Relative pollen productivity estimates for vegetation reconstruction in central-eastern Europe inferred at local and regional scales

Abstract

Understanding pollen-vegetation relationships is crucial for accurate land-cover and climate reconstructions, yet important parameters for quantifying past vegetation abundance are mostly unknown for large parts of Europe harbouring temperate thermophilous ecosystems. We collected pollen and vegetation data in central-eastern Europe, a region covered by patchy cultural landscapes of high biodiversity to estimate relative pollen productivity (RPP) for important pollen-equivalent taxa. Our study area was situated in the south-western part of the White Carpathians (Czechia–Slovakia borderland), where we collected 40 modern moss pollen samples scattered over 250 km² and mapped vegetation within 100 m around each pollen site. Additional vegetation data were compiled from Forest management plans, Natura 2000 habitat mapping and floristic inventories over the entire area. We calculated RPP (referenced to Poaceae) by testing two approaches: the extended R-value (ERV) model by estimating relevant source area of pollen and the REVEALS-based productivity using regional scale vegetation estimates. Two models were applied to depict pollen dispersal: Lagrangian stochastic and the Gaussian plume (Prentice) models. We estimated RPP for 16 taxa using the ERV model and an additional nine taxa using REVEALS. Both approaches found Plantago lanceolata-type to be a high pollen producer, Quercus medium-to-high, Asteraceae subf. Cichorioideae, Anthemis-type, Ranunculus acris-type and Rubiaceae low-to-medium and Brassicaceae and Senecio-type as low pollen producers. Results for other, mainly tree taxa, significantly differed in both approaches mainly due to largely uneven representation in both local and regional vegetation. In comparison with other studies, our data demonstrate a high variability in the estimated RPPs which could be influenced by climatic conditions or potentially vegetation structure. We suggest that the accuracy of RPP estimates could be enhanced by comparing modern pollen data with large-scale vegetation data in the future.

Keywords

Czech Republic ERV model Gaussian plume model Lagrangian stochastic model modern pollen spectra REVEALS vegetation cover

Introduction

Quantifying past land cover has been shown to be crucial for understanding past climate change, evolution of human societies or answering various ecological questions (Gaillard et al., 2010; Roberts et al., 2018; Seddon et al., 2014). Since most of the information on past vegetation comes from palynological data, it is absolutely essential to be able to interpret such records with the highest accuracy. Translation of pollen data into vegetation cover, therefore, has to account for mechanisms of pollen production, transport and deposition. These data are obtained by field measurements for particular plant taxa and regions (Bunting et al., 2013).

Several approaches have been used to investigate parameters of pollen-vegetation relationship. A combination of annual pollen loadings together with vegetation data measured in absolute units proved to provide independent information on pollen production for each studied taxon from individual sedimentation basins (e.g. Filipova-Marinova et al., 2010; Sjögren, 2013; Sugita et al., 2010). Obtaining such data is, however, time consuming since several vegetation seasons are necessary for investigation of pollen accumulation by means of pollen trapping to calibrate a representative mean of interannual variations (van der Knaap et al., 2010). Moreover, most fossil pollen data are available only in percentages for which absolute pollen productivity is not required to reconstruct past vegetation. Instead, deploying pollen surface samples (moss cushions) aggregating several years of pollen deposition provides a viable solution that estimates relative pollen productivity (RPP; Bunting et al., 2013). Using moss cushions provides an effective method for quantifying vegetation reconstructions from percentage pollen data, and is less time consuming.

Estimated RPPs vary across regions in Europe (Abraham and Kozáková, 2012; Baker et al., 2016; Mazier et al., 2012). While their variation is not yet fully understood, the most likely reasons for that are climatic, but other factors linked to methodological issues such as vegetation sampling strategies or vegetation structure certainly play an important role (Bunting et al., 2013). In Europe, RPPs were estimated in different biomes including boreal, boreo-temperate and temperate vegetation (Broström et al., 2008). Yet, estimates of RPPs are still missing from more continental and summer-warm central-eastern Europe, with forest-steppe or steppe biome. Nonetheless, RPP values from more oceanic western and northern Europe are used for land-cover reconstructions in these regions (e.g. Feurdean et al., 2015). However, because there is such a large variation in RPPs across different biomes and regions, this begs to question whether applying RPP values from one region to another is a valid approach (e.g. in Marquer et al., 2017; Trondman et al., 2015). In this paper, we aim to produce RPPs from central-eastern Europe, where thermophilous temperate forests meet with forest steppes. This region also maintains an ancient vegetation structure, originated in the 11th–12th century, shaped by long-lasting human impacts (Hájková et al., 2018). In addition, the application of quantitative techniques with parameters calibrated in the region will help to sort out many important ecological questions regarding the development of cultural landscapes or the persistence of open land during the Holocene (Feurdean et al., 2015; Kuneš et al., 2015).

