**3. Methods**

The data are collected from the NILS programme which uses a strategically and systematically placed gridwork of 55 km squares (Ståhl et al., 2011). There are in total 631 squares and a full rotation of the inventory consists of 5 years. Two types of inventory are undertaken within a central, 11 km square, i) inventory by aerial photo interpretation in

**Zone 2. The Northern boreal zone** has wide heaths and forests of mountain birch or coniferous trees, mainly Norway Spruce (*Picea abies*) but also Scots Pine (*Pinus sylvestris*),

**Zone 3. The Upper middle boreal zone** has mixed coniferous forests in flat or hilly terrain and is rich in mires of mixed types, together with willow and birch swamps. The zone is predominantly above the highest post-glacial coastline and the metamorphic bedrocks are

**Zone 4. The Lower middle boreal zone** is vegetated by coniferous forest and swamp forest, in a hilly terrain rich in lakes. The transition from above to below the highest post-

**Zone 5. The Coastal boreal zone** is characterised by sandy or morainal islands and coastal boreal plains, with hilly terrain and valleys dominated by marine sediments back from the coast. The vegetation ranges from the marshlands, meadows and lowland forests along the Gulf of Bothnia coast to the inhabited parts where cultivated land and towns are situated (sparsely), which brings open grassland and deciduous forest between the coniferous parts. The precipitation is somewhat higher than the inland forest zones.

Fig. 2. The five Elevation Zones from west to east: 1 Arctic/Alpine, 2 Northern boreal, 3 Upper middle boreal, 4 Lower middle boreal, and 5 Coastal boreal zone. The zones are based on the Natural Geographic Regions of Sweden (Helmfrid, 1996). The table in the figure shows the mean, minimum and maximum of elevation heights (m above sea level),

The data are collected from the NILS programme which uses a strategically and systematically placed gridwork of 55 km squares (Ståhl et al., 2011). There are in total 631 squares and a full rotation of the inventory consists of 5 years. Two types of inventory are undertaken within a central, 11 km square, i) inventory by aerial photo interpretation in

mixed with mires. The relief includes scattered mountains and plateaus.

glacial coastline occurs predominantly in this zone.

and the standard deviation of the mean height.

**3. Methods** 

covered by till. Few and scattered human settlements of late date can be found.

stereo models and ii) field-based inventory of sample plots and sample transects. This chapter will deal only with the aerial photo part of the inventory. The data set consists of the first set of air photo samples collected in 2003 to 2005, and is restricted to the northern part of the country. The data used here consists of 116, 11 km squares, which contain a total of 3229 registered polygons interpreted from the Colour Infra-Red imagery (Fig. 3).

Fig. 3. The study area with five Elevation Zones, and with the distribution of the 1x1 km squares. The grid locations of the permanent NILS squares are indicated in red, the yet unsampled squares are not filled.

## **3.1 Aerial photo interpretation**

The scale of aerial photos allows mapping based on vegetation structure and overall vegetation composition (e.g. Ihse, 2007; Morgan et al., 2010). A basic level of classification utilizes the fundamental division into bog (rain-nourished) or fen (mineral soil nourished). Each of the finer divisions is based on different hydrotopographic conditions slope (e.g., flat, sloping, raised), location in landscape, and distinctive vegetation patterning on the surface. At the polygon level the types are described for general dominance of main vegetation in the understory. Within polygons smaller units, such as hummocks, hollows, and flarks, can be identified. This means that one can use the structure of the vegetation and peat surface, along with several other ecological indications, to distinguish between the different main types of mires and vegetation elements inside them. On the landscape scale the different hydrotopographic types of mires can be put together as mire complexes (Rydin et al., 1999). We hypothesize that the geographic distributions and areal extents of types will vary from east to west and north to south in northern Sweden, and also when comparing north with south Sweden.

Main Ecosystem Characteristics and Distribution of Wetlands

below these heights falls into the shrub/small tree variable.

