Spatial distribution of snow chemical load at the tundra

Biogmchcmiarv o/Scuronolly Snow-Covered Cafchtne~s
(Proceedings oTa Boulder Symposium.
July 1995). WtS Publ. no. 228. 1995.
Spatial distribution of snow chemical load at the
tundra-taiga transition
J. W. POMEROY & P. MARSH
~VationalHydrology Research Intlirure, Environment Canada.
I I Innovation Blvd., Saskaroon, Sashchewan S7N 3H5,Canada
H. G.JOrnS
1:VRS-eau. Sre-Foy, Quebec GJ V 4 ~ 7 Canada
.
T. D. DAVIES
,
Climatic Research Unit, Universiry of Enrr Anglia. Now'ch MP4 7iV, UK
Abstract The chemical composition of seasonal snow covers was studied
in the taiga-tundra transition zone near Inuvik, Northwest Territories,
Canada. Concentrations of the major ions, snow water equivalent, and
winter leaf area index were determined for a series of forested, shrubtundra and open tun&rasites along a transect that spans the arctic Weline.
The substantial variation in snow and ion load with leaf area index,
landscape type, and mesoscale site demonstrates that both the local land
surface factors and the broad-scale influences which control snow and ion
deposition, must be addressed in order to spatially extrapolate
measurements of snow quantity and snow chemistry. Ion loads vary by
up to 5-fold in different landscape types within a mesoscale site and up
to 18-foldbetween mesoscale sites. Two factors, operating at two scales,
most suongly affect the load of snow and major geochemical ions in
snow at the arctic treeline. The first, at a small-scale, is the landscape
roughness as parameterized by the leaf area index or by topographic
slope. The second, mesoscale wind exposure and relocation of snow, can
strongly affect the small-scale landscape-snow relationship. This makes
determination of snow accumulation and chemistry from point
characteristics extremely difficult. A combination of two-dimensional,
physically based models of wind transport of snow and snow chemical
loads, operating in a distributed fashion over the mesoscale, must be
developed to predict snow accumulation and snow chemistry in complex,
windswept environments such as the tundra-wigs transition.
The geochemistry of subarctic and arctic catchments is strongly influenced by the
chemistry of snow cover. This is due to the long winter over which chemical species
accumulate in snow and the rapid release of the chemical load from the pack upon melt.
the delivery of major ions to these catchments is not uniform, as several atmospheric
and land surface processes enhance or deplete concentrations in snow and the
":amulation of snow (Woo & Marsh, 1978; Delmas & Jones, 1987; Bamie, 1991;
192
J. W. Pomeroy cr al.
Jones er al., 1993). Pomeroy et al. (1993)showed that wind redistribution of snow tien:
smooth to rough surfaces and preferential dry deposition to forested terrain were tiic
primary processes controlling the distribution of major ions in the late winter snow of
the subarctic-arctic transition zone. The results of Benson's (1982) blowinz-snow
aansport study and the sublimation model of Tabler (1975)suggest that from 50 to 75 R
of annual snowfall is sublimated from open tundra. In addition, Pomeroy & Schmidt
(1993)have shown that sublimationof intercepted snow from coniferous boreal forests
can remove 32% of annual snowfall. Thus both dry and wet deposition and the
redistribution of snow are sensitive to aerodynamic roughnecs. It follows that schemec
Parsons Lake
Fig. 1 Study site in northwestern Canada. l l ~ etransecr extended nonhward frcm
Tundra Lakes to Whitefish Pingo. The treeline is crossed by this transect benvetn
Havikpak Creek and Trail Valley Creek.
Spatial distribution of rhe snow chemical load at the tundra-taiga transition
193
to stratify ion load and concentration by landscape class (e.g. vegetation, terrain
roughness, hydrography) might use techniques developed by Steppuhn & Dyck (1974)
for snow accumulation.
Pomeroy et al. (1994) and Marsh et al. (1995) demonstrated the variability of
snow accumulation in catchments in a subarctic region. Marsh et al. (1995) found that
a small sheltered catchment (Siksik Creek, 0.83 krn2) received a blowing-snow input
equivatent to 25 % of winter snowfall, which more than compensated for its 16% snow
loss due to blowing-snow sublimation. Pomeroy et al. (1994) showed that an adjacent
larger catchment with terrain more representative of the regional average (Trail Valley
Creek. 63 m2), showed no net gain from blowing-snow inputs but rather a loss
equivalent to 23 % of winter snowfall due to blowing-snow sublimation. If we presume
that the chemical concentrations in freshly deposited snow were equal in both basins,
the small catchment would have had a 40% greater winter ion load (per unit area) than
the larger catchment.
