The use of quantitative models to assess long-term climate–vegetation dynamics – A case study from the northern Scandinavian Mountains

Abstract

Quantitative vegetation models combined with fossil pollen records have large potentials for assessing long-term vegetation dynamics. In this study, vegetation cover as a function of July temperatures was modelled for the Dividalen valley area situated in the northern Scandinavian Mountains. Site-specific pollen deposition values of the two dominating tree species pine and birch were simulated and compared with empirical fossil pollen values. The applied vegetation model could not explain the forest dynamics prior to 7400 cal. BP, although after this date, the forest seemed to have received a modern structure, and summer temperatures could potentially explain the long-term variations in forest cover. The most extensive forest cover occurred c. 7400–3800 cal. BP, corresponding to July temperatures of 0.5–1°C above the present. The approach proved sensitive to both site location and data quality. Careful site-selection is of importance for the application, and pollen accumulation rates are to be preferred as pollen percentage values proved problematic.

Keywords

climate change forest dynamics Holocene vegetation change pollen accumulation rate quantitative vegetation model R-value model Scandinavian Mountains

Introduction

Many processes in nature operate on time-scales well beyond what is practically feasible to cover with instrumental measurements. Most obvious is perhaps forest dynamics where trees grow centuries old and regenerate in pulses hundreds of years apart (Zackrisson et al., 1995). In order to understand such long-term dynamics, one has to rely on reconstructions of past vegetation. As written records are scarce and limited, such reconstructions in most cases have to be based on different proxy data from natural archives, such as pollen, macrofossils, tree-rings and environmental DNA. Of these, pollen is the most widely applied, as the pollen grains are produced in vast numbers, easily dispersed and very resilient to decomposition, and on a landscape level, it is the only viable proxy for past vegetation cover.

The amount of pollen deposited in a natural archive as a lake or bog is affected by a multitude of factors such as how much pollen a species produce and how it is dispersed (see Jackson, 1994; Jackson and Lyford, 1999; Prentice, 1985; Tauber, 1965). Given this complexity, most pollen data are traditionally only interpreted in a qualitative (presence of a taxon) and relative way (increase or decrease of a taxon). Quantitative reconstructions, that is, absolute estimates of vegetation abundance, would, on the other hand, provide more detailed information about past vegetation and be more comparable to data expected from modern observations which in turn would allow broader applications. Given this, there has been extensive investigations in how deposited pollen is quantitatively reflected by the surrounding vegetation (e.g. Broström et al., 2008; Sjögren et al., 2008). Once the pollen–vegetation relationship is established, it is possible to quantitatively reconstruct past vegetation.

There are two main approaches to quantitative reconstructions based on pollen data. In the direct approach, the reconstruction is based on the data itself, with as little as possible assumptions according to the past landscape structure (e.g. Marquer et al., 2014; Mazier et al., 2012). The alternative is a more indirect approach, where the base of the reconstruction is a vegetation model rather than the palaeodata itself (e.g. Gaillard et al., 2008, 2010; Sjögren et al., 2014). In the latter approach, the pollen data can be used to validate and/or evaluate the model, but it is also possible to let the proxy data determine model parameter values and thus truly provide a model-based quantitative vegetation reconstruction. As the vegetation reconstruction is model based, it will invariably be dependent on the same parameters as the model and thus, for good and bad, be a measurement of how well these can explain past changes in vegetation.

Despite the considerable work done on the pollen – vegetation relationship and quantitative reconstruction methods – many uncertainties remain (see Matthias and Giesecke, 2014; Theuerkauf et al., 2013), which illustrates the complexity of the issue. Further testing, development and evaluation of quantitative reconstruction methods are thus essential.

The main goal of the present investigation is to test the applicability and performance of a model-based vegetation reconstruction on a northern boreal vegetation data set. Sub-goals are as follows: (1) test the applicability and reliability of the approach on four different palaeo-sites in a transect across today’s tree line forming birch belt, (2) test the applicability and reliability of the approach to pollen percentage versus pollen accumulation data and (3) determine whether summer temperatures can explain the vegetation structure and changes in vegetation cover of northern boreal forests during the Holocene.

Method

Study area

The investigation area is situated in Dividalen valley in the northern Scandinavian Mountains, close to the border between Norway and Sweden (see Figures 1 and 2). These northern boreal forests can be considered perfect for the construction, application, testing and evaluation of such quantitative vegetation models as the composition and main dynamics are assumedly relatively simple. Birch and pine strongly dominate the forest, and their distribution is, given the altitude and northern latitude, likely dependent on summer temperatures (e.g. Gehrig-Fasel et al., 2008; Holtmeier and Broll, 2005). In addition, there exist four pollen records from the Dividalen area situated in a transect from the upper pine forest through the mountain birch belt into the open heath vegetation (Jensen and Vorren, 2008; Sjögren and Kirchhefer, 2012; see Figure 1).

Figure 1.

Site map over Dividalen valley. In the inserted map of Fennoscandia, the investigation area is marked with a star.

Figure 2.

Modelled vegetation cover at increasing temperatures above the assumed present climatic equilibrium with vegetation. Dashed lines (upper left map) show the outline of the investigation area. SKT: Lake Skrubbtjern; JRT: Lake Jervtjern; GPT: Lake Gauptjern; AVM: Anjavassmyra Mire..