**3.1.3 Tree and shrub cover** 

**4. Wetland area and distribution** 

community development.

in Boreal and Alpine Landscapes in Northern Sweden Under Climate Change 199

The variables dealing with tree and shrub cover are important indicators of communityecosystem and population-species biodiversity, land influences, disturbances, and climate change. These variables are important to identify those polygons that have had some influence by forestry, for example, tree planting being indicated by the patterns of tree distribution. In the aerial photo inventory, the distinction between trees and shrubs/small trees are as follows: all woody vegetation above 3 m height and in mountain birch forest above 2 m height are considered trees, regardless of species, and all woody vegetation

The main factors of peatland formation are climate, relief (topography), parent material, biota, and time. The key factor is relief, with flatness or concavity being the most important types of relief. Within the climate regime, the two climatic factors temperature and moisture (humidity) are interrelated in terms of control on wetland and peatland biota and

The total number of wetland polygons in 116 sampled squares was 3229, and the average number of polygons per square was 28.9 (SE 2.1, median 25, minimum 1, maximum 105). Across the Elevation Zones the highest numbers of squares were 34 in the Arctic/Alpine zone, and 32 in the Northern boreal zone, decreasing to 23 in the Upper middle boreal zone, 14 in Lower middle boreal, and 13 in the Coastal boreal zone (Table 1). The higher numbers in the Arctic/Alpine, Northern boreal, and Upper middle boreal are because of the NILS sampling strategy to establish a denser gridwork of

The interpretation methods are described in detail in Allard et al. (2003, 2010). Strict rules are applied for spatial mapping accuracy and timing during the vegetation season. The technology is based on viewing the digital images in stereo in a computer-based photogrammetric system, with ancillary data such as maps and geology in a geographic information system (GIS program). The NILS program uses quantitative and qualitative variables in a context-dependent variable flow, instead of the usual predefined classes. This allows for aggregation of variables in several different ways, making several different classifications possible, after the data is registered into a geographical database. Similar basic variables are used both for the aerial photo interpretation and for the field survey, although the entire square is inventoried in the aerial photos, and sample plots and line inventories are registered in the field. A decision tree was used to define criteria for the polygon delineation in the aerial photo inventory, with a smallest mapping unit of 0.10 ha (31.631.6 m). When this study was made, complete data from the interpretation was available for the years 2003, 2004 and 2005 of the first rotation in NILS.

#### **3.1.1 Variables in NILS relevant to mires**

The NILS program operates on variables, in total 356 of which 156 are documented in the aerial photo interpretation. The main variables documented in wetlands and peatlands come from the recognition of the division between fen and bog; the recent emphasis in peatland classification on physiography, morphology, and hydrotopography; and vegetation elements (e.g., National Wetlands Working Group, 1988; Löfroth, 1991; Gunnarsson & Löfroth, 2009).

There are 17 Hydrotopographic mire types recognized for Swedish conditions (Gunnarsson & Löfroth, 2009). In this study, the mires were the primary focus, but we added a 'Non-mire wetland' type by using the category of semiaquatic land (occasionally flooded by fresh, brackish or salt water). Swamp forests are not included as one of the Hydrotopographic mire types, although they are registered in the aerial photo interpretation. The Hydrotopographic mire types are registered in the aerial photo interpretation of polygons as uniform hydrotropographic units, and can be classified together as large-scale mire complexes.

Thirteen Field and bottom layer vegetation types are recognized. Only one type of field and bottom layer type is registered as the dominant class for each polygon, since several types may occur in one polygon. Seven elements in a Microtopographic series are registered as percentages inside each polygon. The Microtopographic elements are defined on physiognomic as well as vegetation composition aspects of the field and bottom layer, ground level, and firmness of the substrate, following the Microtopographic series tradition in Swedish mire ecology (Rydin & Jeglum, 2006).

#### **3.1.2 Influences and disturbances**

Disturbance descriptors are important when explaining the kind of vegetation and ground cover. In the NILS program they are registered for example as peat harvesting, ongoing or historical, and influences including such disturbances as vehicle tracks, ditching, dredging, and mechanical site preparation. Other facts that can be derived from the data are forestry impacts, mowing, agriculture, erosion by wind or water, and burning. Of particular interest for climate change is the designation of productive forest land, fjell and tree line, and treelimiting climatic impediments below the tree line.

