Spatial and temporal patterns of chloride deposition in southern Sweden

Spatial and temporal

patterns of chloride

depo-sition in southern Sweden

Mats Gustafsson, VTI and Eva Hallgren Larsson, Swedish Environmental Research Institute

pant © © CV en et ar, n id brad S d Hu Hsv] 142)

Swedish National Road and f Transport Research Institute

Spatial and temporal patterns of chloride

deposition in southern Sweden

Water, Air and Soil Pollution 124z345-369, 2000

Mats Gustafsson, VTI

Eva Hallgren Larsson, Swedish Environmental Research Institute

Copyright (2000) with kind

permission from Kluwer Academic Swedish National Road and

Publishers ' .

MATS E. R. GUSTAFSSON and EVA HALLGREN LARSSON2

1 Swedish National Road and Transport Research Institute, Linköping, Sweden; 2 Swedish Environmental Research Institute, Aneboda Research Station, Lammhult, Sweden

(* authorfor correspondence, e-mail: mats.gustafsson©rti. se) (Received 6 January 1999; accepted 14 December 1999)

Abstract. Monthly bulk deposition of chloride at 49 stations in Southern Sweden between 1989 and 1995 was used to produce quarterly and annual deposition maps through ordinary block kriging. Generally, deposition decreases from the west coast and eastward and displays a large annual vari-ation, governed by the frequency and intensity of midlatitude cyclones. The 1st quarter dominates the temporal pattern all years except 1992. The 4th quarter is the second most important and the 2nd and 3rd quarters have generally low deposition. The spatial deposition maximum was often displaced from the west coast to the western fringe of the South Swedish highland, due to orographic enhance-ment of precipitation. On the western rise of the highland, deposition almost entirely co-variates with precipitation. On the west coast and in the eastern part of Southern Sweden, the temporal pattern is a more complex result of precipitation and frequency of strong westerlies. Comparing the quarterly total dataset with precipitation and frequency of westerly gales shows that both the 3rd and 4th quarters have higher mean precipitation than the 1st quarter, but lower deposition, while the second quarter has substantially lower precipitation but almost equal deposition to the 3rd quarter. The frequency of westerly gales shows a clearer, linear relationship to quarterly deposition. The 1st

quarter has the highest variability in deposition and precipitation as well as in frequency of westerly

gales. The importance of single highly salt laden cyclones to the annual deposition is obvious in the 1st quarter of 1993. Changes in cyclone activity due to climate change is therefore of vital importance for the chemical characteristics of the midlatitude atmosphere.

Keywords: chloride, climate, deposition, geostatistics, kriging, pattern, Sweden

1. Introduction

The transport and deposition of marine ions are a natural and important part of the geochemical cycle and has been studied since the early fties in e.g. Scandinavia

(Eriksson, 1959) and North America (Junge and Gustafson, 1957). Sea salt

aero-sols are produced at sea and in coastal areas as waves break. The bubbles in the whitecaps burst and eject droplets of sea water into the air. Depending on particle size and meteorological conditions the aerosols are transported with the wind to eventually deposit. The droplets are the atmosphere s main source for condensation nuclei, why large amounts deposit as wet deposition. During dry conditions the im-portance of dry deposition increases. Spatial studies on deposition data have, so far, been performed with emphasis on acidifying components (Draaijers et al., 1995,

Pk! Water, Air, and Soil Pollution 124: 345 369, 2000.

VyaS and Christakos, 1997). Li, (1992) found annual bulk deposition of chloride in

the United States to vary from approximately 7 8 kg ha 1 y 1 on the south-eastern

coast to 0.3 kg ha 1 y 1 in the western interior parts. Chloride concentration in precipitation has been described by e.g. Van Leeuwen et al. (1996) and Carratala et al. (1998). Van Leeuwen et al. (1996) described precipitation concentration kriged from the EMEP network in Europe and emphasised the large increase in bulk deposition close to the Atlantic coast. Carratala et al. (1998) used kriging to map rain composition in a coastal area on the Spanish east coast at a similar scale as have been performed in the present paper. Their conclusions were that the contours for marine ions followed the coastline with values decreasing from the coast.

Lacking in the above examples are the temporal variation and a focus on marine constituents. The deposited salts are a vital contribution to the base cation supply, but display variable and sometimes less favourable effects. High sea salt dry depos-ition to vegetation can cause direct damage to salt sensitive species (e.g. McCune,

1991; Morris, 1992; Pedersen, 1993). Sea salt episodes has also been shown to cause short-term acidi cation in surface waters (e.g. Heath et al., 1992; Hindar,

1995). Sodium substitutes for hydrogen (H+) and aluminium (Al ) on the soil

colloids causing these ions to be released to the soil solute and further into surface waters. This causes temporary acid reduction in the soil, but increases the acidity of runoff. Soils with low buffering capacity, already subject to acidi cation have been shown to be particularly sensitive to this sea salt effect . On a longer time scale, though, the same sea salt episodes might mitigate acidi cation, depending on grade of acidity of soils and background deposition conditions (Harriman et al.,

1995).

This study aims to investigate the temporal and spatial patterns in deposited amounts of the most abundant marine ion chloride, through precipitation in open eld areas (bulk deposition). Since the anthropgenic contribution to chloride de position is very small in Southern Sweden, the deposition of chloride, in a relevant way, re ects the deposition of marine constituents. The calculated total sea salt contribution to the area is also discussed.

