Evaluating the effect of flowering age and forest structure on pollen productivity estimates
- First Online:
- Cite this article as:
- Matthias, I., Nielsen, A.B. & Giesecke, T. Veget Hist Archaeobot (2012) 21: 471. doi:10.1007/s00334-012-0373-z
- 1.1k Downloads
Pollen productivity estimates (PPEs) are indispensable prerequisites for quantitative vegetation reconstructions. Estimates from different European regions show a large variability and it is uncertain whether this reflects regional differences in climate and soil or is brought about by different assessments of vegetation abundance. Forests represent a particular problem as they consist of several layers of vegetation and many tree species only start producing pollen after they have attained ages of several decades. Here we used detailed forest inventory data from north-eastern Germany to investigate the effect of flowering age and understory trees on PPEs. Pollen counts were obtained from 49 small to medium sized lakes chosen to represent the different forest types in the region. Surface samples from lakes within a closed forest of Fagus yielded disproportionate amounts of Fagus pollen, increasing its PPE and the variability of all other estimates. These samples were removed from further analysis but indicate a high trunk-space component that is not considered in the Prentice–Sugita pollen dispersal and deposition model. Results of the restricted dataset show important differences in PPEs based on the consideration of flowering age and understory position. The effect is largest for slow growing and/or late flowering trees like Fagus and Carpinus while it is minimal for species that flower early in their development like Betula and Alnus. The large relevant source area of pollen (RSAP) of 7 km obtained in this study is consistent with the landscape structure of the region.
KeywordsGermany Relevant source area of pollen Pollen productivity estimates Forest structure Flowering age Surface samples
It is an old dream of quaternary palynologists to translate pollen diagrams into quantitative representations of past plant abundance (Davis 2000). This is especially important where pollen analytical results are used by neighboring disciplines like archeology (Gaillard 2007) and vegetation or climate modeling (Miller et al. 2008; Gaillard et al. 2010). Also the influence of land use and abiotic factors on regional vegetation patterns can only be evaluated if it is possible to reduce the effect of differential pollen production and dispersal in vegetation reconstructions (Nielsen et al. 2012). Studies investigating the pollen vegetation relationship have been conducted for several decades and the concept of finding factors for correcting pollen proportions to reflect the vegetation abundance goes back to Davis (1963). However, the non-linearity between pollen and vegetation proportions hampered the application of simple correction factors. The development of the extended R-value (ERV) model was intended to correct for the non-linearity, and allow the estimation of the relative pollen productivity of different species (Parsons and Prentice 1981; Prentice and Parsons 1983). This approach was combined with knowledge from studies on pollen transport (e.g. Tauber 1965) to develop numerical models that separate the two species-specific components: pollen production and pollen dispersal (Prentice 1985; Sugita 1993, 1994). Using these models, the dispersal and deposition of pollen can be simulated based on the different fall speed of differently shaped pollen types, which can be estimated or measured directly (Eisenhut 1961), provided estimates of relative pollen productivity are available. Over the last decade, continued theoretical development has led to a conceptual framework that realizes the longstanding dream of quantitative vegetation reconstructions (Sugita 2007a, b; Gaillard et al. 2008). Together with the development of user-friendly computer programs (Sugita 2007a, b; Bunting and Middleton 2005), this has given the impulse for many palynologists to work towards quantitative reconstructions. As mentioned before, pollen productivity estimates (PPEs) are a very important component of these models, but estimating the relative (or absolute) production of pollen of different plant species remains a challenge, as many different factors influence the results (Broström et al. 2008). Moreover, each study addressing the pollen-vegetation relationship to obtain PPEs provides new information to test the models and theoretical concepts of pollen transport and deposition, highlighting potential factors influencing the calculation of PPEs and adding confidence to future reconstructions. Recently a number of these studies have been carried out in Europe: e.g. Norway (Hjelle 1998), Sweden (Sugita et al. 1999; Broström et al. 2004; Von Stedingk et al. 2008), Denmark (Nielsen 2004), England (Bunting et al. 2005), Finland (Räsänen et al. 2007), Estonia (Poska et al. 2011), northern Germany (Theuerkauf et al. 2012), and the Czech Republic (Abraham and Kozáková 2012) as well as for the Swiss Jura (Mazier et al. 2008) and the Swiss Plateau (Soepboer et al. 2007). Initial comparisons of different regional studies revealed large differences in estimated PPEs, which may be due to real differences in pollen production, brought about by regionally different climate or soils, but could also stem from methodological differences (Broström et al. 2008). The influence of vegetation survey methods on PPEs have been shown on the small scale for heathland vegetation in Norway (Bunting and Hjelle 2010), but have so far not been investigated for different assessments of forest vegetation on a landscape scale.