The valley itself is north-south orientated and sheltered from the humid westerlies by mountains, resulting in a sub-continental, intramontane climate, with an annual precipitation in the valley bottom of only 282 mm. Mean annual temperature is +0.8°C, with a mean summer temperature of 12.8°C and mean winter temperature of −9.4°C (mean values for 1961–1990 for weather station 89950 ‘Dividalen’, longitude: 19°47″E, latitude: 68°43″N, 228 m a.s.l., Norwegian Meteorological Institute, met.no). Vegetation in the lower northern end of the valley is dominated by pine forests which gradually give way to birch woodland and eventually open heath further up the valley and the surrounding mountain slopes.

Modelling vegetation

A Landsat Thematic Mapper (TM)–based vegetation map from eastern part of Troms county (Johansen et al., 1995) was improved by the use of a digital terrain model and mire and forest mask (N50 data from the Norwegian Mapping authority). It was then compared with vegetation units previously grouped according to their minimum temperature requirements in terms of July temperature mapped on Varangerhalvøya peninsula (Karlsen and Elvebakk, 2003; Karlsen et al., 2005). Most of the mapped vegetation types in the Dividalen area, except pine forests, had related types on Varangerhalvøya. Due to the high variation in species composition and their fine mosaic structures, not easily detected by Landsat TM satellite data, mires were excluded from the study. Subsequently, the mapped vegetation types in Dividalen were grouped according to their minimum temperature preferences in steps of 1°C differences in July temperature. Also, the pine forest types were forced into temperature and habitat groups.

After the grouping of all vegetation types, except mires, according to minimum temperature and habitat preferences, the altitude distribution of each vegetation type in relation to the closest tree line today was analysed. The current tree line is formed by birch and varies across the study area, from less than 500 m altitude in north-facing slopes in the western parts to more than 650 m in south-facing slopes and in the eastern flat parts. Based on the forest data from the mapping authority (N50 data), polylines of the upper tree line were extracted. Outliers where the tree line was low due to supposed anthropogenic reasons were subjectively removed. The upper tree line was then compared with a terrain model, and a raster was created where the pixels had altitude information for the upper tree line.

The final step was to simulate past vegetation according to higher July temperature. The simulation was done by Python scripts in the ArcGIS software. It was assumed that the current positions of the vegetation types and the tree line are in balance with climatic factors. A second assumption was that a 1°C increase in July temperature corresponds with 171 m altitude increase in tree line, based on a lapse rate of 0.585°C/100 m (Tveito et al., 2000). For increases in July temperatures of 0.5°C, 1.5°C and 2.0°C, a 85.5-, 256.5- and 342-m rise in the tree line was used. It was then assumed that current temperature-demanding (thermophilous) vegetation types will expand into colder areas and replace less temperature-demanding vegetation types with equal habitat preferences.

Pollen record

Pollen records from four sites within the investigation area were used: Lake Skrubbtjern, Lake Jervtjern, Lake Gauptjern and Anjavassmyra Mire, placed more or less in a transect from open alpine vegetation to just within the pine forest (see Table 1 and Figure 1).

Table 1.

Modern pollen percentages and R_rel values from the investigation sites.

Site	Modern pine pollen (%)	Site-specific R_rel		Altitude (m a.s.l.)	Size (ha)	Local vegetation
		Local weighting birch × 1	Local weighting birch × 2
Lake Skrubbtjern	50.1	3.8	2.7	790	2	Open
Lake Jervtjern	32.2	2.3	2.2	548	1.5	Open birch woodland
Lake Gauptjern	11.6	0.6	0.7	400	1	Birch forest
Anjavassmyra Mire	66.7	3.0	2.6	370	0.5	Open pine forest

Modern pine pollen percentages are based on the pine and birch sum. R_rel is calculated using log10^1/3 weighting with superimposed linear proportional weighting of birch. R_rel for birch is 1.

Given the low modern abundance in both vegetation and pollen for other taxa than pine and birch, it was not possible to quantify the pollen–vegetation relationship for these, and only the two dominant pollen types are presented (Pinus sylvestris-type and Betula pubescens-type). These two taxa have a very strong dominance in both the modern and fossil assemblages. For Lake Skrubbtjern, Lake Jervtjern, Lake Gauptjern and Anjavassmyra Mire, the modern deposition of these two taxa represent 90%, 83%, 91% and 85% of the total pollen sum, respectively. Average values for all Holocene samples are 76%, 85%, 87% and 94%, respectively. Pine and birch, thus, represent the major changes in the past forest composition. For the full pollen spectra as well as depth–age models, we refer to Jensen and Vorren (2008) and Sjögren and Kirchhefer (2012).

Surface sediments from the lakes were collected with both a Kajak and a Hongve samplers, and the top 3 cm from each core was analysed at 1-cm intervals (Jensen and Vorren, 2008). The modern pollen proportions used here are the average of these six samples. For the mire, the modern pollen values are based on a 15-cm peat monolith treated as a single sample (Sjögren, 2013).