In later decades, as a result of the growing knowledge of acidi cation and air pollutants, national and international monitoring networks have been built up to monitor deposition. This have given scientists a possibility to study the spatial as well as temporal variation of deposition. In Sweden, the Swedish Environmental Research Institute (IVL) has monitored deposition of acidifying compounds since

1985 (e. g. Hallgren Larsson et al., 1995). The network includes precipitation stud

ies in open eld areas and throughfall and soil solution in forest areas at 130 background locations spread throughout Sweden.

2.1. AREA DESCRIPTION

Southern Sweden is characterised by a level landscape, dominated by agriculture close to the west coast and in the whole of the south western part. The country rises relatively fast to %150 m a.s.l. about 10 20 km from the west coast. The South Swedish highland then continues to rise slowly eastward and the crest (% 300 m a.s.l.) is reached approximately south of Lake Vättern. Further to the east the land sinks towards the Baltic. The landscape has a relatively low undulating relief and is mostly covered by coniferous or mixed coniferous/deciduous forests with some minor agricultural areas.

The climate is characterised by the proximity to the ocean and frequent passages of depressions in the prevailing westerlies, with cool stormy winters and cloudy,

mild summers. It is therefore, as a whole, termed maritime, even though the central

part of Southern Sweden is slightly more continental. The sea thus imprints the climate, but also the composition of the atmospheric deposition is strongly affected by marine aerosols.

2.2. SAMPLING

Measurements of bulk deposition were carried out in open areas. such as clear cuts or meadows in the vicinity of the forest observation plots. To avoid influence of throughfall the distance from collection site to nearest tree has to be at least 30 m, when the height of the tree was 20 m. Bulk deposition concists, in open eld measurements, mainly of wet deposition. To be able to calculate total deposition, dry deposition must be estimated for each environment, since it depends on i.e. exposure of the sampler, vegetation cover and season. Dry deposition of sea salt is likely to in uence chloride concentrations in the samples of this investigation, since the collectors were constantly open (Granat, 1974). Sampling of precipitation (for calculation of bulk deposition) was made through a funnel combined with a col lector placed on a pole 1.5 2.0 m above ground level (Figure 1). At the bottom of the funnel, there was a smaller funnel with a plastic netting (maze 2 mm) attached on top, in order to prevent contamination of the collected sample. During winter a snowsack was applied, which means a tubular plastic bag, mounted on PVC plastic rings at the top and a funnel and collector at the bottom. The snowsack was hanging in a holder on a pole (modi ed after LOvblad and Westling, 1989). All collectors were covered by aluminium foil in order to minimise effects of heat and sunlight on the chemical composition of the sample.

Samples were collected monthly. Collected volume was registered and the samp

les were analysed for pH, conductivity, alkalinity, SO4 S, Cl , NO3 N. NH4-N,

and, in some cases, Na+, K+, Ca , Mg , and Mn . All results were registered in a database. Deposited amounts were calculated in kg per hectare and month (kg

"";.

&

is,

&

är

& äg

&är

&.

"u.

%

r.._

äär

& ..

=== ::

====:

="..=:

ts.

&

_ir

Summer

Winter

Figure ]. Equipment for collecting precipitation in open eld areas (bulk).

ha 1 mo ) and can be summarised for different periods. Chloride analyses have been made by ion chromatography, EPA test method 300.0 (EMEP, 1996).

To obtain an estimation of total amount of sea salt deposited, the sea water ratio between total amount of salt and the amount of chloride is used. This is approximately 1.8 and is presented in parenthesis in the map legends.

2.3. GEOSTATISTICS

Data from 49 locations in southern Sweden during six hydrological years, October 1989 through September 1995, were used for the analysis (Figure 2). The criterion chosen was that all stations were to have data for all months for as many consecut-ive years as possible. The resulting network of stations have the advantage that all maps produced have the same spatial input which minimises differences in spatial pattern caused by different network structure when comparing maps. The disad vantage was that the network resolution was renounced. In this case, the criterion resulted in a large area with no stations at all in the north-eastern part of the maps, which must be considered when interpreting the results. The temporal resolution used were quarters of years. Primarily, we found this a suitable balance between resolution and number of maps. Secondly it in a relevant way divides the year in a cooler and a warmer part, which differ regarding cyclonic activity. The 1st (winter early spring) and the 4th (autumn winter) quarters usually have higher cyclonic activity (Martyn, 1992) with relatively strong westerlies and frontal precipitation

Scale: 100 km

A

;

V

Figure 2. Map of Southern Sweden with stations from the Regional forest deposition network used in geostatistical analyses marked as black dots. Circles indicate stations used in Figure 10.

causing accordingly high deposition of chloride. The spring and summer months of the 2nd and 3rd quarters have a lower cyclonic activity with a greater proportion of convective precipitation and lower chloride deposition. Data ranging from the 4th quarter of 1989 to the 3rd quarter of 1995 has been analysed, resulting in 24 consecutive quarters and ve years (1990 1994).

The geostatistical interpolation method kriging (Matheron, 1971) was used to create the deposition maps. This technique uses the semivariogram (Figure 3) in two dimensions to nd the spatial correlation between the point data. The semivari

ance @(h)) is de ned as

A 1 n 2

M) Zn ;tzw Zoe, +h>}

(1)

where Z (x,-) denotes the value of a variable in point xi, h is the sample spacing also called the lag. A plot of h) against h is known as the experimental semivariogram and is a useful tool for determining optimal weight for interpolation. Models can then be tted to the semivariogram scatterplot to determine the range, sill and the nugget. The sill is the value of f(k) where the variogram levels off at a certain range. Beyond this range the variable shows no spatial correlation. The nugget or nugget variance is the residual, spatially uncorrelated noise produced by measurement er-rors together with spatial variations that occur over distances much shorter than the sample spacing, and that consequently cannot be resolved (Isaaks and Strivastava,

1.2 sill % 0.8 0.6

V(

Ihl

)

L i l i l i l i i i l i i i l range l l I I | IJ l l J l l l l l l l l l I I | l I l l l l l l l 1 l 1x l l l 0 50000 100000 150000 200000 250000 300000 350000 400000

Ihl

1 l l l l l l l

Figure 3. Example of experimental semivariogram tted with gaussian model.