#### **3.1.3 Tree and shrub cover**

198 Ecosystems Biodiversity

The interpretation methods are described in detail in Allard et al. (2003, 2010). Strict rules are applied for spatial mapping accuracy and timing during the vegetation season. The technology is based on viewing the digital images in stereo in a computer-based photogrammetric system, with ancillary data such as maps and geology in a geographic information system (GIS program). The NILS program uses quantitative and qualitative variables in a context-dependent variable flow, instead of the usual predefined classes. This allows for aggregation of variables in several different ways, making several different classifications possible, after the data is registered into a geographical database. Similar basic variables are used both for the aerial photo interpretation and for the field survey, although the entire square is inventoried in the aerial photos, and sample plots and line inventories are registered in the field. A decision tree was used to define criteria for the polygon delineation in the aerial photo inventory, with a smallest mapping unit of 0.10 ha (31.631.6 m). When this study was made, complete data from the interpretation was

The NILS program operates on variables, in total 356 of which 156 are documented in the aerial photo interpretation. The main variables documented in wetlands and peatlands come from the recognition of the division between fen and bog; the recent emphasis in peatland classification on physiography, morphology, and hydrotopography; and vegetation elements (e.g., National Wetlands Working Group, 1988; Löfroth, 1991; Gunnarsson &

There are 17 Hydrotopographic mire types recognized for Swedish conditions (Gunnarsson & Löfroth, 2009). In this study, the mires were the primary focus, but we added a 'Non-mire wetland' type by using the category of semiaquatic land (occasionally flooded by fresh, brackish or salt water). Swamp forests are not included as one of the Hydrotopographic mire types, although they are registered in the aerial photo interpretation. The Hydrotopographic mire types are registered in the aerial photo interpretation of polygons as uniform

Thirteen Field and bottom layer vegetation types are recognized. Only one type of field and bottom layer type is registered as the dominant class for each polygon, since several types may occur in one polygon. Seven elements in a Microtopographic series are registered as percentages inside each polygon. The Microtopographic elements are defined on physiognomic as well as vegetation composition aspects of the field and bottom layer, ground level, and firmness of the substrate, following the Microtopographic series tradition

Disturbance descriptors are important when explaining the kind of vegetation and ground cover. In the NILS program they are registered for example as peat harvesting, ongoing or historical, and influences including such disturbances as vehicle tracks, ditching, dredging, and mechanical site preparation. Other facts that can be derived from the data are forestry impacts, mowing, agriculture, erosion by wind or water, and burning. Of particular interest for climate change is the designation of productive forest land, fjell and tree line, and tree-

hydrotropographic units, and can be classified together as large-scale mire complexes.

available for the years 2003, 2004 and 2005 of the first rotation in NILS.

**3.1.1 Variables in NILS relevant to mires** 

in Swedish mire ecology (Rydin & Jeglum, 2006).

limiting climatic impediments below the tree line.

**3.1.2 Influences and disturbances** 

Löfroth, 2009).

The variables dealing with tree and shrub cover are important indicators of communityecosystem and population-species biodiversity, land influences, disturbances, and climate change. These variables are important to identify those polygons that have had some influence by forestry, for example, tree planting being indicated by the patterns of tree distribution. In the aerial photo inventory, the distinction between trees and shrubs/small trees are as follows: all woody vegetation above 3 m height and in mountain birch forest above 2 m height are considered trees, regardless of species, and all woody vegetation below these heights falls into the shrub/small tree variable.

## **4. Wetland area and distribution**

The main factors of peatland formation are climate, relief (topography), parent material, biota, and time. The key factor is relief, with flatness or concavity being the most important types of relief. Within the climate regime, the two climatic factors temperature and moisture (humidity) are interrelated in terms of control on wetland and peatland biota and community development.

The total number of wetland polygons in 116 sampled squares was 3229, and the average number of polygons per square was 28.9 (SE 2.1, median 25, minimum 1, maximum 105). Across the Elevation Zones the highest numbers of squares were 34 in the Arctic/Alpine zone, and 32 in the Northern boreal zone, decreasing to 23 in the Upper middle boreal zone, 14 in Lower middle boreal, and 13 in the Coastal boreal zone (Table 1). The higher numbers in the Arctic/Alpine, Northern boreal, and Upper middle boreal are because of the NILS sampling strategy to establish a denser gridwork of

Main Ecosystem Characteristics and Distribution of Wetlands

size (Number of squares)

Elevation zone Sample

vektor, 1:100 000, Swedish Land Survey)

proportion of the terrestrial land area.

