Rate of Evolution of the Specific Surface Area of Surface Snow Layers

Dec 24, 2002 - and the French Alps, using CH4 adsorption at liquid nitrogen temperature (77 K). The SSA of fresh snow layers was found to decrease wit...
0 downloads 0 Views 127KB Size
Research Rate of Evolution of the Specific Surface Area of Surface Snow Layers A X E L C A B A N E S , L O ¨I C L E G A G N E U X , A N D F L O R E N T D O M I N EÄ * CNRS, Laboratoire de Glaciologie et Ge´ophysique de l’Environnement, B.P. 96, 54 Rue Molie`re, 38402 Saint Martin d’He`res, Cedex, France

The snowpack can impact atmospheric chemistry by exchanging adsorbed or dissolved gases with the atmosphere. Modeling this impact requires the knowledge of the specific surface area (SSA) of snow and its variations with time. We have therefore measured the evolution of the SSA of eight recent surface snow layers in the Arctic and the French Alps, using CH4 adsorption at liquid nitrogen temperature (77 K). The SSA of fresh snow layers was found to decrease with time, from initial values in the range 613-1540 cm2/g to values as low as 257 cm2/g after 6 days. This is explained by snow metamorphism, which causes modifications in crystal shapes, here essentially crystal rounding and the disappearance of microstructures. A parametrization of the rate of SSA decrease is proposed. We fit the SSA decrease to an exponential law and find that the time constant Rexp (day-1) depends on temperature according to Rexp ) 76.6 exp (-1708/T), with T in kelvin. Our parametrization predicts that the SSA of a snow layer evolving at - 40 °C will decrease by a factor of 2 after 14 days, while a similar decrease at -1 °C will only require 5 days. Wind was found to increase the rate of SSA decrease, but insufficient data did not allow a parametrization of this effect.

1. Introduction For over a decade, numerous studies of atmospheric chemistry in polar regions have revealed that the snowpack could interact chemically with the atmosphere. Spring-time Arctic ozone depletion (1) is among the most studied aspects. Recently, studies have shown that significant exchanges of reactive gases between the snowpack and the lower troposphere were taking place, increasing the interest of studying air-snow interactions. Production of NOx by the snowpack has been observed in Greenland and Antarctica (2-5). Emission of carbonyl compounds (formaldehyde, acetaldehyde, and/or acetone) from the snowpack have also been observed in Greenland (6, 7), the Canadian Arctic (8-10), and at middle latitudes (11). Emissions of persistent organic pollutants have also been observed and modeled (12). Numerous models of atmospheric chemistry over snowcovered surfaces (13, 14) lead to calculated concentrations lower than measured ones (8, 15), which demonstrates that the impact of the snowpack must be included to yield reliable * Corresponding author phone: (33) 476 82 42 69; e-mail: [email protected]. 10.1021/es025880r CCC: $25.00 Published on Web 12/24/2002

 2003 American Chemical Society

atmospheric composition. However, the inclusion of this impact requires the knowledge of the physical processes involved, but these are insufficiently understood (16). Possible processes include sublimation/condensation of snow and solutes, adsorption/desorption of trace gases from the snow crystals surfaces, diffusion of species in the ice lattice, and heterogeneous reactions on crystals surfaces (17). Many physical parameters such as the specific surface area (SSA) of the snow must be known to quantify processes involving the surface of snow crystals. SSA is the surface area accessible to gases for a given mass of snow and is expressed in cm2/g rather than m2/g because snow SSA values are usually low (18-21). Fresh snow layers have the highest interaction potential with the atmosphere. Fresh crystals are at the surface of the snowpack and have the highest potential for physical evolution, and their SSA decrease can be rapid (19, 21). The rate of SSA decrease is an essential parameter to quantify the amounts of a given species released by desorption, as mentioned by Hutterli et al. (6) in the case of HCHO. However, there are only few data on the rate of decrease of snow SSA shortly after precipitation (19, 21), and quantitative rate parameters have not been extracted from those data. We have therefore studied the SSA evolution of eight fresh snow layers, sampled in the French Alps and in the Arctic. Two of these layers were thick enough to be divided into sublayers, so that 11 evolution rates were obtained. These rates seemed to be affected by temperature, wind speed, and other environmental parameters, whose values are also reported. We then hope that this first, and necessarily incomplete, set of data will assist modelers of snow-air interactions in estimating the rate of SSA decrease of fresh snow.