Forests are multi-layered and trees require a number of years of growth before they start to produce flowers and shed pollen. While the former factor will influence the assessment of vegetation cover, the latter affects the vegetation-pollen relationship directly. For example, a 15-year-old plantation of Pinus will hardly be the source of much Pinus pollen, while it will appear like a dense forest on an aerial photograph or satellite image. Even when a vegetation survey is conducted on the ground, it is difficult to decide whether or not a Fagus tree has reached flowering age.
The forest inventory data in north-eastern Germany are exceptionally detailed and thus permit the assessment of the influence of flowering age and canopy structure on PPEs. Moreover, as the data are available in digitized formats there is little limitation to the area that can be used in the analysis.
The aim of this study is: (i) to contribute to the emerging dataset on regional relative PPEs with a focus on tree species; (ii) to examine what influence flowering age and understory position have on PPEs of the major tree species.
Material and methods
Sample collection and preparation
Characteristics of the study sites
Sampling was carried out during the spring and summer of 2009 using a HON-Kajak sediment corer (Renberg 1991) with a 10 cm diameter. The short cores obtained were sampled in the field at 1 cm intervals beginning at the sediment water interface (or shortly above it where the interface was diffuse). Samples were stored in small plastic bags at 4 °C until processing.
In the lab a subsample of 1 cm3 was taken from the uppermost sample of each short core using a syringe. In a few cases from cores with a diffuse sediment–water interface this sample yielded a very low pollen concentration and here the second sample in the short core was also subsampled and processed. The subsamples were processed for pollen analysis following the general procedure described by Bennett and Willis (2001), without sieving and using a 2 min acetolysis. The sample residues were mounted in glycerol and counted at 400× magnification, while cereals were identified at 1,000× magnification. The pollen was identified using the keys of Beug (2004) and Moore et al. (1991), and the reference collection of the Department of Palynology and Climate Dynamics, Göttingen University. A minimum of 1,000 terrestrial pollen grains were counted in each sample. The pollen sum is defined as the sum of all terrestrial pollen grains.
All forest inventory data were included (allFID).
Forest data were included if the flowering age of the different trees and shrubs had been reached (allFIDage).
Only data from the canopy layer were included, while information on understory trees was excluded (canopyFID).
- 4.Only data from the canopy layer were included, and only if the flowering age of the different taxa was reached (canopyFIDage). The information on the flowering age for the trees was obtained from the literature (Table 2).Table 2
Pollen fall speed of 16 taxa and the flowering age considered for the tree taxa
Larix (and Pseudotsuga)
The four different sets of assembled vegetation information for a radius of 15 km around the lakes were compared to pollen percentages in the surface samples. For distance weighting, vegetation data was compiled in 250 m-wide rings for the first km and one km-wide rings from 1 to 15 km distance from the center of the lake. To evaluate the effect of ring width on the results we compiled one test set with 200 m rings over the first 3 km. Calculations of PPEs and the relevant source area of pollen (RSAP) were conducted using the Program ERV-Analysis 1.2.3 (Sugita unpublished). Each of the three different ERV submodels (Prentice and Parsons 1983; Sugita 1994) were applied using the Ring Source model (Sugita et al. 1999) to distance-weight the vegetation data. Wind speed was set to 3 m/sec and pollen fall speeds were taken from the literature or estimated (Table 2). Lake radii were set to an average radius of 140 m (Nielsen and Sugita 2005; Soepboer et al. 2007). We used Pinus as a reference taxon as it is abundant in the vegetation and pollen samples from all sites and represents mainly the single native species, Pinus sylvestris. Initial analysis of the full dataset showed that model performance was improved by reducing the samples to a subset of 39 lakes, omitting lakes where Fagus exceeded a cover of 5 ha within a radius of 250 m from the center of the lake.