0

1 Arctic/Alpine 2 Northern

**4.1.1 Variation and adequacy of sample** 

and the adequacy of sample (Table 1).

boreal

3 Upper middle boreal

**Elevation zone**

Fig. 4. Estimated total areas of wetlands in the five Elevation zones: Arctic/Alpine, Northern boreal, Upper middle boreal, Lower middle boreal, and Coastal boreal. Error bars indicate

There was high variation among squares owing in part to the variation of wetlands within the zones, and also to the variation in numbers of squares amongst the zones. Relative standard error (Standard error divided by the estimate) indicates the quality of the estimate

The lowest numbers for relative standard error for the wetland area estimate were obtained for the Northern and Upper middle boreal zones, indicating an adequate sample size of 32

4 Lower middle boreal

5 Coastal boreal

2000

standard errors.

4000

6000

8000

**Total area (km2)**

10000

12000

14000

\*

in Boreal and Alpine Landscapes in Northern Sweden Under Climate Change 201

Total terrestrial land area (km2)\*

Terrestrial land area (excluding water) obtained from the Swedish road map series (GSD-Vägkartan,

Table 1. Wetland statistics in the 116 11-km2 squares. Estimated wetland area as a

1 Arctic/Alpine 34 33905 12.39 (2.42) 0.20 2 Northern boreal 32 45084 25.86 (3.57) 0.14 3 Upper middle boreal 23 37769 26.54 (4.10) 0.15 4 Lower middle boreal 14 23860 16.35 (6.02) 0.37 5 Coastal boreal 13 12127 14.36 (3.51) 0.24 Total 116 152747 20.56 (1.87) 0.09

Estimated percentage wetland of terrestrial area, with standard

Relative standard error

errors

squares in the Alpine and adjacent interior land where landscape data are redundant or missing (Ståhl et al., 2011).

#### **Calculation procedure for determining wetland areas for zones:**


\* The distribution of sample squares follows a stratified design with least sampling effort in the central boreal parts of Sweden (Fig. 3; Ståhl et al. 2011). Stratum borders do not coincide with zone borders.

#### **4.1 Wetland areas in squares and zones**

The total wetland areas compared amongst Elevation Zones were significantly different (One-way ANOVA, F = 4.76 and P = 0.001; Fig. 4). The Northern and Upper middle boreal zones achieved the highest total wetland areas, and the Coastal boreal zone the lowest. The minimum cover achieved among squares was 1180 m2 in the Coastal boreal zone; the maximum was 898163 m2 in the Upper middle boreal zone, which is close to filling a full 11-km2 square (1 million m2). Expressed as percentages of the 11 km2 square (106 m2) these are 0.12 and 89.8%, respectively.

Total wetland area depends on the areal sizes of the elevation zones, which are not equal. To obtain a truer picture of the relative area covered by wetlands, we calculated the relative percentages of land covered by wetland in each zone (Table 1). The highest percentages of wetlands are in the Upper middle boreal zone followed closely by the Northern boreal zone. The Arctic/Alpine and the Coastal boreal zones have the lowest percentages, with a slightly higher value in the Lower middle boreal zone. The percentage cover of wetlands in the whole study area was 20.56% (SE 1.87). This matches well with other estimates of about 20% for Swedish wetlands (Hånell, 1989; Joosten & Clarke, 2002; Paavilainen & Päivänen, 1995; Christensen et al., 2007; Gunnarsson & Löfroth, 2009). However, all these references refer to the whole of Sweden, whereas the present paper deals only with Northern Sweden. A summary of wetlands comparable to the present one could be done for all southern Sweden below the study area.

squares in the Alpine and adjacent interior land where landscape data are redundant or

1. for each stratum\*, divide the total wetland area per inventoried 11 per km square in the zone by the number of squares sampled to get '*mean wetland cover /* 

3. multiply by the total number of squares from which the sample was taken in

4. for each zone, calculate the sum all strata areas to get '*estimate of the total wetland* 

5. obtain the '*estimate of the total terrestrial area per zone*' by doing step 1 to 4 for

6. obtain the *'wetland proportion'* by dividing the estimated total wetland area

7. multiply the wetland proportion by the corresponding terrestrial area of the

\* The distribution of sample squares follows a stratified design with least sampling effort in the central boreal parts of Sweden (Fig. 3; Ståhl et al. 2011). Stratum borders do not coincide with

**Calculation procedure for determining wetland areas for zones:** 

2. multiply by 25 to get 'wetland cover / 55 per km square'

terrestrial land cover instead of wetland cover.

zone to obtain '*ratio estimate of total wetland area*'.