2. Snow Sampling Snow layers studied in this work were sampled in the French Alps in early 1999 and 2001, hereafter referred to as winters 1999 and 2001, in the Canadian Arctic during winter and spring 2000, and in Svalbard during spring 2001. In the French Alps, snow samples were collected at three sites. Two of them were located at Col de Porte (45°12′ N, 5°44′ E) at an elevation of 1330 m, in the Chartreuse range, about 10 km north of Grenoble. The first sampling site was located within the meteorological station of Centre d′Etude de la Neige (CEN) and is referred to as site P. Continuous monitoring of air temperature and humidity, surface snow temperature, snowpack height, liquid and solid precipitation rates, and wind speed at 10 m height are performed by CEN. Air temperature was recorded in a ventilated shelter 1.5 m above the snow surface. Snow surface temperature was measured by its IR emission (22). The second site was located 500 m west of CEN in a flat forest clearing of about 3000 m2. This site, referred to as site P′, was preferred to site P when chemical measurements were coupled to our physical measurements because it is further from the road. Meteorological parameters were assumed to be similar to those at site P. This was verified for air temperature, that was found to be within 0.5 °C of the CEN values on several occasions. The third site was located at Col du Lautaret (45°02′ N, 6°24′ E), 55 km east of Grenoble, in a small south-facing sheltered basin at an elevation of 2058 m. This higher elevation site, referred to as site L, had to be used because of the unusually warm temperatures during winter 2001, that VOL. 37, NO. 4, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

661

TABLE 1. Snow Layers Studied in the French Alps and the Arctic no.

date of snowfall

1

Feb 6-9, 1999

2

Mar 4, 1999

3

Feb 9, 2001

4

Feb 3 , 2000

5

Feb 7 , 2000

6

Apr 13-14, 2000

7

Apr 25-28, 2000

8

Apr 27, 2001 a

period of study

sampling site

Feb 10, 1999 Feb 16, 1999 Mar 4, 1999 Mar 9, 1999 Feb 9, 2001 Feb 15, 2001 Feb 4, 2000 Feb 22, 2000 Feb 7, 2000 Feb 22, 2000 Apr 14, 2000 Apr 23, 2000 Apr 25, 2000 Apr 29, 2000 Apr 27, 2001 May 3, 2001

Col de Porte site P Col de Porte site P′ Col du Lautaret site L Alert site A Alert site A Alert site A Alert site A Ny-Ålesund site N

Density measured at the bottom of the layer.

b

initial layer thickness (cm)

fresh snow crystal shape

-11

0.21a

plates, needles and columns

38

-2

0.10

plates, needles and columns

24

-0.5

0.20a

dendritic crystals

3 to 8

-40

0.16

columns, bullets combination

1

-40

0.08

columns, bullets combination

0.3

-27

0.07

columns, bullets combination

5

-15

0.016 to 0.21b

variably rimed dendritic crystals

0.14

columns, rimed dendritic crystals

6

-1.5

In wind-packed accumulation.

3. Experimental Method The SSA of each snow sample was determined by CH4 adsorption at liquid nitrogen temperature (77.15 K), as detailed earlier (21, 22). CH4 was used rather than N2 because of its lower saturating vapor pressure at 77 K, which allows a much better precision (21). Briefly, we used a volumetric method which requires the measurement of the adsorption isotherm of CH4 on snow, followed by a BET treatment (23). This yields the SSA of the snow sample and the net heat of adsorption of CH4 on ice, ∆QCH4. This latter parameter was used to test the reliability of SSA measurements, as a value ∆QCH4 ) 2240 ( 200 J/mol must be obtained for the measurement to be considered reliable (21). The reproducibility of the method was within 6%, and its accuracy, taking into account systematic errors due to the BET treatment, was within 12% (21).