We calculated PPEs for 16 taxa (Table 2), of which one is a group of taxa. This group is labeled “wild herbs” and consists of the uncultivated terrestrial herb pollen, including e.g. Poaceae, Plantago lanceolata, Rumex acetosa, R. acetosella, and Chenopodiaceae, with Poaceae being the most abundant. The fall speed for the wild herbs was estimated, using the fall speed of the most frequent taxa in this group.
Average cover and s.d. (%) of the tree taxa within the RSAP
cover ± s.d.
cover ± s.d.
cover ± s.d.
cover ± s.d.
0.85 ± 0.71
0.85 ± 0.71
0.84 ± 0.71
0.83 ± 0.69
2.07 ± 1.11
2.05 ± 1.10
1.92 ± 1.00
1.89 ± 0.99
0.74 ± 0.91
0.64 ± 0.89
0.34 ± 0.53
0.33 ± 0.53
6.36 ± 5.61
4.23 ± 4.34
2.77 ± 2.75
2.55 ± 2.55
0.13 ± 0.14
0.10 ± 0.13
0.10 ± 0.12
0.09 ± 0.11
Larix and Pseudotsuga
2.21 ± 1.31
1.84 ± 1.19
1.98 ± 1.24
1.77 ± 1.17
2.32 ± 1.79
1.55 ± 1.22
1.89 ± 1.47
1.26 ± 1.04
33.61 ± 17.79
32.43 ± 16.87
33.08 ± 17.25
32.04 ± 16.57
0.18 ± 0.17
0.17 ± 0.17
0.18 ± 0.17
0.17 ± 0.17
4.36 ± 2.76
3.01 ± 2.08
2.89 ± 1.93
2.49 ± 1.84
0.39 ± 0.27
0.25 ± 0.27
0.16 ± 0.15
0.14 ± 0.15
Reference taxon and the representation of herb pollen
As calculation of absolute PPEs on a landscape scale is difficult and requires pollen accumulation rates (Sugita et al. 2010), PPEs are commonly estimated relative to a reference taxon, which needs to be present in the pollen count and vegetation of all samples. A common choice is Poaceae (Broström et al. 2008) as it occurs almost everywhere. Here forest lakes were targeted and the cover of herbaceous vegetation on the forest floor was not assessed. Consequently, vegetation information on Poaceae is incomplete, and probably particularly biased near the lakes. Poaceae was therefore not chosen as a reference taxon and not calculated separately (but combined with the other wild herbs in an attempt to minimize this bias). The scatter-plots in Fig. 6 show a poor fit for herbaceous pollen which may reflect the sampling design, where lakes seldom border non-forested areas. As the commonest tree in the study region, Pinus fulfilled all requirements of a reference taxon including a good fit between vegetation and pollen data.