Pollen accumulation rates (PARs) were presented only for Lake Jervtjern, as calculation proved problematic for the other sites and/or they had an, to this approach, unfavourable location. It was not possible to calculate modern PARs from the surface sediments, and the values had to rely on assumed representative parts of the deeper sediments. As it is uncertain how representative these are, as well as the general difficulties related to obtaining reliable PARs, the average values of two different sets of samples were used. For the first set, the top five fossil samples (1000–400 cal. yr) were used, providing almost the exact same percentage value as derived from the surface samples (pine 31.3% vs 32.2%). This suggests that the vegetation was similar to the modern for this period. For the second set of samples, the uppermost pollen zone of Jensen and Vorren (2008) was used (J11; 2500–400 cal. BP). Given the Holocene development of the fossil pollen record, it does not seem likely that the modern PARs should be higher than provided by the first set or lower than provided by the second set, and thus, they provide a reasonable upper and lower limits of potential modern PAR values.

Pollen–vegetation relationship

The pollen–vegetation relationship is based on the R-value model (Davis, 1963), which is as follows: R_a = Species_a pollen percentage/Species_a vegetation percentage. As the R-value is based on percentage values, it will vary between investigations, although the ratio between the R-values of two species is constant and referred to as the R_rel (Andersen, 1970). When absolute pollen and vegetation data are used instead of the relative data, we use the term R_abs (in grains/cm²/yr).

As nearby vegetation contributes more pollen than distant, it needs to be weighted accordingly. In the present investigation, the vegetation is measured in rings with 10^1/3 logarithmic increasing radius (15 rings with an outer radius of 10 m–464 km), where each ring is assumed to contribute equally to the pollen assemblage (Sjögren, 2013; Sjögren et al., 2008; Van der Knaap et al., 2001). This mathematical weighting gives similar results as Sutton’s (1953) empirically derived equation if a voluminous pollen source, moderate wind speed and limited maximum dispersal are assumed (Sjögren, 2013), and the logarithmic weighting applied here provides the simpler alternative. As the choice of distance weighting, or dispersal function, is important for the results (Theuerkauf et al., 2013), a second alternative weighting is provided for comparison (percentage data only). Here, a proportional weighting of birch is superimposed on the logarithmic (Sjögren et al., 2015), with an increased weighting of the innermost ring of +100% followed by a proportional decline in the weighting of each ring resulting in a weighting of the outermost ring of −100%. This follows the general perception among playnologist that birch (and most other trees) disperses its pollen less effectively than pine.

Cover estimates of pine and birch were allocated to each vegetation class, which allowed the calculation of species-specific cover. The species-specific cover of rings 1–12 (outer radius: 10 m–46 km) is entirely based on these data, and areas extending into Sweden were ignored (i.e. the cover percentage of the Norwegian part of the ring is used, 16.6% of ring 12 extended into Sweden). For the more distant rings 13–15 (outer radius: 100–464 km), the vegetation cover extending into Sweden was approximated based on general regional cover data (Forest Statistics, 2012). Slightly more than 1/3 of the area beyond 46 km is situated in Sweden, which makes up 7% of the total weighted area.

Simulating pollen deposition

For each of the modelled landscapes (for July temperatures of +0.5°C, +1.0°C, +1.5°C and +2.0°C), proportional pollen deposition at the four localities was simulated. For the purpose of calculating simulated pollen deposition, a modelled area of 46 km around the localities was used. The more distant vegetation change (rings 13–15 with an outer radius of 100–464 km) was not modelled; instead, a conservative cumulative increase of birch and pine cover of 12.5%/0.5°C was assumed (resulting in a total increase of 60% at +2.0°C). This is a more conservative increase compared to the modelled area (see below), and the reason for this is that sea (in the west and north) and forest (in the east and south) cover much of the more distant areas. Thus, potential forest expansion is much more limited there than within the mountainous investigation area.

Linking simulated and fossil pollen deposition

Based on the simulation results for the four localities and available fossil pollen data, only Lake Jervtjern was selected for further analysis. The main advantages of the Lake Jervtjern record are as follows: (1) it is situated in the upper birch belt which gives a clear response in simulated pollen to the modelled vegetation change for the temperature interval of interest; and (2) it has reasonable, and thus probably unbiased, modern as well as past pollen percentages and PAR values. By assuming linear interpolation between the simulated pollen values at different temperatures intervals, it was possible to assign each fossil pollen sample to a modelled vegetation cover, that is, when an empirical fossil pollen sample had the same PAR or percentage value as that of a simulated pollen sample. As the simulated birch PAR values peaked between +0.5°C and +1.0°C (see below) two problems arose: (1) high birch PAR values could only be assigned to the temperature interval +0.5°C–1.0°C and not to a specific temperature. For these samples, the mid-point of the temperature interval, that is, +0.75°C, was used in order to simplify presentation. (2) Moderate birch PAR values could potentially be assigned to two different temperatures, either below or above the peak. In such cases, the values have consequently been assigned to the lower temperature. In the by far most instances, this is evident as the values are part of trends, although false assignments are theoretically possible.