In the case of chloride deposition in Southern Sweden, the spatial correlation is not uniform in all directions. Since the chloride source is the sea, the data often show a pattern parallel to the west coast causing values in this orientation to have higher correlation. This is called anisotropy. In the geostatistical software Variowin (Pannatier, 1996), it is possible to study variograms in different directions. These can then be modelled using nested structures to obtain a variogram surface. When possible, variograms for the direction of highest and lowest correlation were used for modelling. If directional variograms could not be accomplished or if the

cross-validation (see below) resulted in large differences between kriged and real data,

omnidirectional variograms were used instead.

Ordinary block kriging in a 10 >< 10 km2 grid was performed using the nested directional or omnidirectional variogram models and the software GEO-EAS 1.2.1.

(Englund and Sparks, 1988). All models were either gaussian or spherical or a

combination of both. The kriged maps were then processed in the rasterbased GIS software Idrisi for Windows (Eastman, 1997). The many grayscales of the legend (Figures 4 8) was not made primarily for reading of deposition amounts to speci c grid cells, but to be able to discern patterns in the map parts and quarters where deposition is low.

The kriging technique works best on normally distributed data. Since the data used here has many low and few high values, we have transformed the data using its natural logarithm. The kriged results was back-transformed using the exponential function. This gives a biased distribution and the median of the kriged block instead of the mean. Kriging the transformed data provides better kriging results, but the kriging standard deviations (ksd) may not be used as error estimators. They can, though, be used for intercomparison between maps.

% <1 (<1.8) 1-2 (1.8-3.6) 2-3 (3.6-5.4) 3-4 (SA-7.2) 4-5 (7.2-9) 5-6 (9 10.8) 6-7 (m_s 12.6) 7-8 (12.6-14.4) 8-9 (MA-16.2) 9-10 (16.2-18) 10-12 (IS-21.6) 12-14 (21.6 25.2) 14-16 (252-288) 16-18 (28.8-32.4) 18-20 (32.4-36) 20-25 (36-45) 25-30 (45-54) - 30-35 (54-63) _ 35-40(63-72) . 40-45 (72-81) _ 45 50(81-90) . 50-55 (90-99) . 55-60 (99-108) >60 (>108)

lst quarter Bulk deposition of chloride (Calculated bulk deposition of sea salt)

(kg/ha)

Lower right: Mean quarterly contribution to annual bulk deposition

(%)

<1 (<1.8)

1-2 (1.8-3.6) 2-3 (3.65.4) 3-4 (SA-7.2) 12-14 (21.6-25.2) 14-16 (252-288) 16-18(28.8-32.4) .;;i;.-;z.gs;i§' 35-40 (63-72) - 40-45 (72-81) . 45-50(81 90) _ 50-55 (90-99) . 55-60 (99 108) . >60 (>108) 2nd quarter Bulk deposition of chloride (Calculated bulk deposition of sea salt)

(kg/ha)

Lower right: Mean quarterly contribution to annual bulk deposition

(%)

<1 (<1.8) 1-2 (LS-3.6) 2-3 (3.6 5.4)

!#:E'äfl'kxzöfåå i.» ' 4??-%% 3-4 (5.4-7.2)J:! ':'f' ' a 4-5 (72-9) 5-6(9-10.8) 6-7 (10.8 12.6) " 7-8(12.6-14.4) 8 9 (14.4 16.2) 9-10 (16.2-18) ' 10-12(18-21.6) _ 12-14 (21.6-252) 14-16 (252-288) ' 16-18(28.8-32.4) 18-20 (32.4-36) 20-25(36-45) _ 25-30(45-54) 30-35 (54-63) - 35-40(63-72) - 40-45 (72 81) - 45-50 (81-90) _ 50-55 (90-99) - 55-60 (99-108) . >60 (>108)

3rd quarter Bulk deposition of chloride (Calculated bulk deposition of sea salt)

(kg/ha)

Lower right: Mean quarterly contribution to annual bulk deposition

(%)

<1 (<1.8) 12 (1.83.6) 2-3 (3.6-5.4) 3-4 (5.4-72) as? 4-5 (72-9) 5-6 (9-10.8) 6-7 (10.8 12.6) 7-8 (12.6-14.4) 8-9 (14.4-16.2) 9-10 (16.2 18) 10-12 (ls-21.6) 12-14 (21.6-252) 14-16 (252-288) 16 18 (28.8-32.4) 18-20 (324-36) 2025 (36-45) 25-30 (45-54) 30-35 (54 63)

35-40 (63 72) _ 40-45 (72-81) . 45-5o(81-90) . 50-55(90-99) - 55-60 (99-108) - >60(>108) 4th quarter

Bulk deposition of chloride (Calculated bulk deposition of sea salt)

(kg/ha)

Lower right: Mean quarterly contribution to annual bulk deposition

(%)

<1 (<1.8) 1-2 (1.8-3.6) 2-3 (3.6-5.4) ; 3-4 (54-72) 4-5 (7.29) 5 6 (9-10.8) 6-7 (10.8-12.6) 7-8 (12.6-14.4) 8-9 (14.4-16.2) 9-10 (16.2-18) 10-12 (18-216) 1214 (21.6-25.2) 14-16 (252-288) 16-18 (288-324) 18 20 (32.4-36) 20-25 (36-45) 25-30 (45-54) 30-35 (54-63)