4. Results The characteristics of the eight snow layers studied are summed up in Table 1. 4.1. Description of the Snow Layer Fallen on February 6-9, 1999 at Col de Porte. This snowfall ( no. 1 in Table 1) 9

initial snow density

105

resulted in rare snow events at site P. Unfortunately, there is no meteorological station at or near site L. In the Canadian Arctic, snow samples were collected near the Alert base (Ellesmere Island 82°30′ N, 62°21′ W) during the ALERT 2000 campaign. The sampling site, referred to as site A, was located 5.4 km south of the base at about 180 m above sea level (18). In Svalbard, the snow was sampled at Ny-Ålesund (78°55′ N, 12°56′ E) during the NICE campaign in April-May 2001. The sampling site N was located about 50 m south of the Italian station, at the edge of the Ny-Ålesund village. The sampling method has been described in detail earlier (19, 21-22). Briefly, vertical faces were dug to observe the stratigraphy and to locate the snow layer of interest. For each sample, about 150 cm3 of snow was collected in two glass vials using a stainless steel spatula thermally equilibrated with the snow. Great care was taken to pertub the snow as little as possible during sampling (21). The vials were immediately dropped in liquid nitrogen to stop metamorphism and remained immersed until being transferred into the volume used to measure SSA. Snow and air temperatures were measured at different heights with a mercury or alcohol thermometer, or a thermocouple. According to the thickness of the layer, snow was sampled at up to three levels, and density was measured using a sampler of known volume.

662

mean air temp

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 4, 2003

FIGURE 1. Air temperature (1.5 m above the surface) and snow surface temperature (measured by IR emission) in early February 1999 at Col de Porte (site P). Data from CEN, Me´ te´ o France. was studied at Col de Porte at site P. It fell in several episodes in 4 days until reaching a thickness of 105 cm on February 10 and covered a hard melt-freeze layer. According to CEN precipitation and snowpack height recordings, there were six major precipitation events from February 6 to February 9, three on February 6, one on February 7, one on February 8-9, and the last one on February 9. Sampling was done at several heights on February 10, 12, and 16, but a wind event that started late on February 10 at 4 p.m. (GMT) and lasted until the morning of February 12 perturbed the layer so that only the bottom sampling, at a height of 10-15 cm above the underlying layer, is reported here. This snow was made up of a mixture of plates, needles, and columns with little or no rime. On February 10, it was sampled at 12:00 GMT, while the air temperarture was -9.4 °C. The temperature of the sampled snow was -2.7 °C, its density was 0.21, and its SSA was 414 cm2/g. From CEN precipitation and snowpack data, and from our density measurement, we estimated that the snow sampled was 90 h old. The temperature started dropping on February 12 (Figure 1), while wind speed became low (0.95 in 3 cases out of 11 with the linear equation, in 5 cases for the exponential equation, and in 8 cases with the logarithmic equation. The average R2 values for eqns 1-3 are 0.873, 0.911, and 0.936, respectively. So the best fits appear to be obtained with the logarithmic equation, but the differences in R2 with the other equations are small and may not be significant. Considering the empirical character of our approach, it is clear that there is no strong justification to actually prefer one equation over another one at this stage. Indeed, the physical reason why eq 3 works best is unclear. Other physical transformations of solids, such as creep (27), follow logarithmic laws, but no physical explanation has been given for these processes either. For model parametrization, it is desirable to relate R values to environmental variables. Wind clearly has an effect, as the fastest decreases were observed for the April 25, 2000 layer at Alert, which was exposed to strong winds (up to 11 m/s). However, most other layers were subjected to low or negligible windspeeds (