The quantitative vegetation model did not consider temperatures below modern. In order to improve the presentation of the modelled values as well as provide approximate values for temperatures below the present day, a very simplified model of forest cover was applied to these samples, based on the lapse rate of 171 m/1°C. For −0.5°C, it is assumed that there is no birch forest within 200 m and no pine forest within 1000 m from Lake Jervtjern. Regional birch cover (until ring 12, 100 km) is assumed to decline with 10% and regional pine cover with 40%. Extra regional cover (rings 13–15, 100–464 km) of pine is assumed to decline with 20%. At −1°C, all values are assumed doubled. Empirical PAR and pine percentage values below modelled −1°C values were assigned to −1°C. All values for temperatures below the present should only be considered approximate estimates.

Results

Modelled landscapes

Vegetation cover maps modelled for July temperatures of +0.5°C, +1.0°C, +1.5°C and 2.0°C are provided in Figure 2. Within the investigated area (see Figure 2), the modelled increase in pine is from 1.1% total cover at modern temperatures to 14.1% at +2.0°C. For mixed birch–pine forest, the change is from 2.0% to 20.0% and for birch forest from 14.2% to 36.7%. The largest area decline is seen in exposed heaths (from 15.4% to 5.2%), grass heaths (from 13.0% to 3.3%) and sparse vegetated mid-alpine vegetation types (from 17.1% to 0.6%). Thus, the model predicts a fully different landscape at a July temperature increase of 2°C, from a predominantly open landscape (17.3% forest cover) to a forested landscape (70.8% forest cover). This change is even more profound considering that shadows are not mapped (5.6%), wetlands are not modelled (1.4%) and there is some water (2.8%), leaving less than one-fifth of the potentially vegetated areas with open vegetation types at July temperatures of 2°C above modern.

R-values

The pine R_rel values calculated for the four localities vary substantially, from 0.6 to 3.8 (birch R_rel = 1; see Table 1). If local weighting is increasing for birch, these differences are reduced, with all localities without Gauptjern producing values in the 2.2–2.7 interval. This might thus be a more accurate dispersal function, although the low number of sites makes conclusions tentative. Lake Gauptjern still shows a very low pine R_rel value of 0.7, suggesting that the applied dispersal model and/or function may not be accurate for this locality.

For Lake Jervtjern, the pine R_abs values of 7700 and 10,800 grains/cm²/yr are similar to what is derived from mires in the region (regional mean 11,800 ± 2600 grains/cm²/yr; Sjögren, 2013), while the birch R_abs values of 3500 and 4700 are considerably higher (regional mean 2100 ± 300 grains/cm²/yr; Sjögren, 2013; see Table 2). A nearby comparison between lake and mire PARs (Jensen et al., 2002) shows that the lake PARs are much higher than the mire PARs, and perhaps more important for the present result; the relative representation of pine is twice as high in the peat as the lake sediments. Thus, the R_abs values seem to be reasonable in comparison with earlier results.

Table 2.

Pollen depositional and productivity values used in the investigation.

Parameter	Modern^a pollen values	PZ1 pollen values
Pine %	32.2	29.9
Pine PAR	555	399
Birch PAR	1184	899
Pine R_rel	2.3	2.1
Pine R_abs	10,800	7700
Birch R_abs	4700	3500

PAR: pollen accumulation rate.

PZ1 pollen values are the average percentage and PAR values in the uppermost pollen zone in Jensen and Vorren (2008) 2500–400 cal. BP. PAR values are in grains deposited per square centimetre per year. R_abs values are in grains released per square centimetre taxon cover per year.

PAR values based on the five uppermost pollen samples (1000–400 cal. BP), with a mean proportional value almost identical to the modern.

Simulated and fossil relative pollen deposition

The simulated pollen depositions show that the four sites respond very differently to the modelled changes in vegetation (see Figure 3). For the uppermost site, Lake Skrubbtjern, the simulated response curve of pine pollen percentages to temperature is almost flat and temporarily even decline. The reason for this is that pollen from local encroachment of birch outweighs the increase in more distant pine forest. Lake Jervtjern and Lake Gauptjern positioned in the birch belt show similar curves – first a rather steep increase in pine pollen percentages with increasing temperatures after which the curves flattens out. This shift occurs when the local forest changes from birch to pine, which for modelled vegetation is around +1.5°C at Lake Jervtjern and at +0.5°C at Lake Gauptjern. Anjavassmyra Mire only shows a very weak increase in pine percentages with increasing temperatures. The reason for this is that today, it is situated just within the pine forest, so there is no local vegetation change within the modelled temperature interval.

Figure 3.

Simulated relationship between modelled vegetation and relative pollen deposition of pine at the four investigated sites. The temperature indicates the corresponding temperature to the modelled vegetation cover (see Figure 2) resulting in the simulated pollen deposition. Dashed line shows an alternative relationship if increased local distance weighting is assigned to birch (see text).

Fossil pine pollen percentage records

In the fossil pollen records, the development throughout the Holocene for the three lowermost localities is very similar, with a clear peak in pine pollen percentages around 8000 cal. BP (see Figure 4). The pollen record from the highest locality, Lake Skrubbtjern, shows an entirely different development. This could very well be explained by the simulated difference in response to climate change (see Figure 3). All pollen records with exception of Lake Gauptjern show pine pollen percentage values that are mostly below modern. Given that Lake Gauptjern also has very low pine R_rel values compared with the other localities and with other investigations (e.g. Andersen, 1970; Sjögren, 2013), the modern pollen measurements and/or the applied pollen dispersal model might not be representative for this fossil pollen record.