_{;; 35 40(63 72)} _ 40-45 (72-81) - 45-50(81-90) _ 50-55 (90-99) _ 55-60(99-108) . >60(>108)

?r, Year 3522 ." ägg?/)?]; -: ' RSV-23.41.44" . . ** gå / f.:./ '.. Bulk deposmon of åxå-'%';''fl

W _{chloride (Calculated bulk}

deposition of sea salt)

(kg/ha)

_Standard

deviation

3. Results

3.1. MAPS

Figures 4 7 shows mean quarterly bulk deposition of chloride from October 1989 to September 1995. Each figure also has a map of mean quarterly contribution to yearly bulk deposition.

Generally, bulk deposition of chloride decreases from west to east. The lst quarter (January March) (Figure 4) dominates deposition during the year. Only in 1992 does it compare with the rest of the quarters. It is mainly the western part of southern Sweden that is affected by the higher deposition while the eastern part receives lower and less variable amounts of chloride. In 1990 and 1995 the higher deposition is obvious along the entire west coast. In 1994 a more central part stands out and in 1991 two distinct deposition peaks are found approximately 50 100 km from the coast. In 1993, high deposition affects also central and a small part of the east coast of southern Sweden.

The 2nd quarter (April June) (Figure 5) is a clear contrast to the preceding quarter, with the lowest bulk deposition throughout the whole period. Except for the eastward decreasing trend an obvious feature is the maximum in the central western part of the area, which corresponds to the western part of the highland. In 1990, 1991, 1993, and 1995 this maximum is very distinct. It also spatially coincides with the large maximum in the lst quarter, 1991. Secondary maxima on the east coast can be seen in 1994 and 1995.

It is interesting to notice the disappearance of the central western maximum in

the 3rd quarter (July September) (Figure 6). Instead a more continuos decrease

from west to east dominates the spatial pattern. The bulk deposition is somewhat higher than in the 2nd quarter. In 1991 two small maxima are found in the

south-western part. In the years 1990, 1993, and 1994 the coastal area close to the Laholm

Bay (Figure 2) on the south-western part of the coast have the highest deposition.

During the 4th quarter (October December) (Figure 7), deposition in the west ern part increases. The increase is pronounced in 1990, 1991, and 1994. In 1992,

the 4th quarter has the highest deposition during the year, but it is still rather low. In 1993, the deposition pattern in the 4th quarter resembles the patterns of the 3rd. In 1990 and 1994 an area in the north westem part shows high deposition amounts. In 1990, this area stretches from the coast and approximately 100 km inland, while in 1994 the area is probably related to a single station at the south-west tip of Lake Vänern. 1989 shows a continuos deposition decrease from west to east.

The yearly mean bulk deposition (Figure 8) is strongly imprinted by the de position patterns of the lst quarter, but also dominating spatial features from other quarters are superimposed. Maximum bulk deposition are generally found on the west coast or in the western part of the highland, while the minimum is normally connected to the south-eastern part except for the southern coast. Most obvious

80 60 40

me

an

pe

rc

en

t

of

an

nu

al

ch

lo

ri

de

bu

lk

de

po

si

ti

on

20

quarter

Figure 9. Yearly percentage of chloride deposited in each quarter, 1990 1994.

coast received high amounts of chloride. The year is dominated by deposition from the rst quarter. 1991 shows a similar pattern, but with the high deposition area restricted to the central part of the west coast. Further, in 1992, deposition was relatively low while in 1993 large parts of southern Sweden received high chloride deposition. Only the south-east comer are unaffected by the extensive deposition related to the lst quarter. The yearly map of 1994 resembles 1990 1992, but with a rather narrow maximum on the central west coast. All years, except 1994 Shows a small, but well noticeable, inland maximum in the western central part. The yearly deposition of chloride is often dominated by deposition from one single month or even one Single storm.

Mean deposition during the period re ects the overall pattern for the area. The maximum deposition is connected to the western highland and a broad zone along the west coast have similar bulk deposition. The south-central part of the east coast has the lowest deposition and also the lowest variation.

Figure 9 shows the yearly percentage of chloride deposited in each quarter from 1990 to 1994. The years 1990 and 1993 have 54 and 60%, respectively, deposited in their rst quarter. In 1991 and 1992, though, the 4th quarter have a higher proportion of the yearly deposition with 46 and 37%, respectively.

__ mm __d ha' 600 _. pf ,)T. ff. f ..f ...ep(.kg x)60 westcoast-500'. 350 _ 1 400F 140 i o. ' (D A _ *c & 300l _30 ;: c:. ; _{: g} 200} \/ $20 100". lx, _"10 0 . . L . . . .410 öm. ..._{i ! {west highlandi}.. -. 60 500". ' .*50 400"- 540 : : o. ,. . - % å 300: . 30 g? a : ] : g 200". . 120 100; .;10 Z 1 0 . . i . . . .- . id ._..*() 600 ,...v...,...f._{- east hi ghland-}...60 500'. :50 400"-_. -40_{l g} A _ 4 U & 300; _"30 & :=. _ ' g-r _ J 200C 520 h of 100'- ; -10 35 * W Vf vu vr- v ' ' Tu ' ' : l . 1 30? 3 C * 25L L 8 20} _: ä : : A » . a 15? '. z . 10:-5.