Subarctic snowmelt occurs in late May or early June. The ion load released to
streams and soils is the result of 8 to 9 months of wet and dry deposition and various
concomitant transformation processes. Historical observations suggest that the annual
solute delivery to lakes and streams is dominated by snowmelt in the high Arctic (de
March, 1975), whiist in the low Arctic it is most important in years with low rainfall
(Welch & Legault, 1986). Climate change models predict relatively large temperature
increases (24°C) for high latitudes and also anticipate changes in regional s t o m tracks
and cyclonic activity (Goodess & Palutikof, 1992). Because of this warming, biophysical
models predict a northward advance of the boreal forest into the Subarctic and low
Arctic (Monserud er al., 1993; Smith et al., 1992). However, the productivity of boreal
forests is limited by availability of nitrogen. This limitation extends over much of the
northern boreal forest (Boring et al., 1988; Hudson et al., 1994) and is amplified by
further growth restrictions due to anthropogenic sulfur deposition at certain treelines
(Bucher, 1987). A physically-correct understanding of the processes of snow accumulation in northern landscapes is necessary to predict the delivery of snow and major ions
to this globally-significant ecotone. It is therefore important to document the (1) spatial
patterns of ion load at mesoscales, and (2) relationships to vegetation and terrain which
may be used to distribute the ion load using available landscape classifications.
This paper examines the small-scale (10-100 m) and mesoscale (100 m-10 Ian)
variability of snow water equivalent, snow chemical concentration, and snow chemical
load in the tundra-tai,oa transition zone (200 km north-south) in order to identify the
processes and landscape properties that need to be modelled in order to predict the
spatial distribution of winter ion accumulation in treeline catchments.
Study site
The study encompasses a 200-km north-south transect spanning the treeline,
2pproxirnately 50 km east of, and parallel to, the Mackenzie Delta, Northwest
Territories, Canada (Fig.I). The transect reaches southward from the Arctic Ocean, its
Xrthern point being an arctic tundra coastal plain near a prominent pingo (ice-cored
'
.
I. W. Pomeroy er al.
194
Table 1 Mesoscale sites at which landscape-stratifiedsnow and snow chemistry surveys were conducted.
near Inuvk. N.W.T.
Mesoscale site
Distance from Dominant
coast (km)
vegetation
Terrain
(N)
Longitude
(w)
67'37'49'
133'45'36"
198
Spruce, taiga and
forest
Low relief muskeg
and lakes
Havikpak Creek 68'19'12'
133'30'59'
115
Taiga and shrubtundra
Rolling hills and
valleys
Trail Valley
Creek
68'44'32"
133'29'60'
84
Shrub and open
tundra
Plateau, hills. and
incised valleys
ParsonsLake
69'00'13"
133'35'59'
42
Open tundra
Undulating plain
and lakes
Open tundra
Coastal plain with
isolated hills
Tundra Lakes
Latitude
Whitefish Pingo 6g023'30'
133'32'1 1'
0.25
-
hill) on the Tuktoyaktuk Peninsula. The line passes just east of Inuvik, through a
transitional forest-tundra and finishes north of Arctic Red River, in a boreal forest.
Long-term snowfall records suggest that snowfall declines sharply with Iatitude in this
region, as the 30-year mean annual snowfalls for Inuvik and Tuktoyaktuk are 177 and
86 mm, respectively. These values may underestimate true snowfall (Woo et al., 1983)
by up to three-fold, especially in the tundra region, and are not likely to represent snow
accumulation because of r'edistribution processes such as blowing snow in open areas
and interception of snow in coniferous forests. Snow covers along the transect are
uniformly cold and dry during the winter, mean daily high temperatures are below
-5°C from October through April and Iows below -30°C for 4 mo. Mean winter wind
speeds and hence the frequency of blowing snow are much higher at Tuktoyaktuk
(6 m s-') than at Inuvik (2.6 m s"). Straddling the transect are two research basins, Trail
Valley Creek and Havlkpak Creek. These basins lie on either side of the treeline and are
dominated by tundra and black spruce forest, respectively. Along the transect.
landscape-stratified snow surveys and snow chemistry measurements were taken at five
"mesoscale sites" which include the two basins. The mesoscale sites are described in
Table 1.