Flowering age and forest structure
The results show that flowering age and forest structures have a marked influence on the calculation of PPEs for some taxa. The autecology of the different tree species is modulating this effect in different ways and, based on the forest inventory data from Brandenburg, we can identify three different groups. Betula and Alnus are fast growing and reach their flowering age quickly. Thus, consideration of flowering age and understory have little influence on the estimates of their abundance, while their relative PPEs may be affected due to the reduction in effective abundance of other species. A second group consists of fast growing trees that are rarely found in the understory, while their flowering age may directly affect the pollen-vegetation relationship. In this dataset, examples of this type are Fraxinus and Picea. For the third group, mainly Fagus and Carpinus, the abundance in the understory has a larger effect than the flowering age on the pollen-vegetation relationship. These different ways in which the autecology of the tree species influences their pollen–vegetation relationship would also play a role in the natural woodlands of the past that we are aiming to reconstruct. However, nearly all European forests today are heavily managed and naturally occurring age structures are likely to be skewed towards younger non-flowering trees. The harvesting age of trees is often the time when they have started to produce ample pollen and thus the strong pollen producers are removed from the forest. Thus present tree plantations and woodlands may not be good analogies for the past, and it would be preferable to estimate relative pollen productivity from flowering trees only. Regional differences in the age structure of woodlands are a potential reason for differences in PPEs between regions in previous studies, e.g. those compared by Broström et al. (2008). However, omitting trees based on their flowering age is also not without problems, as trees start to flower at different ages depending on biotic and abiotic conditions (Schütt et al. 2006). Most trees will not go from not flowering to abundant pollen production in a single year, but rather gradually increase the amount of pollen produced with age. It is also likely that a free-standing tree produces more pollen than one that grows within a dense stand (Aaby 1994). This effect of increased pollen production with forest openings has so far not been quantified, but, if important, it would greatly hamper our ability to reconstruct small openings in the forest using pollen data.
The occurrence of species in different stories in the canopy (e.g. superstructure, matured forest) is easy to assess in the field, but is difficult, if not impossible, to gauge from remote sensing. Tree abundance for PPEs has been estimated in a number of different ways: field estimates of basal area (e.g. Jackson and Kearsley 1998), cover of individual plants, considering different forest layers and leading to values above 100 % (e.g. Soepboer et al. 2007), canopy cover as seen for one forest layer (e.g. Mazier et al. 2008) based on land-cover classifications in combination with field surveys and forest inventory, land-cover classification, and combinations with field surveys (e.g. Poska et al. 2011), remote sensing forest inventory data (Räsänen et al. 2007) or detailed forest inventory data (Theuerkauf et al. 2012 and present study).
Based on different assessments of heathland vegetation cover in western Norway, Bunting and Hjelle (2010) have shown that the way in which vegetation data are collected influences the PPEs and RSAP. We do not find differences in the RSAP, but show that PPEs are markedly influenced by the inclusion or exclusion of understory trees.
Considering the understory trees also has theoretical implications for the modeling of pollen dispersal. Tauber (1967) argued that the pollen carried to the lake through the trunk space is a significant component of the total pollen deposited at the site. However, this trunk space component is not considered in the dispersal models used here (Prentice 1985, 1988; Sugita 1993, 1994). In the present study, we omitted 10 lakes from the analysis that were situated in closed-canopy forest dominated by Fagus. The inclusion of this subset resulted in inflated PPEs for Fagus and less stable results for other taxa, indicating that the local abundance of Fagus around the lakes was not adequately considered in the model. The reason may be the dispersal model used in the ERV-program (Theuerkauf et al. 2012) and the large trunk space component of Fagus pollen in these situations. Sjögren et al. (2010) address the problem of the large trunk space component by dividing the pollen source area into different components and applying different dispersal functions. This approach seems rather ad hoc, as it is difficult to determine the contribution of the different components. However, this method can potentially accommodate a large trunk-space component (Filipova-Marinova et al. 2010). Pollen from understory trees may not contribute to above-canopy pollen transport in the same way as canopy trees, but it should be important in the trunk-space component (Tauber 1965). In this sense, the abundance of trees in the understory could be important near a site, while it may have little influence several hundred meters away from the site. Thus it could be meaningful to include the understory trees within a short distance around the site and exclude them from vegetation data further afield.