An approach using relevant source area of pollen (RSAP) has proved to be one of the most useful techniques for estimation of RPPs (e.g. Bunting et al., 2013; Sugita, 1994), because it does not require accurate large-scale vegetation maps. Several studies have also tested the applicability of different weighting methods for calculating pollen-vegetation relationships; to date using Lagrangian stochastic dispersal model (LSM; Mariani et al., 2016; Theuerkauf et al., 2013) seems to be the most useful approach. Here we use two approaches of estimating RPPs using the same pollen dataset, one estimating both RPPs and RSAP, and the other using large-scale vegetation data. The latter is in fact the one most often used for quantitative vegetation reconstructions, for example, in application of the REVEALS model (Sugita, 2007).

Our main aim is to estimate RPPs of the main dominant plant taxa in a patchy cultural landscape of high biodiversity (including species richness and community diversity) in an area of temperate thermophilous vegetation. We will test the performance of two different approaches in estimating RPPs combining two spatial extents of vegetation data, and discuss the results across similar vegetation regions and structure. Our main question, therefore, concerns, whether it is realistic to locally estimate RPPs and extrapolate these parameters to regional land-cover reconstructions.

Study area

Our study region encompasses the south-western portion of the White Carpathians. This low mountain range (highest peak 970 m a.s.l.) belongs to the Outer Western Carpathians and borders the Pannonian Basin, a warm and dry forest-steppe environment (Chytrý, 2012; Rasser and Harzhauser, 2008). The mean annual temperatures in the region are approximately 7–9°C and the annual precipitation is approximately 550–750 mm (Tolasz et al., 2007). The prevalence of easily eroded sedimentary bedrock consisting of alternating layers of calcium-rich claystones and sandstones has resulted in the development of a landscape with long gentle slopes, broad valleys and rounded ridges. Because of its peripheral position within the country and legal protection as a Protected Landscape Area since 1980, modern agriculture, industry and urbanization have not altered the area substantially. A diverse vegetation mosaic was preserved here, showing some features of a traditional rural landscape (Jongepierová, 2008) arranged into patches of broad-leaved forests (Fagus sylvatica, Quercus robur and Carpinus betulus being the dominant forest canopy taxa), semi-natural dry and mesic grasslands and other agricultural land, often used in a low-intensive manner (hay making, grazing). Plants from various floristic elements (central-European, continental, sub-Mediterranean, montane) grow in the area together creating a global hotspot of plant species richness (Roleček et al., 2014; Wilson et al., 2012). Study sites were scattered over an area of 250 km² (Figure 1) and at altitudes of 205–685 m a.s.l.

Figure 1.

Location of the study area, sampling design and vegetation cover. Vegetation survey design with indicated land cover (zones and outlines) is depicted at different scales. Ring widths corresponding to each survey zone are shown in Figure A.1, available online. (a) 0–10 m (Figure A.1a, available online) tree canopies in semitransparent colour illustrates their overlaps; (b) 0–10 m (Figure A.1a, available online) radial position of releves used for open habitats and forest undergrowth (squares); (c) 10–100 m (Figure A.1b, available online) plant communities (different colours) sampled by releves (squares) except for surveying of the forest communities (walked through as indicated by arrow); (d) 100–2000 m (Figure A.1c, available online) vegetation model map compiled from various sources (see ‘Materials and methods’ section), only three major habitats are shown, codes of pollen sites (red dots) correspond to Figure 2; (e) regional scale 2–60 km (Figure A.1d, available online) CORINE land-cover maps with pollen sites (black dots); and (f) study area within Europe.