the stratum to get '*estimate of the total wetland area per stratum*'

(from step 4) by the estimated total terrestrial area (from step 5);

The total wetland areas compared amongst Elevation Zones were significantly different (One-way ANOVA, F = 4.76 and P = 0.001; Fig. 4). The Northern and Upper middle boreal zones achieved the highest total wetland areas, and the Coastal boreal zone the lowest. The minimum cover achieved among squares was 1180 m2 in the Coastal boreal zone; the maximum was 898163 m2 in the Upper middle boreal zone, which is close to filling a full 11-km2 square (1 million m2). Expressed as percentages of the 11 km2 square (106 m2)

Total wetland area depends on the areal sizes of the elevation zones, which are not equal. To obtain a truer picture of the relative area covered by wetlands, we calculated the relative percentages of land covered by wetland in each zone (Table 1). The highest percentages of wetlands are in the Upper middle boreal zone followed closely by the Northern boreal zone. The Arctic/Alpine and the Coastal boreal zones have the lowest percentages, with a slightly higher value in the Lower middle boreal zone. The percentage cover of wetlands in the whole study area was 20.56% (SE 1.87). This matches well with other estimates of about 20% for Swedish wetlands (Hånell, 1989; Joosten & Clarke, 2002; Paavilainen & Päivänen, 1995; Christensen et al., 2007; Gunnarsson & Löfroth, 2009). However, all these references refer to the whole of Sweden, whereas the present paper deals only with Northern Sweden. A summary of wetlands comparable to the present one could be done for all southern Sweden

missing (Ståhl et al., 2011).

*square*';

*area per zone*'

zone borders.

**4.1 Wetland areas in squares and zones** 

these are 0.12 and 89.8%, respectively.

below the study area.


\* Terrestrial land area (excluding water) obtained from the Swedish road map series (GSD-Vägkartan, vektor, 1:100 000, Swedish Land Survey)

Table 1. Wetland statistics in the 116 11-km2 squares. Estimated wetland area as a proportion of the terrestrial land area.

**Elevation zone**

Fig. 4. Estimated total areas of wetlands in the five Elevation zones: Arctic/Alpine, Northern boreal, Upper middle boreal, Lower middle boreal, and Coastal boreal. Error bars indicate standard errors.

#### **4.1.1 Variation and adequacy of sample**

There was high variation among squares owing in part to the variation of wetlands within the zones, and also to the variation in numbers of squares amongst the zones. Relative standard error (Standard error divided by the estimate) indicates the quality of the estimate and the adequacy of sample (Table 1).

The lowest numbers for relative standard error for the wetland area estimate were obtained for the Northern and Upper middle boreal zones, indicating an adequate sample size of 32

Main Ecosystem Characteristics and Distribution of Wetlands

**4.3 Relation of wetland area to height (elevation) above sea level** 

formed by increasingly large sizes of polygons.

elevational variability of the landscapes there.

lower in Lower middle boreal and Coastal boreal (Fig. 3).

**5. Hydrotopographic types distribution and area** 

in Boreal and Alpine Landscapes in Northern Sweden Under Climate Change 203

parts with the greatest amounts of weakly sloping, flat and concave topographies, the wetlands would have had the longest time to develop and form larger homogeneous units. The mean size of polygons in each square is positively correlated with the total wetland area (r = 0.330, P = 0.000, n = 116) which indicates that when there is higher wetland cover, it is

The Elevation Zones reflect a gradient in height above sea level, with the Arctic/Alpine zone having the highest mean height and the Coastal boreal zone having the lowest. Standard deviation shows how the variation in heights within squares varies. Standard deviation values are highest in the Arctic/Alpine zone, and lowest in the Coastal boreal zone (Fig. 2). The three interior zones have standard deviations that are quite similar, increasing only slightly in the Lower middle boreal zone. This indicates a similar degree of

Simple Pearson correlations were done for the mean wetland areas with means of height (above sea level), standard deviation of height, and mean area of polygons per square. Mean wetland area is strongly negatively correlated with mean standard deviation of height (r = - 0.367, P = 0.000, n=116), in other words, the lower the relief within a square, the greater the amount of wetland. Mean wetland area is not significantly correlated with height (r = -0.096, P = 0.305); this is explained by the bell-shaped relationship of wetland area to elevation, low in the Arctic/Alpine, higher in Northern boreal and Upper middle boreal, and dropping