Field measurements
Measurements were conducted in late April and early May 1993, severaI weeks before
meIt. All measurements were made on landscape-stratified snow courses, established in
six prominent landscape classes:
(1) open tundra - short grass or lichen tundra with vegetation Iess than 30 cm tall;
(2) shrub tundra - alder or willow bush tundra with vegetation from 30 cm to 3 m tall;
(3) taiga - open canopy spruce forest of stunted trees, 3-8 m tall;
(4) forest - closed canopy spruce forest of taIl trees, 5-15 m tall;
(5) valleys - sheltered valley bottoms of open or shrub vegetation;
(6) drifts - hillsides with slope gradients greater than 9% (8") and an open upwind
fetch.
Spatial distribution of the snow chemical loud ar the tundra-taiga transition
195
Care was taken to center snow courses in each landscape class in order to minimize,edge
effects and to include the characteristic variation of terrain and vegetation in each
landscape over the course length. Snow courses were located at each mesoscale site
where an example of a particular landscape class could be found.
Leaf Area Index (LAT) measurements were made during snow surveys and in June
1994, before deciduous leaves formed and after all snow had melted from the sites. At
least 10 LA1 measurements were made along each snow course, with some repeat series
of measurements to eliminate any effects of variable sky conditions. LA1 was measured
with a LICOR 2000 Plant Canopy Analyzer which measures scattered light intensity in
5 cones through the vegetation canopy and compares these intensities to a background
unobscured-sky reference intensity (Welles & Norman, 1991). From the resulting
vesetation transmissivity values, the extinction cefficient of vegetation is calculated and
then presuming a random leaf distribution, the LAI. For this late winter case, LAI is
defined as the cumulative horizontal area of stems, needles, and branches per unit area
of ground and is dimensionless.
Snow water equivalent was sampled using snow depth rods and an ESC-30 density
sampler along snow courses established within each landscape class. Snow depth for
each course was measured at 25 points, spaced 5 m apart and density at five points
selected from the 25 depth points.
Snow for chemical analysis was colIected from three pits in each course seIected
amongst points for density measurements. All snow was cold and dry and had not
undergone melt. Snow was removed uniformly from each layer using a Teflon scoop and
placed in 1-1 HDPE bags, one bag corresponding to one pit.
Laboratory measurements
Snow was kept frozen until required, rapidIy melted, and filtered through 0.2-pm
nucleopore filters. The filtrate was kept at a temperature of 4°C and shipped to
Saskatoon for ion chromatograph and AAS analysis. Anion analyses were performed
using a Dionex 2010i ion chromatograph with a 100-pl injection loop, 0.75 m M
NaHC03/1 .5 mM Na2C0, eluent and suppressed conductivity detection. Samples were
introduced by Technicon sampler into the loop after a 9 to 1 sample to eluent dilution.
A Dionex 4270 Integrator measured peak area and a Linear 100 recorder measured peak
height of Dionex concentration traces. Numerous blanks on scoops, bottles, filters, and
analytical equipment indicated no measurable contamination of samples. The level of
arecision and limits of detection for the ion chromatograph are normally at least an order
of magnitude less than the levels measured.
Data analysis
LA1 measurements were averaged for each snow course and then averaged for each
landscape class (except drift, which was defined by topography) across the uansect.
Snow water equivalent was calculated by assigning densities derived from the ESC-30
neasurement to the nearest (or most similar in the case of a vegetation gradient) snow
depth and multiplying. Averages and coefficients of variation were calculated frcm the
25 derived snow water equivalent values in each course. Ion Ioad was calcuIatcd by
.
J. W. Pomeroy
196
et al.
mulriplyin~bulk ion concentration by the snow water equivalent. The range of ion ~ 5 3 .
centrations was quite low in each course, therefore ion loads for each course could L??
confidently determined by multiplying mean concentrations from the three pits by the
mean snow water equivalent as determined by the five densities assigned to the adjacent
25 deprhs for each course. Mean concentrations for samples from the mesoscale sites
were not weighted by the areal extent of the landscape type where samples were collected. However the same number of samples were collected from each landscape type at
each mesoscale site. Thus, although the site sample means are not true "regional means"
of concentrations for each mesoscale site, they do provide local ranges of values which
are comparable to each other because of the consistent manner of collection. The stronn
similarity of values within each mesoscale location suggests that ion concentrations chr:
be lumped in this manner to calculate means without prejudicing the analysis.