The obtained RSAP of 7 km is rather large in comparison with similar analysis based on lakes of similar size, where RSAP values of 0.8–2 km were obtained (Nielsen and Sugita 2005; Soepboer et al. 2007; Hjelle and Sugita 2012; Poska et al. 2011). Poska et al. (2011) showed that vegetation classification can influence the RSAP for lakes, as was shown for moss polsters and heath vegetation by Bunting and Hjelle (2010). We did not find this effect using the different datasets with regards to flowering age and understory trees being omitted. The reason may be that the configuration of vegetation units remained the same in all situations and the omission of trees generally only affected their abundance rather than their presence. Hjelle and Sugita (2012) found that the distribution of lake-sizes could also influence the RSAP. The restricted dataset has a lake-size distribution nearing a normal distribution, which is slightly right skewed. The full dataset has a lake size distribution with a stronger skew to the right but yields a similar RSAP. It is most likely that the result is in fact due to the distribution of vegetation types and forest species in the study area, which is a well-known factor controlling the size of the RSAP (Sugita et al. 1999; Bunting et al. 2004; Nielsen and Sugita 2005; Hellman et al. 2009). Soil substrate, and therefore nutrient and water availability in north-east Germany are linked to the glacial geomorphology of the area. The Quaternary geology of the region consists of glacial sediments several hundred meters thick, including sands, clays and tills. The extensive outwash plains often only support Pinus trees and are rarely used for agriculture, while much of the nutrient-rich till plains are cleared for agricultural purposes. Nutrient-rich soils also developed on terminal moraines that are too steep for agriculture and are often covered by Fagus-dominated forests. These three geomorphological units dominate the landscape structure and vegetation-type patch size. They may occur in their pure form over tens of kilometres or mix with other geomorphological features creating smaller-scaled vegetation patterns.
The lakes in Switzerland from which Soepboer et al. (2007) obtained an RSAP of 800 m were often surrounded by a belt of trees, and woodland patches in the surroundings have diameters of several hundred meters. The size of woodland patches in Estonia range from a few hundred meters to several kilometers and Poska et al. (2011) find a RSAP of 1,500–2,000 m. In Brandenburg extensive forests (which often stretch over more than 20 km) alternate with extensive agricultural areas. Thus, extrapolation from the woodland patch-size versus RSAP in Switzerland and Estonia to the size of forest patches in Brandenburg may explain the large RSAP obtained. This example shows the dependency of RSAP on the patch size (Bunting et al. 2004; Hellman et al. 2009) for the selected lakes and indicates large effective patch size for the region investigated here.
Our estimates for Alnus agree well with studies based on lakes (Poska et al. 2011), while studies using moss samples tend to produce much lower values (Southern Sweden, Broström et al. 2004; Central Bohemia, Abraham and Kozáková 2012). This may again be a result of the trunk-space component, which is generally not considered in the pollen dispersal model. Alnus is often found on the lake shore where a large amount of pollen may be carried in the trunk space, or simply drop into the lake, while moss samples from upland vegetation are generally distant from this tree species and thus collect Alnus pollen that is transported by the wind above the canopy. The good agreement for the PPE of Betula relative to Pinus between studies is encouraging, especially as Betula is often mixed in with other trees or occurs near the lake shore and it is thus difficult to estimate its abundance.
Few studies have obtained a PPE for Carpinus and our result agrees with Soepboer et al. (2007) in that its relative pollen production is higher than Pinus, although the relative amount differs greatly between the two studies. This may in part be due to the high uncertainty in the PPE for Pinus obtained for Switzerland (Soepboer et al. 2007). Our results show that the PPE of Carpinus depends strongly on whether understory trees are included in the vegetation estimate (Table 4), a factor which was not or only partly considered in several previous studies. More studies have estimated the PPE for Fagus with varying results. As discussed earlier, we suppose that the trunk-space transport of Fagus pollen may have a strong influence at sites within Fagus-dominated forest. This may contribute to the high PPE estimated by Theuerkauf et al. (2012) for the area just north of this study (calculated by the ERVv1.2.3. software). The value obtained here compares well with that estimated in southern Sweden, where, as in Brandenburg, both Pinus and Fagus are abundant forest trees. Also the PPE for Picea compares well to the estimate from southern Sweden, although the Picea trees in Brandenburg are planted and the distribution considered natural is situated just to the south of the study area. Larix and Pseudotsuga are planted widely in Brandenburg and were thus considered in the analysis. As the two trees cannot be differentiated by their pollen type, the PPE is difficult to interpret.