Figure 4.

Pine pollen expressed as percentage of the sum of pine (Pinus sylvestris-type) and birch (Betula pubescens-type) pollen at Lake Skrubbtjern, Lake Jervtjern, Lake Gauptjern and Anjavassmyra Mire. Five-sample running average was applied to the Gauptjern, Jervtjern and Skrubbtjern curves. Black field shows pollen values above modern values.

Modelled past vegetation

The simulated pollen depositions at Lake Jervtjern are presented in Figure 5 and PARs from Lake Jervtjern in Figure 6. The modelled vegetation cover expressed in ΔJuly temperatures corresponding to the fossil pollen samples is presented in Figure 7, while the spatial distribution of the vegetation is provided in Figure 2. Interpretations consider the results from both the alternative R_abs applied, using the interval of temperatures. Seen individually, the trends are identical, but the R values based on the uppermost pollen zone (J11) of Jensen and Vorren (2008) indicate somewhat higher temperatures, up to +0.3°C.

Figure 5.

Simulated pine pollen percentages (in % of pine + birch), and pine and birch PARs (in grains/cm²/yr) at Lake Jervtjern for modelled vegetation at increasing temperatures in +0.5°C steps. The solid lines are based on modern percentages and assumed modern PARs, and the dashed lines are based on pollen values from the uppermost pollen zone (Jensen and Vorren, 2008), that is, 2500–400 cal. BP.

Figure 6.

Pollen accumulation rates (in grains/cm²/yr) for pine (Pinus sylvestris-type) and birch (Betula pubescens-type) from Lake Jervtjern. Five-sample running average was applied. Black infill shows PAR values above modern (i.e. 1000–400 cal. BP).

Figure 7.

Inferred ∆ July temperatures are based on the intercept between model-based simulated pollen deposition and fossil empirical pollen deposition at Lake Jervtjern. Results based on modern pollen values are shown in black. Results based on uppermost pollen zone (Jensen and Vorren, 2008) values are shown in grey. Values below zero shown by dashed lines are approximations. Five-sample running average was applied to all curves.

Pine PARs indicate that the largest extension of pine forest occurred in the time-span 8200–3800 cal. BP, corresponding to a modelled vegetation at July temperatures c. 0.5°C above the present. Birch PARs show similar development from 7400 cal. BP onwards but with slightly higher maximum forest cover corresponding to a modelled vegetation at +0.5–1.0°C. Prior to 7400 cal. BP, corresponding temperatures for pine and birch show opposite trends. From 10,000 to 9000 cal. BP, birch PARs correspond to above modern temperatures, while pine PARs correspond to considerably lower temperatures. In the period 8200–7400 cal. BP, the situation is the opposite, where pine PARs according to the vegetation model equal July temperatures c. +0.5°C above the present, while birch PARs equal July temperatures considerably below the present. Pine percentages indicate maximum temperatures in the same interval, 8200–7400 cal. BP, which is clearly an artefact of low birch PAR values. Otherwise, pine percentages indicate a mid-Holocene maximum forest cover similar to or slightly above the present, which decrease c. 4100 cal. BP.

Thus, all three semi-independent variables show the same general trends from c. 7400 cal. BP onwards but strongly diverge in the previous time period (10,000–7400 cal. BP), which demonstrates that vegetation dynamics were different than the modelled prior to 7400 cal. BP. Maximum forest cover occurs in the time period 7400–3800 cal. BP and corresponds to modelled vegetation at July temperatures of +0.5–1.0°C, equivalent to an altitudinal increase of the forest and tree line of 85–170 m.

Discussion

Internal consistency of results

Pollen percentage and PAR data gave quite different results for the period 8200–7400 cal. BP, and it is obvious that pollen percentage data might give unreliable results using the present approach and should only be interpreted together with PAR data. Change in cover from different taxa may give no, little or even negative response in the pollen percentages. PAR data are here much more informative as they provide independent assessments of each taxon. The problem with PARs is that they are more difficult to obtain than pollen percentage data and are prone to error if the sedimentation regime changes. Pollen percentage data may here be useful to identify potential bias in the PAR record, as it is independent of sedimentation rate.

Based on the PAR data, the two variables tested, pine and birch covers, were internally consistent with the model for the middle and late part of the Holocene. The results showed that variations in summer temperatures could explain past changes in forest cover after 7400 cal. BP, that is, modelled past birch and pine covers independently indicated more or less the same variations in temperature. In the time span before that, 10,000–7400 cal. BP, the model was unable to explain the changes in birch and pine covers, that is, modelled past birch and pine covers indicated dissimilar variations in temperature. It is possible that other factors, for example, migration lag, soil maturation and precipitation, affected the forest structure to a larger degree in this time period.

Comparison with climate and vegetation reconstructions

This region of the Scandinavian Mountains has been thoroughly investigated concerning both climate and vegetation changes throughout the Holocene (e.g. Barnekow, 1999; Berglund et al., 1996; Bigler et al., 2002, 2006; Bjune et al., 2004; Grudd et al., 2002; Seppä and Birks, 2002; Sonesson, 1968), including investigations from the valley itself (Jensen, 2007; Jensen and Vorren, 2008; Sjögren and Kirchhefer, 2012).