Concentrations along the latitudinal transect
Mean values and coefficients of variation (CV) for ion concentrations (Cl', SO^^-. NO3?
collected from all landscape types are shown for each mesoscale site along the transect
in Table 2. The highest concentrations and CVs were found in the more northerly and
windswept locations. The increase in concentration with latitude is extreme in the case
of a sea-salt derived ion such as C1-, which increases 8-fold in concentration from
and NO; which increase
Tundra Lakes to Whitefish Pingo but still notabIe for so4*2.4- and 1.7-fold, respectively. Two significant environmental transitions occur as one
travels northward along this latitudinal gradient; i.e. increasing proximity to sources of
sea spray in the Arctic Ocean and increasing frequency of blowing snow events as the
treeline is crossed. The effect of these transitions on snow chemistry is not uniform for
all chemical species. The input of sea spray to precipitation and to dry deposition
primarily enhances C1' concentrations whilst blowing-snow phenomena enhance all
concentrations to some degree. Pomeroy et al. (1991) reported similar evidence of
preferential scavenging of sea-salt ions by blowing-snow particles and a general ion
concentration increase as the blowing-snow particles sublimated. Their study showed
that preferential scavenging during transport approximately doubled the concentration
of sea-salts in wind-blown snow (in addition to subIimation-derived concentration
changes) after one afternoon's blizzard. The concentrations of so4'' increased 20%, in
Table 2 Concentrations in peq 1'' and coefficients of variation (in parentheses) of chrec major ions in
snow. Inuvik regioa. spring 1993.
Mesoscale site
Cl-
Tundra Lakes
6.2
10.5
17.2
16.6
47.1
Havikpak Cnek
Trail Valley Creek
Parsons Lake
Whitefish Pingo
NO,'
(0.11)
(0.24)
(0.12)
(0.26)
(0.21)
2.6
2.6
3.4
2.7
4.5
SO-:
(0.05)
(0.10)
(0.06)
(0.14)
(0.19)
.
4.5
5.1
6.6
6.7
10.8
(0.09)
(0.15)
(0.12)
(0.14)
(0.19)
Spatial dirrribution of the snow chemical load at the tundra-taiga transition
197
proportion to the sublimation loss from blowing snow but the NO,' increased somewhat
less than proportionately (Pomeroy er al.. 1991).
The blowing snow model of Porneroy & Gray (1994) was applied using
meteorological data collected at the Tuktoyaktuk weather station (wind speed, air
temperature, humidity, snow depth, snow fall) to estimate the annual blowing snow
sublimation loss over the winter of 1992-1993. The annual snowfall and annual
sublimation may be used to calculate enhancement of ion concentrations in wind-blown
snow if the ions are presumed to be "conservative" with respect to sublimation, i.e. they
do not volatize. This enhancement effect, Ef,is defined as
where L, is the annual load of ion (x) in snow, L, is the annual snowfall (mmsnow water
equivalent) and Lb is the annual blowing snow sublimation loss (rnm snow water
equivalent). Pomeroy & Gray's (1994) model predicts that for tundra with a fetch of 3
km near Tuktoyaktuk in the winter of 1992-1993, 65% of annual snowfall sublimates
as blowing snow. Using this result and presuming that (I) parent snowfall concentrations
and dry deposition of NO,' and SO,'' are similar across the transect (likely because of
level terrain and the remoteness of the region from any sources); (2) blowing snow
sublimation is negligible in the forest at the southern edge of the transect; and (3) both
species are conservative during sublimation, then blowing snow sublimationshould lead
to an Ef value of 2.2 in concentrations from forest to tundra. Such an increase is quite
consistent with the 1.7- and-2.4 fold increases in NO3- and SO,:'
respectively, in the
transect from Tundra Lakes to Whitefish Pingo. The latitudinal increase in wind speed
and decrease in terrain roughness suggest that the latitudinal increase in SO': and NO3concentrations are almost entirely due to. blowing snow sublimation and that the
latitudinal increase in C1- concentrations are due to a combination of blowing snow
sublimation and enhanced wet and dry deposition to snow because of proximity to the
ocean and sea spray and the high incidence of blowing snow.
Snow ion load and vegetation density
Relationships between snow water equivalent, snow ion load, and LA1 as a vegetation
density index were developed for the three mesoIocations where all landscape classes
were represented. In latitudinal order along the transect these are Tundra Lakes.
Havikpak Creek, and Trail Valley Creek (Fig. 2). The LAI values represent
measurements from all landscape classes except drift areas. Drift areas are excluded
from the vegetation density analysis because the preferential wind deposition is caused
more by topography than by vegetation. Three environmental gradients are of note in
this comparison: wind speed, snow accumulation, and ion concenuation increm
Ia:imdinally for the same landscape type. In all mesoscale locations the regions with low
L.11 accumulated the least snow and the least ions. Interestin_ely. though higher LM
developed greater ion and snow loads. a linear increase in ion or snow loads with LM
-"E
1-
E
-0.6
=
rn
C
0.4
---
b
e
.