Even though non-significant the mean size of the polygons increased steadily from a low in the Coastal boreal to a high in the Arctic/Alpine zone (Table 2). However, polygon size was not correlated to elevation (r =0.011, P = 0.515). Polygon size was largest at intermediate elevations. In fact we would expect smaller polygons at higher elevations. The high value for polygon size in the Arctic/Alpine zone can be explained by the high variation of height within this zone. In addition to high elevation squares with smaller polygons it also contains

Each polygon was classified into Hydrotopographic wetland types, slightly modified after the wetland types used in the Swedish Wetland Inventory (Gunnarsson & Löfroth, 2009). Of the 18 wetland types (compared with 17 types in Swedish VMI), 14 were recorded in the 116 squares for northern Sweden. There were five types that were relatively common in terms of area and percentage of total wetlands: Flat (topogenous) fen with 44%, Sloping fen with 14%, Flat or weakly raised bog with 12%, String flark fen with 11%, and Non-mire wetland with 6% (Table 3, Fig. 5). Three other types achieved moderately low levels: Limnogenous fen 4%, String mixed mire 5%, and Mosaic mixed mire 3%. The remaining six types were quite uncommon. It is significant that Flat fens are by far the most abundant wetlands. It is also noteworthy that the Non-mire wetlands are among the more common of the wetlands. The wetland types missing from the inventory were Plateau bog, Domed bog, Blanket bog, and Palsa mire. Plateau bogs and Domed bogs are more common in the south of Sweden where there is more heat and more precipitation, allowing these bogs to grow upwards into distinct raised bodies (similar to Finland, Seppä, 1996). Blanket bogs are missing in our

a number of intermediate elevation squares with large mean polygon sizes.

and 23, respectively. The other zones had higher relative standard errors for estimated wetland areas indicating that the numbers of required squares for adequacy exceeded the actual numbers of squares sampled. However, for all 116 squares combined, the relative standard error was below 15%. In other words, for sampling the whole study area the sample was adequate, but for the five individual zones, only the two zones with the highest wetland covers were adequately sampled.

#### **4.2 Sizes of wetland polygons**

Mean number of wetland polygons varied amongst squares and zones (Table 2). The occurrence of wetland polygons was highest in the Upper middle and Northern boreal zones, with 39.6 and 31.5 polygons per km2, respectively. The other three zones had similar number of polygons per km2. The mean areas of individual polygons were largest in the Arctic/Alpine zone and decreased progressively towards the Coastal boreal zone. However, mean polygon size among zones was not significantly different, (ANOVA, F = 0.76 and P = 0.552; Table 2). The smallest wetland polygon was 331 m2, sampled in the Coastal boreal zone. The largest wetland polygon was 428620 m2 sampled in the Arctic/Alpine zone.

The number and area statistics for wetland polygons (Table 2) give an estimate of the configuration of the landscape with respect to wetland components, and hence are indicative for landscape biodiversity. The complexity of wetlands in the landscape, as indicated by the size of uniform areas for mapped polygons, is least in the Arctic/Alpine zone, and is greater progressing towards the Coastal boreal zone. These results indicate that the landscape configuration along the coast generally provides the highest potential for wetland biodiversity, and that the potential decreases westwards towards the alpine zone.

It is known that the landscape near the Coastal boreal zone is more influenced by man's activities, and is ecologically younger owing to the ongoing glacio-isostatic land uplift (e.g. Svensson, 2002), consisting of patches of wetlands in earlier successional stages. The polygon sizes also indicate smaller areas (Table 2). These small patches tend to combine and merge together over time, such that in the oldest exposed parts of Northern Sweden, and the


Table 2. Polygon statistics in the data set of 116 squares. Given are the mean, minimum and maximum number of polygons in a square, estimates of the number of wetland polygons per km2, and the mean area per polygon in m2, in the five Elevation Zones and the total study area.

and 23, respectively. The other zones had higher relative standard errors for estimated wetland areas indicating that the numbers of required squares for adequacy exceeded the actual numbers of squares sampled. However, for all 116 squares combined, the relative standard error was below 15%. In other words, for sampling the whole study area the sample was adequate, but for the five individual zones, only the two zones with the highest