U : 119.89- 119.90. 1901. 199'2 i993] i9§4l 1995

Figure IO. Quarterly deposition (dep) and precipitation (p) at three stations situated as seen in Fig ure 2. The lower diagram shows quarterly frequency of days with westerly wind speeds of 21 m s 1

WESTERLIES

The quarterly bulk deposition and precipitation for three stations in a transect across southern Sweden is shown together with the frequency of strong westerlies in Figure 10. This frequency is here defined as days with westerlies exceeding 21 m s"1 on the west coast (SMHI, 1989 1995). The stations are situated respectively on the west coast, in the western part of the highland and in the eastern part of the highland. The location in the western part of the highland is more exposed to airborne pollution since it is situated on a south westem slope, while the others are situated in level areas. The spatial variation can be large, why only the temporal patterns Should be interpreted.

The generally high frequency of strong westerlies during the rst quarter, in combination with high precipitation causes deposition peaks. This is especially true for the western highland station, where the deposition, precipitation and strong westerlies co variates to a higher degree than at the other stations. An exception is the precipitation peak in the 3rd quarter of 1993 related to a very wet summer but with a low frequency of strong westerlies.

The west coast and eastern highland stations lack the dominating precipitation peak in their lst quarters. Maximas in the 3rd and 4th quarters are at least as common. Since precipitation during these quarters to a greater extent is convective rather than frontal and therefore not generally connected to strong westerlies, this results in a more complex deposition pattern. The very high frequency of strong westerlies in the lst quarter of 1990 yields high deposition at the west coast station, despite relatively low precipitation. For the eastern station, the lst quarter of 1993 caused a marked deposition peak with 3 4 times higher chloride deposition than during the rest of the period. This peak is also dominant at the western highland station, but surprisingly anonymous at the westernmost station.

3.3. TOTAL DATASET DEPENDENCE ON PRECIPITATION, WESTERLY GALE FREQUENCY AND DISTANCE FROM WEST COAST

Figure 11 shows that the lst quarter has the highest mean deposition and the highest variability but not the highest mean precipitation amounts. The 2nd quarter has the lowest precipitation and deposition. The 3rd and 4th quarters have similar precip itation, but the 4th quarter displays a higher deposition and a higher variability in both precipitation and deposition. The quarterly total datasets show weak power law relationships between bulk deposition and precipitation, in accordance with

i.e. Beverland et al. (1998), with R2 equalling 0.33, 0.12, 0.22 and 0.40 for each

quarter, respectively.

The quarterly frequency of westerly gales over the whole period, is strongly related to bulk deposition of chloride (Figure 12). Bulk deposition increases with increasing frequency as does variability in both bulk deposition and frequency. Again the lst quarter stands out as the primary period for chloride deposition,

20 ' * ' ' 1 l I | 1 * i -4 15 :x r-Im ,_ a 1-m _| £4, 10 1 Cl _ r'1 # _9 - # LJ oå 1 -& 5 "* P14 _ "d 1 Jr åå

₌

_

_lz

_{_}

₁₁₁₃

₁

1 _D & TI

o _

T

_

l _5 _ 1 1 1 l 1 | 1 1 I 1 1 1 1 | 1 1 1 1 I 1 1 1 1 l 1 1 1 1 50 100 150 200 250 300 350

1) (mm)

Figure l]. Mean quarterly bulk deposition and precipitation for all years and all stations in the

dataset. Error bars show standard deviation.

20 I I I I I I I I I I I I T I | I I I T I I I I I 15 -:*Im _ _. .:::_co _ _ .::: 10 e -v L- 1 Cl rn _9 L m _ og ' 3 5 * r 14 _ FD ,_ " LJ

&

₌

_ 3 1

_{L m .} "D __{_} L1..1 J'_{2 l} 0 - _{l. _} _5 P 1 1 1 1 J 1 1 1 1 l 1 1 1 1 l 1 1 1 1 l 0 5 10 15 20 25 N(u>21 ms)

Figure 12. Mean quarterly bulk deposition and frequency of westerly gales for all years and all

_ d _ _ >- ». P I i- m ul _ Wo . v- u -ca F nu q h .. O . Q .- °. . 0 | 0 .. _ Ö m . .. _ Q.. _ 0 . "Ö 4 -< - 4 1- i I- _! - . h '! h 1 III]! I I O O 1 1 1 4 o v C ' v l O

rITT T III IIIT 1 I [IlllllT 1

I 1 M .. H . r- O . II b & ' O .- O . C? n O 1 _ %! . 1 O -1 1 ( >- . H . >= . h- C! - q _ '! I- + . F 1 r- 1 1111111 1 D 0 wt FA O H ' ' CD (Perl 351) dev 10 0 15 0 2 0 0 2 5 0 3 0 (1 (k m) 50 10 0 15 0 20 0 25 0 30 0 (1 (k m) 50

I- ++ ++ -4 . + +++ _ ru _| L . I + . _ i++ . _ % +++ _ _ o . 1- N" . _ M . »- -u_(* -O FO. n Q 1- 0 d " m -1 dZ |- h d .- . » 0._d) + -i- 'O -1 r.. _ p _ I- . i- . b . 1111 1 1 1 111111 1 1 1 8 O H 1 3 1 4 H o (121133) dap IIIII får [IIIII I #. _ H . - O _. . N" . _ M . .v"lh . O i- O . # I'- O -. 'n . Ö! _ N . & ll . . >x . _ ++ +'H' _ 1- _ 1111 1 1 1 11.1111 LL! O O 1 ( H E: H <0 (113531) dov 10 0 15 0 20 0 25 0 30 d (k m) 50 10 0 15 0 2 0 0 2 5 0 3 0 0 d (k m) 50 Fi gu re 13 . Qu ar te rl y bu lk de po si ti on at ea ch st at io n as re la te d to di st an ce fr om we st co as t.

followed by the 4th quarter. The 2nd and 3rd quarters are very similar despite the large difference in precipitation amounts.