-
. '.
..--------
+
_ _ _ _ - _ d - - - - -
2 0.2
c
-0
0.
0
0.2
0.4
0.6
0.8
Leaf Area lndex
1
0
0.2
0.4
0.6
0.8
Leaf Area lndex
1
0
0.2
0.4
0.6
0.8
Leaf Area lndex
1
Fig. 2 Ion load and snow water equivdent as functions of leaf area index for Tundra
Lakes, Havikpak Creek, and Trail Valley Creek. Note increasing scales for ion load
and SWE from top to bottom.
was not consistent and the highest LAI did not necessarily correspond to the highest
water equivalent or ion load. The nonlinear relationshipbetween snow accumulationand
LA1 may explain why Timoney et at. (1992) found no significant correlation to a
proposed linear relationship between snow depth and tree density in the subarctic near
Great Slave Lake, N.W.T. A peak in snow and ion load occurs at an LAI between 0.25
and 0.35, corresponding to transitional vegetation which tends to be deciduous and at
the leading edge of high roughness areas where wind-blown snow is deposited. This
peak is strongest at Trail Valley Creek where the wind speeds are highest. Vegetation
along the leading edge receives wind-blown snow and its chemical constituents from
adjacent low LA1 areas. The snow in the low LAI areas undergoes scouring, relocation,
and sublimation, resulting in low snow retention. The highest LA1 areas are coniferous
forests which do not usually receive blowing snow inputs because they are fringed by
more open taiga and shrub-tundra. The coniferous forests may lose some snow over the
winter due to sublimation of intercepted snow. however snow surveys suggest that this
loss is low for the subarctic.
Spatial distribution of the snow chemical bad at the tundra-taiga transition
199
Open Tundra
80
W
loo
150
Distance from Coast (km)
0
200
+
A Shrub-Tundra
-
loo
0
150
Distancefrom Coast (bn)
0
50
100
150
Distance from Coast (km)
60
-
r3.5
3
B
200
Fomt
8
240
160
120
\\\
2.5
2
en
c 1.5
\
-
u
- 0.5
'-..-.
-
9 0
0
0
50
loo
150
Distance from Coast (bn)
50
100
150
200
.
200
Dlstance from Coast (km)
Fig. 3 Ion load and snow water equivalent as functions of distance from the coasr for
open tundra. shrub-tundra, taiga. forest, and drift landscape classes. Note differing
scales for ion load and SWE for each landscape class.
1Mesoscale rariation in snow ion load
A revizw of Fig. 2 shows that the range of ion loads differs substantially amongst the
Tundra Lakes, H a v i k ~ a kCreek, and Trail Valley Creek mesoscale locations. The load
of ions and snow water equivalent within a landscape type are shown for five
landscape classes as a function of distance from the Arctic Ocean coast in Fig. 3.
Greater distance from the coast leads to lower sea salt inputs, lower wind speed.
warmer temperature, less frequent blowing snow, and increasing domination of land
cover by dense vegetation landscapes (shrub-tundra, taiga, and forest). Snowfall is at
a minimum near the coast, reaches a peak at Trail Valley Creek (80 krn south of the
coast), and declines somewhat south of that location. The nature of snow and ion
redistribution in the landscape result in varying relationships between load and latitude
in different landscape classes.
Chloride shows enhanced concentrations near the coast and in more open regions,
trends which distort the relationship between CV and snow load. Because sea-salts are
readily incorporated in wind-blown snow, ion load declines with increasing distance
from the coast. The differential in ion load amongst landscape classes varies
dramatically with distance from the coast. At Whitefish Pingo the C1 load in drifts is
3 times greater than that on open tundra, but 45 km inland at Parsons Lake it is 15
times greater and at Tundra Lakes 200 km inland it has returned to being only 4 times
greater. Open areas exhibit a steady decline in snow water equivalent (SWE) and more
dramatically (except for NO<), ion load with distance from the coast. Snow load
halves and C1- load drops &fold in the first 50 km, with a slighter decline as distance
increases. The change in SWE is due not primarily to a change in snow depth but in
density; lower wind speeds at lower latitudes result in less frequent blowing snow and
Iess densification during snow relocation. Sublimation of blowing snow causes a
concentration enhancement which is most noticeable near the coast. This concentration
enhancement leads to a load enhancement because in windswept, open areas the
maximum possible snow depth is largely controlled by the height of vegetation. With
a fixed vegetation heizht, regions that experience more blowing snow will deveIop
denser snowpacks and higher concentrations of most ions in these snowpacks, leading
to higher loads of ion and snow. The exception is NO,' which does not appreciably
change in load with mesoscale site, only changing with landscape type. It would
appear that a loss of NO;, proponional to the loss of snow due to sublimation,
reduces the enhancement effect and hence the mesoscale variation in load.