Jensen (2007) concluded that the forest composition in the investigation area prior to 9200 cal. BP might be explained by the good expansion ability of birch and migration lag for pine (and other species) rather than a climate favouring birch compared to pine, while the forest composition 9200–7400 cal. BP could have been affected by variations in humidity in addition to summer temperatures. This is consistent with the results of the present investigation in the sense that other parameters than July temperature seem to have had a defining influence on the forest cover and distribution prior to 7400 cal. BP.

After 7400 cal. BP, both pine and birch PARs respond in a similar way to changes in July temperatures (according to the model), and the relative trends in forest cover are consistent with reconstructed climate and vegetation changes in the region. The present results suggest a maximum forest cover c. 7400–3800 cal. BP equivalent to temperatures of 0.5–1°C above the present (see Figure 2 for spatially explicit cover). Climatic reconstructions based on transfer functions of pollen assemblages show much higher July temperatures during the Holocene thermal maximum (varying between investigations but occurring within the time frame 8500–4500 cal. BP). In Lake Jervtjern itself, temperatures reach above +3°C compared to the present, although only +1.5–3°C is reached in the nearby Lake Gauptjern (Jensen and Vorren, 2008). Less than a 100 km to the north-east summer temperature reconstruction at Lake Dalmutladdo shows temperatures of +2°C (Bjune et al., 2004) and at Lake Toskaljavri of +1.5–2°C (Seppä and Birks, 2002). Less than a 100 km to the south-west, climatic reconstructions based on pollen, chironomids and diatoms suggest temperatures of +1–2°C in the millennia around 7000 cal. BP (Bigler et al., 2002, 2006). So, there is a discrepancy between the present investigation that suggests that the maximum forest cover during the Holocene was equivalent to July temperatures of 0.5–1.0°C above the present and most climate reconstructions in the region which suggest temperatures around 1.5–2.0°C above the present. It is, thus, possible that factors other than summer temperature limited the maximum extension of forest cover in this period, for example, soil conditions, precipitation, wind, snow cover and/or browsing (e.g. Henne et al., 2011).

Tree line reconstructions based on macrofossils also suggest that the tree line of pine was 300 m higher than the present (Bjune et al., 2004; Jensen and Vorren, 2008), considerably more than the approximately 100–150 m suggested by the present results. One possible explanation of this discrepancy is that altitudinal forest line was lower in relation to the tree line than it is today, that is, there was a broader belt of sparse, low-productive scattered trees above the forest line.

The results have not been adjusted according to land uplift, and the altitude was c. 100 m lower 9000 cal. BP and then rapidly increased (Møller and Holmeslet, 2002). Adjusted for lapse rate along the period with maximum forest covers at 8000–4000 cal. BP would equal a temperature of 0.4–0.2°C lower, although increased oceanity could have had an opposite effect. This gives an additional uncertainty when compared to large-scale climate change but has little effect on the comparison above, as these have not been adjusted and can be assumed to have been affected in a similar way.

Evaluation of the modelling approach

The case study yielded interesting results concerning long-term forest dynamics. Different parameter values were tested without changing the general conclusions, and the method as applied on the present data set can be considered robust. So, for assessing past vegetation dynamics and testing environmental parameters, the approach proved useful. Modelled pollen deposition at different localities also reduced the risk of errors in interpretation of the fossil pollen records, as it showed the importance and effect of location on the pollen records.

The application of the approach was, on the other hand, more difficult than expected. Out of the four potentially useful fossil pollen records only one provided reliable results and then only based on the PAR record. If this is a general rule, then application would be severely restricted by the site location and quality of the fossil records. Uncertainties considering many of the parameter values also make it difficult to provide absolute values in detail, making it necessary to present ranges rather than values, although this is likely to improve in future investigations. Such improvements could be to model a larger area, using a wider temperature span, finer temperature increments in modelled vegetation, improved cover estimates, locally site-specific adjusted models, better modern PAR values, more accurate dispersal functions and fossil pollen data from mires.

One potentially important aspect that has not been considered in the present approach is the direct effect of climate and altitude on pollen productivity, which has been shown to decline with temperatures (Hicks, 2006). This may bias the result as the pollen production of high altitude vegetation is overestimated compared to vegetation at lower altitude. In addition, this effect might vary over time as the climate change. If the present vegetation distribution lags behind the recent warming, it is possible that pollen productivity remains high close to the present tree line, while it is much lower when the vegetation reaches equilibrium with the climate. This may explain the discrepancy between the present results and temperature reconstructions of the region. Further research is required to clarify and in necessary correct for this effect.

It could be noted, however, that pollen-based climatic reconstructions further south in Fennoscandia also show a distinct Holocene thermal maximum at c. 8000–4000 cal. BP, with a summer temperature (JJA) at 6000 cal. BP of +0.7°C compared to 500 cal. BP (Salonen et al., 2014), that is, more or less identical to the present results. In this investigation, the climatic reconstructions were based on boosted regression trees (BRT) rather than the more commonly used weighted average (WA). BRT resulted in 0–1°C lower reconstructed temperatures than WA, and if this is generally valid, it could explain some of the discrepancy seen between the present investigation and regional climatic reconstructions.