Mean number of wetland polygons varied amongst squares and zones (Table 2). The occurrence of wetland polygons was highest in the Upper middle and Northern boreal zones, with 39.6 and 31.5 polygons per km2, respectively. The other three zones had similar number of polygons per km2. The mean areas of individual polygons were largest in the Arctic/Alpine zone and decreased progressively towards the Coastal boreal zone. However, mean polygon size among zones was not significantly different, (ANOVA, F = 0.76 and P = 0.552; Table 2). The smallest wetland polygon was 331 m2, sampled in the Coastal boreal zone. The largest wetland polygon was 428620 m2 sampled in the

The number and area statistics for wetland polygons (Table 2) give an estimate of the configuration of the landscape with respect to wetland components, and hence are indicative for landscape biodiversity. The complexity of wetlands in the landscape, as indicated by the size of uniform areas for mapped polygons, is least in the Arctic/Alpine zone, and is greater progressing towards the Coastal boreal zone. These results indicate that the landscape configuration along the coast generally provides the highest potential for wetland biodiversity, and that the potential decreases westwards towards the alpine

It is known that the landscape near the Coastal boreal zone is more influenced by man's activities, and is ecologically younger owing to the ongoing glacio-isostatic land uplift (e.g. Svensson, 2002), consisting of patches of wetlands in earlier successional stages. The polygon sizes also indicate smaller areas (Table 2). These small patches tend to combine and merge together over time, such that in the oldest exposed parts of Northern Sweden, and the

1 Arctic/Alpine 19.3 (3.1) 1 73 11644 (3856) 2 Northern boreal 31.5 (3.7) 1 105 9208 (1233) 3 Upper middle boreal 39.6 (4.9) 4 91 7160 (821) 4 Lower middle boreal 22.0 (5.3) 1 55 6169 (1502) 5 Coastal boreal 25.3 (4.7) 1 47 6119 (763) Total 28.9 (2.1) 1 105 8422 (957) Table 2. Polygon statistics in the data set of 116 squares. Given are the mean, minimum and maximum number of polygons in a square, estimates of the number of wetland polygons per km2, and the mean area per polygon in m2, in the five Elevation Zones and the total

Minimum number of polygons per sample

Maximum number of polygons per sample Mean wetland polygon area (m2) with standard errors

wetland polygons per km2 with standard

errors

wetland covers were adequately sampled.

Elevation zone Mean number of

**4.2 Sizes of wetland polygons** 

Arctic/Alpine zone.

zone.

study area.

parts with the greatest amounts of weakly sloping, flat and concave topographies, the wetlands would have had the longest time to develop and form larger homogeneous units. The mean size of polygons in each square is positively correlated with the total wetland area (r = 0.330, P = 0.000, n = 116) which indicates that when there is higher wetland cover, it is formed by increasingly large sizes of polygons.

#### **4.3 Relation of wetland area to height (elevation) above sea level**

The Elevation Zones reflect a gradient in height above sea level, with the Arctic/Alpine zone having the highest mean height and the Coastal boreal zone having the lowest. Standard deviation shows how the variation in heights within squares varies. Standard deviation values are highest in the Arctic/Alpine zone, and lowest in the Coastal boreal zone (Fig. 2). The three interior zones have standard deviations that are quite similar, increasing only slightly in the Lower middle boreal zone. This indicates a similar degree of elevational variability of the landscapes there.

Simple Pearson correlations were done for the mean wetland areas with means of height (above sea level), standard deviation of height, and mean area of polygons per square. Mean wetland area is strongly negatively correlated with mean standard deviation of height (r = - 0.367, P = 0.000, n=116), in other words, the lower the relief within a square, the greater the amount of wetland. Mean wetland area is not significantly correlated with height (r = -0.096, P = 0.305); this is explained by the bell-shaped relationship of wetland area to elevation, low in the Arctic/Alpine, higher in Northern boreal and Upper middle boreal, and dropping lower in Lower middle boreal and Coastal boreal (Fig. 3).

Even though non-significant the mean size of the polygons increased steadily from a low in the Coastal boreal to a high in the Arctic/Alpine zone (Table 2). However, polygon size was not correlated to elevation (r =0.011, P = 0.515). Polygon size was largest at intermediate elevations. In fact we would expect smaller polygons at higher elevations. The high value for polygon size in the Arctic/Alpine zone can be explained by the high variation of height within this zone. In addition to high elevation squares with smaller polygons it also contains a number of intermediate elevation squares with large mean polygon sizes.