Shrub tundra receives much of the blowing snow from adjacent open areas. Its
peak snow load (and ion load except for C1-) 80 km south of the coast reflects the
excellent snow-trapping abilities of relatively tall shrub vegetation in windy
environments. Although shrub height increases to the south, snowfall declines, the
frequency of blowing snow decreases, and overall snow loads are smaller. Drift areas
show extremely high snow and ion loads. Similar to shrub tundra. the highest snow
loads in drifts are not found near the coast because the flat terrain there provides
limited areas where drifts may form. The size of drifts thus reaches a maximum south
of the coast and then declines with increasing distance. The limited transects for taiga
and forest landscapes (no examples were found less than 80 krn from the coast)
suggest a decline in snowfall south of Trail Valley. Both forest and taiga have
relatively small ion loads.
Spatial disrriburion of the snow chemical load ar the tundra-taiga trunsition
20 1
DISCUSSION
The variations in snow and ion load with LAI, landscape type, and mesoscale site
demonstrate that both the local land surface factors and the broad-scale influences must
be addressed in order to spatially extrapolate measurements of snow quantity and snow
chemistry. This can be conceptualized in a simple mathematical model where the load
of an element (e.g. snow, ion in snow) to a point in a complex landscape is a result of
a two-stage process. In the first stage there are atmospheric fluxes in three dimensions,
not necessarily in steady-states, to a parceI of air directly over the landscape surface.
The vertical turbulent and energy fluxes in this parcel of air are strongly influenced by
the aerodynamic and energy flux characteristics of the underlying surface. The rate of
change in the concentration of the element in this parcel of air, aXJat, is a function of
nonsteady fluxes in three dimensions, where
and Fdenotes an instantaneous flux (comprised of mean and fluctuating components) in
the vertical (w) or two horizontal (x, y) directions. The second stage is a onedimensional, vertical flux, FL,fkom the parcel of air to the underlying surface under
conditions which tend towards a steady-state,
where K is a turbulent diffusivity for Xp. Because equations (2) and (3) are only in
steady-state over long time periods it ~sdifficult to solve the systems of equations
necessary for solution. However, this conceptual model identifies important
characteristics of the system with respect to scale. For an air parceI within 1 or 2 m of
the surface the horizontal scale necessary for steady state flow is usually the order of
tens of meters. Where the horizontal fluxes, F, and Fyare small (low wind speeds, no
blowing snow) the system resolves into a classical one-dimensional flux between the
atmosphere and the surface. Most snow and chemical deposition models are based on
such a flux and require no further information on upwind landscape conditions (outside
of the immediate area) nor fluxes. However, when the horizontal fluxes are lase as in
windswept regions, then the local landscape flux FL is dependent upon the array of
upwind conditions and fluxes. For blowing snow, Tabler (1975) suggests an average
snow-particle transport distance of 3 km before complete sublimation, confiied by
?enson (1982) for Arctic Alaska. This finding suggests that the scale for equauon (2)
fuxes in wind-blown, open areas is several kilometers. Summing the fluxes over a
winter season calculates the annual deposition of snow and ions. Hence, the local spatial
configurations of landscapes and fluxes are necessary to calculate the seasonal flux at
a point in windswept regions.
The load of snow and ions in a landscape type is therefore not due entirely to the
characteristics of that landscape type but to the spatial configtu-ation of the landscape
type within the mesoscale landscape. It is therefore errreme& difficultto calculate the
snow water equivalent and ion load to a landscape type in a rvindnvept region using on&
soint knowledge of the armosphetic deposition rate and the specific aero@numic
202
J. W.Porneroy ef al.
characrensrics of rite landscape type. This multiple-scale effect becomes most
pronounced in the higher wind speed environments that undergo wind relocation of snow
and more pronounced for ions such as C1- and SO," that are conservative during
sublimation or scavenged by blowing snow processes. Therefore, to model snow and ion
accumulation in windswept environments,a combination of twodimensional, physically
based models of wind transport of snow and snow chemical accumulation, operating in
a distributed fashion over the mesoscale will be required. These models can be operated
and their results accounted for in existing geographical information syster;:
environments. Methods that do not explicitly take into account the regional configuratior,
of landscape types must do so implicitly, in an empirical manner. Unfortunately, such
empirical models cannot be confidently applied outside of their region of origin.