Distance from the west coast has an obvious relationship with sea salt depos-ition. Figure 13 illustrates the quarterly inland decrease. The resulting curve ts have similar inclinations and re ect the quarterly deposition patterns well. Correl ation in the 1st quarter is low due to the high inter-annual variability and the high deposition in the western part of the highland. The 2nd quarter also shows low cor relation due to the inland maximum. The 3rd and 4th quarters show relatively small inter-annual variation and the inland maximum is less apparent and consequently have higher correlations.

4. Discussion

The spatial pattern of bulk deposition of chloride in southern Sweden is related to distance from the west coast superimposed by orographic precipitation reinforce-ment induced by the western part of the South Swedish highland. This latter effect often causes deposition maxima to be displaced from the west coast to the western fringe of the highland. The eastern half of southern Sweden has generally low chloride deposition and little variation, but with a few exceptions, e.g. in the 1st quarter of 1993.

The most important temporal features are the cyclic pattern and the large annual variation in bulk deposition. These are effects caused by the variation in frequency and intensity of mid latitude cyclones. Studying the weather statistics (SMHI, 1989 1995) reveals that single intense depressions sometimes are responsible for the major part of yearly bulk deposition. This is the case in the lst quarter of 1993, when, in January, a very deep low passed Southern Sweden with westerlies around 30 m s l. This episode caused high concentration and deposition of chloride at some inland locations. Also several locations north-east of the area in this study showed considerably larger amounts of chloride compared to more western loca-tions. The mechanism behind this phenomenon is not clari ed, but is probably due to dry deposition (Hallgren Larsson and Westling, 1994). The Sea salt deposition of this cyclone also caused temporary stream water acidification in southern Norway

leading to severe sh death (Hindar et al., 1995).

In 1990, twice as many days as in 1993 with strong westerlies where recorded. It is likely, though, that each of these cyclones where not as salt laden as the single storm of 1993. Nevertheless, they certainly caused a deposition maximum at the westernmost station in Figure 3, which indicates either that dry deposition con tributed to a large part of the deposition, or that the concentration of the relatively sparse precipitation was suf ciently high to cause this peak. It is likely that at

this site, situated %5 km from the coast, as well as at further inland stations, dry

deposition constitutes a relatively large part of the high deposition in the 1st quarter, 1990.

considered an artefact related to a single station. It is true that this station has gen-erally higher deposition (see below), but often, one or more neighbouring stations con rm the peak. To a certain extent this peak is probably due to the situation of the collector on a south-western hillside, making it more exposed to westerlies compared to other locations.

During the 2nd and 3rd quarters, deposition is low throughout the area. A differ-ence between the quarters is that in the 2nd quarter, the maximum of the western part of the highland is fairly dominant, while in the 3rd quarter the maximum is situated along the west coast. An explanation would be that on the coastal plain, the 2nd quarter normally is the yearly precipitation and deposition minimum, while on the western part of the highland, the 3rd quarter is the normal minimum. The western highland deposition in the 2nd quarter is higher than the coastal depos ition, but in the 3rd quarter the opposite condition prevails. This is supported by the time series in Figure 10. It is speculated that these alterations are connected to the proportion of regional westerlies to local wind systems as sea breeze and convective precipitation. These winds are generally not strong, but quite persistent. Convective precipitation in coastal areas, which is normally a part of the inland sea breeze circulation, have been shown to be important for the deposition of marine

ions because it is formed almost exclusively on marine aerosols (Inglis et al., 1995).

Deposition patterns of the 4th quarter are similar to those of the lst, but with a generally narrower coastal maximum. The reason for the maximum at the south-western tip of Lake Vänern in 1994 can not be traced in the meteorological bullet-ins. The yearly maps conclude that even though the deposition during the year is very variable, the total amount and distribution of the bulk deposition of chloride is quite conservative.

The study of three single stations partly explains the effects on bulk depos ition of co-variation, or lack of co-variation, between precipitation and frequency of strong westerlies. At the west highland station, the precipitation maximum is strongly related to the lst quarter, characterised by cyclone passages with high frequency of strong westerlies with enhanced precipitation amounts as the air raises above the highland. At the west coast station, precipitation is seldom orographic-ally induced and convective precipitation is likely to give high precipitation in the third quarter. Convective precipitation is normally not connected to strong west

erlies but might, as mentioned above contain rather high amounts of sea salt in

coastal areas. The marked deposition peak in the lst quarter of 1990 correlates to the very high frequency of strong westerlies. The concentration of marine salt in the air increases exponentially approaching the shore during strong onshore winds due to a succesively increased gravitational fallout of large droplets approaching the coastal zone (Gustafsson, 1996). It is probable that the coastal station, during strong westerlies is within reach of receiving a relatively large proportion of this coastal fallout and that this is the reason to the high chloride deposition. Convective pre cipitation is also the most probable cause to 3rd quarter high precipitation at the

eastern station. At the west coast and eastern stations also the 4th quarter has high precipitation which sometimes causes relatively high deposition. The precipitation in this quarter is most likely to be frontal but it seems as if the cyclones during this quarter generally do not involve as frequent strong westerlies as the cyclones of the

rst quarter, resulting in lower chloride deposition.