The values obtained for Quercus are surprisingly low and similar values have only been obtained from Central Bohemia (Abraham and Kozáková 2012). The dataset has a good spread of Quercus tree and pollen abundance, which gives us confidence in the results. In recent years many areas of Quercus trees in Brandenburg have been severely damaged by insects (particularly the moth Thaumetopoea processionea) and/or the fungus Microsphaera alphitoides (Möller et al. 2010) which can completely defoliate the tree, influencing its resource allocation and potentially reducing if not stopping the production of pollen. Also the age structure of Quercus in Brandenburg is skewed towards younger trees which may not produce the same amount of pollen as an abundance of older trees. While these factors would explain a somewhat lower PPE, the difference compared to the estimates from Switzerland is about threefold and thus difficult to explain by these factors alone.
The PPE for Tilia cordata compares well with the value for Tilia estimated in Southern Sweden (Sugita et al. 1999). Tilia cordata is the more abundant species in our dataset, while T. platyphyllos is also present. ERV model runs combining both pollen types and trees yield similar PPEs and thus we suggest that the value presented here may be appropriate for both species. However, T. platyphyllos is rare in the region and we thus restricted the final analysis to T. cordata.
It is desirable that the relationship between pollen and vegetation should be as good as possible for the reference taxon, because all the calculated PPEs depend on this taxon. We propose to use PPEs from ERV submodel 3 for vegetation reconstruction, because Pinus shows the best fit. Submodel 3 also produced lower likelihood function scores than the other submodels (Fig. 5), indicating a generally better fit to the data across all taxa. Based on the insights gained when working with the data, we recommend the further usage of the results from ERV submodel 3, based on the inclusion of all flowering trees, i.e. considering the effect of flowering age.
Both forest structure and flowering age have strong effects on the calculation of PPEs for most tree species, which should be considered when estimating PPEs or reconstructing vegetation. Forest structure is more important for slow growing trees like Fagus and Carpinus, while flowering age is most important for fast growing trees like Fraxinus and Picea, where flowering age is reached late. Based on this dataset, these effects may cause variations in PPEs in the order of up to 100 % for trees that are slow growing and/or reach their flowering age late. Larger differences in PPEs between regions will most likely have different reasons, as will differences observed for trees that reach their flowering age early.
Trees growing at or near the lake shore contribute a large amount of pollen to the lake through the trunk-space component of pollen transport. This component is generally not considered when estimating PPEs, but could have a large effect and may thus be another factor explaining the differences between studies.
The large RSAP of 7 km is consistent with the landscape structure consisting of large units shaped by ice-marginal processes during the last glaciation. The resulting differences in soil substrates and geomorphology are reflected in different land use today, and it would also have supported different forest types in the past.
This study adds to the growing dataset of PPEs for Europe and thus improves our ability to reconstruct past vegetation cover.
We thank Ralf Köhler from the Landesumweltamt in Brandenburg for the authorization to do our fieldwork as well as Sabine Busch from Landesbetrieb Forst Brandenburg (LFB). We would like to thank Martin Theuerkauf for support in handling of the forest inventory data and long and good discussions. We thank Shinya Sugita for providing us with the unpublished software ERV-Analysis 1.2.3 and a critical discussion of the results. We also thank Jörg Christiansen for technical support. We thank the forest authorities of Brandenburg and Mecklenburg-Vorpommern for providing us with the forest data, particularly Konrad Müller from LFB, Potsdam. We also like to thank Simon Connor and an anonymous referee for helpful comments upon an earlier draft of this manuscript. This study was funded by the German Research Foundation (DFG, GI 732/1-1).
This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.