CONCLUSIONS
Ion loads vary by up to 5-fold in different landscape types within a mesoscale site and
up to 18-fold between mesoscale sites, along a transect that spans the arctic treeline. The
loads of ions and snow are sensitive to small and mesoscale variation in blowing-snow
regime in the following order
C1- >
r S W E > > NO3-
Two factors, operating at two scaIes, most strongly affect the load of snow and major
geochemical ions in snow at the arctic treeline. The first is the landscape roughness at
a small-scale (tens of meters), as parameterized by the leaf area index or by topographic
slope. There is a trend for rougher landscapes to receive greater inputs of snow and ions.
However, this trend becomes masked and eventually nonlinear where high wind speeds
and well-exposed regional landscapes promote relocation of snow. The second factor
operates at the mesoscale and is due to frequent blowing snow, causing the maximum
snow and ion loads to develop in moderately rough landscapes. In the subarctic these
landscapes are "snow sinks, " fringing the low-roughness landscapes from which blowing
snow is removed. The mesoscale (roughly 3 km)wind exposure and relocation of snow
can therefore strongly affect the small-scale landscape-snow relationship, making
determination of snow accumulation and chemistry from point measurements
impossible. A combination of twodimensional,' physically based modeIs of wind
transport of snow and snow chemical accumulation, operating in a distributed fashion
over the mesoscale will be required to predict snow accumulation and snow
accumulation chemistry in complex, windswept environments such as the tundra-taiga
transition.
Acknowledgments This work was supported by the authors' respective institutions, the
NATO Collaborative Grants Programme, the Canadian GEWEX Programme, the
Natural Sciences and Engineering Research Council of Canada (NSERC), the Polar
Continental Shelf Project, and the Science Institute of the Northwest Territories. The
efforts of Cuyler Onclin, William Quinton, Robert Reid, Ken Dion, Natasha Neumam,
and Art Dalton in the field and Joni Onclin and Ken Supeene in data reduction and
chemical analysis are greatly appreciated.
Sparial distribution of the snow chemical load at the tundra-taiga tramition
203
REFERENCES
3arrie. L. A. (1991) Snow formationand processes in the atmosphere that influence iu chemicalcomposition. In: Seasonal
Snowpackr. Procrssesof CompositionalChange(ed.by T. D. Davies. M. Tranter & H. G. Jones). 1-20. NATO AS1
Series G, vol. 28. Springer-Verlag. Berlin.
Benson. C. S. (1982) Reassessment of wincerprecipitationon Alaska's Arctic Slope and masuremenu on the flux of wind
blown snow. UAG R-288. Geophysical fnsrinm. Universiry of Alaska.
Boring. L. R.. Swank. W. T.. Waide. I. B. & Henderson. G. S. (1988) Sources. faces and impacu of nitrogen inputs to
terrestrial ecosystems: review and synthesis. Biogmchemirrty 6. 119-159.
Bucher. I . B. (1987) Forest damage in Swinedand. Aunria and adjacent pans of France and I d y in 1981. In: Effecrs of
Atmospheric Pollur~LSonForesrs. We~lnndr
rmdAgricuIrural&cosysrem (ed. by T. HuuhiruonL K. Meenn). 43-58.
NATO AS1 Series. vol. G16. Springer-Verlag. Berlin.
Delmar. V. & fonts. H. G. (1987) Wind as a factor in the d i r a t measuremenf of the dry deposition of wid pollupnu ID
snowcoven. In: S ~ U O M ~ S W W O ~ ~ ~ SChemisrty.
: P ~ ~ S ~Hydrology
CS.
(ed. by 3.G. Jones & W.Orville-Thormr).
321-336. NATO AS1 Series, vol. C211. D. Reidel. Dordrecht.
de march. L. (1975) Nutrient budges for a high arctic lake (Char Lake. N.W.T.). Verhandl. fnfemat. Verein. Thcoret.
Angewand. Limtwl. 19,496-503.
Goodess. C. M. & Paluukof. J. P. (1992) The development of regional climlc scenarios and the ecological impra of
greenhousegas cnnning. In: The&cologicalConrequmcuof Global CUmte Change(ed. by F. I. Woodward).33-62.