The quarterly dataset for the entire period and area shows that, as a whole, the frequency of westerly gales is the Single most important factor deciding the bulk deposition of chloride in Southern Sweden. The galeS are intimately connected to intense cyclones, often with large amounts of precipitation. Distance from the west coast explains the main features of the spatial pattern. The high temporal and Spatial variability of the lst quarter makes the correlation during this quarter low. Also the 2nd quarter has low correlation, not due to high variability, but to the maximum on the western highland, which recurs every year throughout the dataset. The 3rd and 4th quarters have markedly higher correlation, though still moderate. This mainly because these quarters have lower inter-annual variation in chloride bulk deposition and a more obvious west coast maxima. An additional explanation would be a higher proportion of dry deposition during the 3rd and 4th quarters. The dry deposition have previously been shown to be more directly related to downwind

distance from the west coast (Franzén, 1990; Gustafsson and Franzén, 1996), why

a larger proportion should cause a higher correlation.

Considering the importance of both cyclonic intensity and frequency on sea salt deposition, it is important to study the possible effects of changes in these paramet ers as well as cyclone tracks and temporal distribution due to climate change. In southern Sweden, increasing cyclone activity would mean a higher input of marine base cations to support acid sensitive soils. However, ion exchange between Na+ and H+ might take place in acidic soils causing acute toxic effects in surrounding waters (Hindar et al., 1995). Also, extreme sea salt input could certainly cause events of the sea salt effect in the south-westem part of Sweden.

4.1. CROSS-VALIDATION AND ERROR ANALYSES

Variogram quality can be investigated using cross-validation. The measured values at each station is compared to the kriged value at the station using the neighbouring stations (Van Leeuwen et al., 1996). Two measures of the quality of the kriged maps were used

__ __2

(D

D )

(D D')

A _

ksa

and B _ ;(st

where D denotes the measured value and D the interpolated value at each station, ksd is the kriging standard deviation and ksd2 is the kriging variance. If A ap-proaches zero and B one, the interpolation is satisfactory. Table I shows that the kriging results are very variable. Some maps previously based on directional vari ogram models had to be replaced with Simpler omnidirectional model based maps

Results of cross validation using measures from Van Leeuwen et al. (1996) Year Quarter A B 1989 4 0.028 1.368 1990 1 0.038 1.103 1990 2 0.007 0.968 1990 3 0.026 0.964 1990 4 0.014 1.052 1990 y 0.030 0.814 1991 1 0.013 0.875 1991 2 0.013 2.438 1991 3 0.038 1.440 1991 4 0.007 2.703 1991 y 0.012 1.792 1992 1 0.000 1.336 1992 2 0.016 1.007 1992 3 0.001 3.074 1992 4 0.015 1.214 1992 y 0.022 2.059 1993 1 0.059 1.061 1993 2 0.010 2.003 1993 3 0.001 0.582 1993 4 0.021 1.148 1993 y 0.021 1.559 1994 1 0.013 1.810 1994 2 0.015 1.001 1994 3 0.003 1.261 1994 4 0.008 1.444 1994 y 0.007 1.493 1995 1 0.020 1.270 1995 2 0.017 1.760 1995 3 0.004 1.227

to improve the quality measures. The sometimes low quality of the kriging can probably be assigned to the stochastic characteristics of precipitation. The results of Van Leeuwen et al. (1996) shows that marine ions typically Show somewhat lower quality measures than anthropogenic ions because of the high spatial variab-ility due to distance from the coasts. The authors state that these ions ideally need

(kriged standarddeviation)

(kg/ha)

3rd quarter

4th quarter

Figure 14. Example of maps of kriged standard deviations (ksd) for each quarter in 1990. Note the difference between spherical models (2nd and 4th quarter) and gaussian models (quarter l and 3).

different variogram models at different distances from the sea but that this makes it impossible to fit together a large scale map. Nevertheless, many of the maps show good cross validation results.

The errors of the bulk deposition maps can be divided into two groups; errors caused by assumptions of the methods and errors related to the interpolation. Due to the high variation over small distances in precipitation it is dif cult to decide whether or not a station is representative for its surroundings. An additional prob-lem, not addressed in the analysis, concerning mainly the dry deposition of sea salt, is the importance of exposure to westerlies. Especially important is this to the ex-posed station on the western part of the highland which is situated on a south-west facing slope and often the centre for the west highland maximum. The generally high deposition at this point is, at least partly, a result of generally higher dry de position. The high bulk deposition is often followed by one or more neighbouring stations, indicating a more widespread maximum. Also, sometimes the maximum is closer connected to other stations on the western fringe of the highland. The frequency of different wind direction during the quarters naturally also plays a part

when it comes to dry deposition. Van Leeuwen et al. (1996) estimated this error to

on altitude and wind conditions the error can be as large as 30% (Kovar et al.,

1991).

The kriging technique provides error estimates through the kriging standard deviation (ksd). As mentioned, using logarithmisised values makes the ksd less valuable, but intercomparison between maps might be made. An example of ksd maps of 1990 is shown in Figure 14. The kriging standard deviation is larger in areas which lack stations, like the north eastern part of the maps. The patchy appearance of the ksd-map of the 4th quarter illustrates the effect of choice of variogram model. This quarter is modelled using a spherical model, as is the an quarter, while the other quarters use gaussian models. The reason why the second quarter does not display the patchy appearance is probably due to the legend clas sification. The sometimes strong dominant maximum in the western part of the highland is probably further enforced by the lack of stations situated directly on the coast, presumed to occasionally receive high amounts of dry deposition.

5. Conclusions

From this investigation of chloride bulk deposition in Southern Sweden it is con cluded that:

. The spatial pattern is dominated by a general decrease from west to east, but this pattern is often superimposed by an orographically induced precipitation pattern, causing the western part of the South Swedish highland to be the area of maximum deposition.