Conclusion

The site location of a fossil pollen record largely defines how past changes in forest cover are perceived in the pollen data, and there is a risk that especially pollen percentage data may be misinterpreted. Modelling of theoretical site-specific pollen deposition may reduce the risk of such misinterpretations;

Application of the modelling approach to fossil pollen records proved more difficult than anticipated. To ensure good results, high-quality modern and fossil PAR records should be available. It is possible that mires, with their simpler deposition environment, perform better than lakes;

The approach was successful in assessing past vegetation dynamics, and the relative changes in past vegetation cover (or corresponding temperature) were similar compared to other investigations. Because of the uncertainty of many parameter values, only ranges could be presented. Even so, the present result suggests less forest cover than could be deduced from other climate and vegetation reconstructions in the region, suggesting that that the maximum extension of forest cover is limited by other factors than July temperature;

According to the results, the forest gained its modern structure c. 7400 cal. BP. After this, the forest cover varied with summer temperature, with a maximum forest cover in the period 7400–3800 cal. BP corresponding to July temperatures of 0.5–1°C above the present. Spatial distribution of the vegetation is provided in Figure 2 (modelled vegetation at +0.5°C and +1°C).

Footnotes

Acknowledgements

We would like to thank Professor Arve Elvebakk, University of Tromsø, for encouragement and valuable suggestions. Three anonymous reviewers contributed with valuable and comprehensive comments. The investigation is a contribution to the DYLAN (DYnamic LANdscapes) project.

Funding

This work was supported by the Research Council of Norway (grant or award number: 190044/S30).

References

Andersen

(1970) The relative pollen productivity and pollen representativity of North European trees and correction factors for tree pollen spectra. Danmarks Geologiske Undersøgelser (Series II) 96: 1–99.

Barnekow

(1999) Holocene tree-line dynamics and inferred climatic change in the Abisko area, northern Sweden, based on macrofossils and pollen records. The Holocene 9: 253–265.

Berglund

Barnekow

Hammarlund

. (1996) Holocene forest dynamics and climate change in the Abisko area, northern Sweden – The Sonesson model of vegetation history reconsidered and confirmed. Ecological Bulletins 45: 15–30.

Bigler

Barnekow

Heinrich

. (2006) Holocene environmental history of Lake Vuolep Njakajaure (Abisko National Park, northern Sweden) reconstructed using biological proxy indicators. Vegetation History and Archaeobotany 15: 309–320.

Bigler

Larocque

Peglar

. (2002) Quantitative multiproxy assessment of long-term patterns of Holocene environmental change from a small lake near Abisko, northern Sweden. The Holocene 12: 481–496.

Bjune

Birks

HJB

Seppä

(2004) Holocene vegetation and climate history on a continental-oceanic transect in northern Fennoscandia based on pollen and plant macrofossils. Boreas 33: 211–233.

Broström

Nielsen

Gaillard

M-J

. (2008) Pollen productivity estimates of key European plant taxa for quantitative reconstructions of past vegetation: A review. Vegetation History and Archaeobotany 17: 461–478.

Davis

(1963) On the theory of pollen analysis. American Journal of Science 261: 897–912.

Forest Statistics (2012) Official Statistics of Sweden/Swedish University of Agricultural Sciences. Umeå: Swedish University of Agricultural Sciences.

10.

Gaillard

M-J

Sugita

Bunting

. (2008) The use of modelling and simulation approach in reconstructing past landscapes from fossil pollen data: a review and results from the POLLANDCAL network. Vegetation History and Archaeobotany 17: 419–443.

11.

Gaillard

M-J

Sugita

Mazier

. (2010) Holocene land-cover reconstructions for studies on land cover-climate feedbacks. Climate of the Past 6: 483–499.

12.

Gehrig-Fasel

Guisan

Zimmerman

(2008) Evaluating thermal treeline indicators based on air and soil temperature using an air-to-soil temperature transfer model. Ecological Modelling 213: 345–355.

13.

Grudd

Briffa

Karlén

. (2002) A 7400-year tree-ring chronology in northern Swedish Lapland: Natural climatic variability expressed on annual to millennial timescales. The Holocene 12: 657–665.

14.

Henne

Elkin

Reineking

. (2011) Did soil development limit spruce (Picea abies) expansion in the Central Alps during the Holocene? Testing a palaeobotanical hypothesis with a dynamic landscape model. Journal of Biogeography 38: 933–949.

15.

Hicks

(2006) When no pollen does not mean no trees. Vegetation History and Archaeobotany 15: 253–261.

16.

Holtmeier

Broll

(2005) Sensitivity and response of northern hemisphere altitudinal and polar treelines to environmental change at landscape and local scales. Global Ecology and Biogeography 14: 395–410.

17.

Jackson

(1994) Pollen and spores in Quaternary lake sediments as sensors of vegetation composition: Theoretical models and empirical evidence. In: Traverse

(ed.) Sedimentation of Organic Particles. Cambridge: Cambridge University Press, pp. 253–286.

18.

Jackson

Lyford

(1999) Pollen dispersal models in Quaternary plant ecology: Assumptions, parameters, and prescriptions. The Botanical Review 65: 39–75.

19.