Advances in Ecolog~ulResearch vol. 22. A d e m l c Press, London.
Hudson. R. J. M.. Gherini. S. A. dc Goldstein. R. A. (1994) Modelling the globalcarbon cycle: Nitrogen feniliutionof
the terrestrial biosphere and the 'missing* C 4 sink. Gfobol Biogmchem. Qc. 8.307-333.
Jones. H. G.. Pomeroy. J. W.. Davies. T. D.. Marsh. P. k Tnnter. M. (1993) Snow-rtmosphere inrcracciom in M c
snovacks. net fluxes of NQ. SO, and influenceof solar ndiation. In: proceeding^ of rhe5GthAnruurl Eusrem Snow
Conference,255-264.
Marsh. P.. Quinton W. & Pomeroy. J. W.(1995) Hydrologidprocessesand runoffu the Arctic trecline in Norchwcstern
Canada. NordicHydrol. (in press).
Monserud. R. A.. Tchebakova. N. M. k Leemans. R. (1993) Global vegetationchange predicted by the modified Budyko
model. Climaric CllMge25.59-83.
Pomeroy, J. W. & Gny. D. M. (1994) Sensitivicyofsnow relocationand sublimationtocIirmte and surfacevegetation. In:
Snowandfce Covers:fnteractionswirhtheAmrosphereandEcarysr~(cd.
by H . G. Jones. T. D. Davies, A. Ohmura
& E. M. Morris) (Proc. Yokohama Symp.. July 1993). 213-225. IAHS Publ. no. 223.
Pomeroy. J. W., Marsh, P. k Gray. D. M. (1994) Snow accumulation and sublimation at the m d n - a i g a m i t i o n
Prescncation to the European Conferenceon rhc Gbbal Energy and Wurer Cyct. The Royal Society. Londoa. July
1994.
Pomeroy. I. W. & Schmidt. R. A. (1993) The use of fraccal geometry in modelling intercepted snow accumulation and
sublimation. Proceedingsof rhe 50rh Amurrl Earem Snow Conference. 1-10.
Pomeroy. J. W..Manh. P. & Lesack. L. (1993) Relocationof major ions in snow along the tundra-taigaecotone. Nordic
Hydrol. 24. 151-168.
Pomeroy. J. W., Davies. T. D. & Tranter. M. (1991) The impact of blowing snow on mow chemistry. In: Seasonal
Snowpacks. Processesof Compo~irionalChange ( 4 . by T. D. Davies. M.Tnnter k H. G. Jones). 71-114. NATO
AS1 Series G, vol. 28. Springer-Verlag. Berlin.
Smith. T. M.. Shugan, H. H.. Bonan. G. B. &Smith, 1. B. (1992) Modellingthe potential response of vegentiontoglobl
climate change. In: n e Ecological Consequences of Global Climare Change (ed. by F. I. Woodward). 93-116.
Advances in Ecological Research vol. 22. Academic Ress. London.
Steppuhn. H. & Dyck. G. E. (1974) Escimating.me basin snowcover. 1n:Advonced C o n c e p u d Techniquesinrhc Study
of Snow andfce Resowces. 314-318. Natlonal Academy of Sciences. Washingon, D.C.
fabler. R. D. (1975) Estimating the tmsponand evaporationof blowing mow. In: h w Munugentemon the Gzcm Phh
Symposium, 85-105. Publ. 73. Great Plains Agriculrural Council. Universi~,of Nebraska. Lincoln.
f imoney, K., Kenhaw. G. P. &Olesen. D. (1992) Lace winter snow-ldseaperelouonships in h e subarcticnear H&n
River. Great Slave Lake. Northwest Territories. Canada. Waf.Resour. Res. 2
5
0
.1991-1998.
'xc!ch. H. E. & Legault. I. A. (1986) Precipitation chemistry and chemical limnolog of fertilized and natural lakes at
Saqvaqjuac, N.W.T. Can. J. Fish. Aqwr. Sci. 43. 1104-1134.
Welles, 1. M. & Norman. J. M. (1991) Inst~umentfor indirect mearurementofcanop~uchiteccure.Agron. 1.83.818-825.
Woo. M.-K. & Marsh. P. (1978) Analysis of error in the determination of snow m n g e for small High Arctic b a s h .
I . Appl. Mer. 17. 1537-1541.
Woo. M.-K.. Heron. R.. Manh. P. & Steer. P. (1983) Comparison of wearherstation snowfall wirh winter snow
accumulation in high arctic basins. I. Amos.-Ocean 31. 312-325.