. In the 2nd quarter, the above mentioned inland peak is frequent, while in the 3rd quarter it practically never occurs. This is probably due to the fact that the western part of the highland normally has a precipitation minimum in the 3rd quarter, while the coastal area has it s minimum during the 2nd quarter.

. The temporal variation during the year is substantial, with a pronounced de position peak during the winter half of the year. The lst quarter dominates the yearly deposition followed by the 4th quarter. The 2nd and 3rd quarters expose low deposition.

. The frequency of high westerly wind speeds in combination with high precip itation amounts are the prime factors affecting the temporal pattern. These fea-tures are normally related to frequency and intensity of mid-latitude cyclones. In coastal environments a high frequency of strong westerlies alone can cause high deposition due to dry deposition. The 3rd quarter has the highest precip itation, but it is seldom associated with strong westerlies and the proportion caused by convection over land is high, resulting in low chloride deposition. . Single, intense and salt laden cyclones can contribute to a major part of the

ition pattern of the lst quarter, mainly caused by a single storm, totally domin-ates the yearly pattern making up approximately 60% of the yearly deposition.

Acknowledgements

The authors are grateful to the following persons: Dr. Johan Knulst for help with data management and station knowledge, Mr. Jonas Norrman, Ass. Prof. Lars Fran-zén, Ms. Cecilia Akselsson, Ass. Prof. Deliang Chen and Mrs. Agnetha Gustafsson

for valuable points of view on the manuscript, Dr. Mats Söderström for valuable

discussions on the geostatistics and Ms. Marie Eriksson for lending us necessary computer equipment.

References

Beverland, I. J ., Crowther, J. M., Srinivas, M. S. N. and Heal, M. R.: 1998, Atmospheric Environment 32(6), 1039.

Carratala, A., Gomez, A. and Bellot, J.: 1998, Water, Air, and Soil Pollut. 104, 9.

Draaijers, G. P. J., Van Leeuwen, E. P., Potma, C., Van Pul, W. A. J. and Erisman, J. W.: 1995, Water, Air, and Soil Pollut. 85, 2389.

Eastman, J. R.: 1997, IDRISIfor Windows v. 2.0, Clark University, Graduate Schoool of Geography,

Worcester, U.S.A.

EMEP: 1996, Manualfor Sampling and Chemical Analysis, EMEP/CCC-report 1:95. NILU, Kjeller.

Enlund, E. and Sparks, A.: 1988, GEO-EAS (Geostatistical Environmental Assessment Software): User s guide, EPA/600/4 033a, Environmental Monitoring Systems Laboratory, U.S. Environ

mental Protection Agency, LaS Vegas, Nevada.

Eriksson, E.: 1959, Tellus 11(4), 375. Franzén, L. G.: 1990, AMBIO 19(4), 180.

Granat, L.: 1974, On the Deposition of Chemical Substances by Precipitation , Report A 27. Dep. of Meteorology, University of Stockholm, Sweden.

Gustafsson, M. E. R. and Franzén, L. G.: 1996, Atmospheric Environment 30(6), 977.

Hallgren Larsson, E. and Westling, O.: 1994, Airborne Pollution in Central Sweden, Deposition and Effects October 1992 September 1993. Yearly report (in swedish). Swedish Environmental

Research Institute, IVL. Report B 1139.

Hallgren Larsson, E., Knulst, J., Malm, G. and Westling, G.: 1995, Water, Air, and Soil Pollut. 85, 2271.

Harriman, R., Anderson, H. and Miller, J. D.: 1995, Water, Air, and Soil Pollut. 85, 553. Heath, R. H., Kahl, J. S. and Borton, S. A.: 1992, Water Resources Research 28(4), 1081. Hindar, A., Hendriksen, A., Kaste Q. and Torseth, K.: 1995, Water, Air, and Soil Pollut. 85, 547. Inglis, D. W. F., Choularton, T. W., Wicks, A. J., Fowler, D., Leith, 1. D., Werkman, B. and Binnie,

J.: 1995, Water, Air, and Soil Pollut. 85, 2119.

Isaaks, E. H. and Strivastava. R. M.: 1989, An Introduction to Applied Geostatistics, Oxford

University Press, 561 p.

Junge, Chr. E. and Gustafson, P. E.: 1957, Tellus 4, 164. Li Yuan-Hui: 1992, Water, Air, and Soil Pollut. 64, 561.

Lövblad, G., Westling, G.: 1989, in Nihlgård, B. and Gyllin, M. (eds), Methods for Integrated

Monitoring in the Nordic Countries, The Working Group for Environmental Monitoring, Nordic

Science Publishers B.V., Amsterdam, The Netherlands.

Matheron, G.: 1971, The Theory of regionalizea' Variables and its Applications, Les Cahiers du centre

de morphologie; Mathematique de Fontainebleau, Ecole des Mines de Paris, v. 5. McCune, D. C.: 1991, Environmental Pollution 74, 176.

Morris, E. C.: 1992, Australian Journal of Ecology 16, 141.

Pannatier, Y.: 1996, Variowin Software for Spatial Data Analysis in ZD, Springer Verlag, New York, NY.

Pedersen, L. B.: 1993, Forest and Landscape Research 1, 35.

SMHI (Swedish Meteorological and Hydrological Institute): 1989 1995, Monthly Weather Bulletins

(Väder och Vatten).

Van Leeuwen, E. P., Draaijers, H. P. J. and Erisman, J. W.: 1996, Atmospheric Environment 30(4),

2495.