Jensen

(2007) The dynamics of the Boreal forest-line in response to Holocene climate variability. Palaeobotanical reconstructions from Dividalen, Norway. A dissertation for the degree of Doctor Scientarium (Synthesis 1-51), University of Tromsø.

20.

Jensen

Vorren

K-D

(2008) Holocene vegetation and climate dynamics of the boreal alpine ecotone of northwestern Fennoscandia. Journal of Quaternary Science 23: 719–743.

21.

Jensen

Kunzendorf

Vorren

K-D

(2002) Pollen deposition rates in peat and lake sediments from the Pinus sylvestris L. forest-line ecotone of northern Norway. Review of Palaeobotany and Palynology 121: 113–132.

22.

Johansen

Karlsen

Tømmervik

(1995) Vegetasjonskartlegging og beitevurdering i norsk- svenske konvensjonsområder. Delområder: Indre Troms og Nordre Norrbotten [Vegetation mapping and pasture assessment in Norwegian-Swedish convention areas: Indre Troms and Nordre Norrbotten]. NORUT rapport IT261/2-1995. Tromsø: NORUT, 62 pp.

23.

Karlsen

Elvebakk

(2003) A method using indicator plants to map local climatic variation in the Kangerlussuaq/Scoresby Sund area, East Greenland. Journal of Biogeography 30: 1469–1491.

24.

Karlsen

Elvebakk

Johansen

(2005) A vegetation-based method to map climatic variation in the arctic-boreal transition area of Finnmark, north-easternmost Norway. Journal of Biogeography 32: 1161–1186.

25.

Marquer

Gaillard

M-J

Sugita

. (2014) Holocene changes in vegetation composition in northern Europe: Why quantitative pollen-based vegetation reconstructions matter. Quaternary Science Reviews 90: 199–216.

26.

Matthias

Giesecke

(2014) Insights into pollen source area, transport and deposition from modern pollen accumulation rates in lake sediments. Quaternary Science Reviews 87: 12–23.

27.

Mazier

Nielsen

Broström

. (2012) Signals of tree volume and temperature in a high-resolution record of pollen accumulation rates in northern Finland. Journal of Quaternary Science 27: 564–574.

28.

Møller

Holmeslet

(2002) Havets historie i Fennoskandia og NV Russland. University of Tromsø (Computer program). Available at: http://geo.phys.uit.no/sealev/ (accessed January 2014).

29.

Prentice

(1985) Pollen representation, source area, and basin size: Toward a unified theory of pollen analysis. Quaternary Research 23: 76–86.

30.

Salonen

Luoto

Alenius . (2014) Reconstructing palaeoclimatic variables from fossil pollen using boosted regression trees: Comparison and synthesis with other quantitative reconstruction methods. Quaternary Science Reviews 88: 69–91.

31.

Seppä

Birks

HJB

(2002) Holocene climatic reconstructions from the Fennoscandian tree-line area based on pollen data from Toskaljavri. Quaternary Research 57: 191–199.

32.

Sjögren

(2013) A pilot study on absolute pollen productivity estimates from mires in Norway. The Holocene 23: 1487–1492.

33.

Sjögren

Kirchhefer

(2012) Historical legacy of the old-growth pine forest in Dividalen, northern Scandes. International Journal of Biodiversity Science, Ecosystem Services & Management 8: 338–350.

34.

Sjögren

Solem

Stenvik

. (2014) Quantification of past plant abundances based on the R-value model – Exemplified by an impact assessment of pre-historic iron production and summer farming in Budalen, central Norway. The Holocene 24: 454–465.

35.

Sjögren

Van der Knaap

Huusko

. (2008) Pollen productivity, dispersal and representation factors of major tree taxa in the European Alps based on pollen trap results. Review of Palaeobotany and Palynology 152: 200–210.

36.

Sjögren

van der Knaap

van Leeuwen

JFN

(2015) Pollen dispersal properties of Poaceae and Cyperaceae: first estimates of their absolute pollen productivities. Review of Palaeobotany and Palynology 216: 123–131.

37.

Sonesson

(1968) Pollen zones at Abisko, Torne Lappmark, Sweden. Botaniska Notiser 121: 491–500.

38.

Sutton

(1953) Micrometeorology. New York: McGraw-Hill.

39.

Tauber

(1965) Differential Pollen Dispersion and the Interpretation of Pollen Diagrams. Copenhagen: C.A. Reitzels Forlag.

40.

Theuerkauf

Kuparinen

Joosten

(2013) Pollen productivity estimates strongly depend on assumed pollen dispersal. The Holocene 23: 114–124.

41.

Tveito

Førland

Heino

. (2000) Nordic Temperature Maps. Klima report 09/00. Olso: Det Norske Meteorologiske Institutt, 54 pp.

42.

van der Knaap

van Leeuwen

JFN

Ammann

(2001) Seven years of annual pollen influx at the forest limit in the Swiss Alps studied by pollen traps: Relations to vegetation and climate. Review of Palaeobotany and Palynology 117: 31–52.

43.

Zackrisson

Nilsson

M-C

Steijlen

(1995) Regeneration pulses and climate–vegetation interactions in nonpyrogenic boreal Scots pine stands. Journal of Ecology 83: 469–483.