements
3.4.7 Radiative transfer modeling
3.5 Results
3.5.1 Optimization results
3.5.2 Sea ice IOPs and transmittance simulations
3.5.3 Impacts of snow impurities on sea ice transmittance
3.6 Discussion
3.6.1 Uncertainties about sea ice IOPs and optical measurements
3.6.2 B and impurity content estimates
3.6.3 General implications for radiative transfer of sea ice
3.7 Conclusion
4 Augmentation de l’´eclairement dans la banquise li´ee au m´etamorphisme de la neige, et effets sur les algues de glace 
4.1 Abstract
4.2 R´esum´e
4.3 Introduction
4.4 M´ethodes
4.4.1 Pr´esentation du jeu de donn´ees
4.4.2 Mod´elisation du transfert radiatif
4.5 R´esultats et discussion
4.6 Conclusion et perspectives
5 Conclusion g´en´erale 
5.1 Synth`ese des r´esultats
5.2 Perspectives
Liste des figures
Liste des tableaux
R´ef´erences bibliographiques


Related Posts