Materials and methods

Pollen sampling and vegetation mapping in the field

During three field seasons between 2012 and 2014, 40 sites were located and sampled following the CRACKLES protocol (Bunting et al., 2013), with some modifications described below. Sites were placed with an approximate spacing of 1.5 km (1697 ± 810 m) to avoid autocorrelation. With regard to earlier studies (Abraham and Kozáková, 2012), we aimed at even distribution of both types of sites to gain sufficient representation of tree taxa. We also avoided forest edges as these may significantly affect pollen dispersal and distance weighting. At each of the forested sites, we aimed at collecting a moss cushion in a canopy opening, since dispersal models assume canopy component as the main vector of pollen transport. We also gave preference to places with a high diversity of tree species to increase the number of replications for individual species. In non-forest sites, moss cushions were positioned in open habitats with the minimum distance of 10 m from growing trees. Each moss cushion (of a minimum size of 100 cm²) represents the centre of each vegetation site and was marked by four iron nails to enable future identification by a metal detector. In addition, each moss cushion was covered by a metal lid during the vegetation survey to avoid erroneous pollen contamination. Each moss cushion was revisited and sampled for pollen at the end of each vegetation season. We mapped 20 forest sites in 2012, 17 non-forest sites in 2013 and remaining 3 non-forest sites in 2014. Regarding pollen sampling, 17 forested sites were sampled in 2012, while remaining 3 forested sites and 20 non-forest sites were sampled in 2013. We assumed minimal interannual variation of the grassland and forest vegetation. One forest site showed very poor preservation of pollen, consequently 19 forest and 20 non-forest sites were used for the final analyses. We also encountered poorer pollen preservation at other sites that might have influenced pollen representation in the particular sample. Yet, we considered those samples as representative since we did not find any abnormal quantities of decay-resistant pollen such as Tilia or Asteraceae subf. Cichorioideae.

The vegetation within the first zone (0–10 m ring) was surveyed according to the CRACKLES protocol (Bunting et al., 2013) using an array of 21 1 × 1 m² quadrats distributed according to Figure 1a and b. First, all vascular plant species were recorded in the mapped quadrat, and then abundances of each species were estimated. After all 21 quadrats were mapped, the remaining area was surveyed and the presence of unrecorded species was noted. The spatial position of individual trees and shrubs in the first zone was recorded using a freehand sketch illustrating a vertical projection of the canopy structure, including overlaps and gaps.

In the second zone (10–100 m ring; Figure 1c), the main vegetation types were mapped in the field with the assistance of aerial photographs. Non-forest and shrub habitats were described using four randomly placed vegetation quadrats. In the case of the forest sites, we refrained from the use of the CRACKLES protocol (we considered the 6-m transects as non-representative). Instead we improved our mapping by surveying the whole area and recording the relative abundances of all tree species. Information on the herb layer was added using the procedure described in the ‘Data processing and analysis’ section.

Pollen analysis

Collected moss cushions were prepared for pollen analysis using standard procedures (Faegri et al., 1989). Lycopodium spores were added to the sample as a tracer prior to the treatment (Stockmarr, 1971). Moss was first processed with 10% KOH in a hot water bath, and then acetolysed for 2 min. The final pollen residue was stored in glycerin. Pollen samples were counted under light microscope at 400× magnification until the sum of 500 pollen grains was reached. For pollen determination, pollen keys (Beug, 2004; Punt et al., 1976–2009), pollen atlases (Reille, 1998 [1992]) and the reference collection of the Institute of Botany, CAS, in Brno, were used.

Data processing and analysis

Field data on vegetation composition were digitized using the Turboveg 2.0 software (Hennekens and Schaminée, 2001) and further analysed in R (R Core Team, 2017). Vegetation data for the first zone outside the central 1 m² were merged into concentric rings with varying widths of 1 m (0.5–1.5 m in radius), 1.5 m (1.5–3 m in radius), 3 m (3–6 m in radius) and 4 m (6–10 m in radius) (Figure A.1a, available online). Canopy gaps within each forested site were calculated for each ring using GIS. The area of shrub and herb layer was adjusted (decreased) in order to keep the sum of all the vegetation layers as 100 %. If the forest canopy cover had 100% cover, the herb layer was excluded from the final calculation.

Data on open and shrub communities situated in the second zone (10–100 m) were extracted using rings of 10 m width (Figure 1c; Figure A.1b, available online). Data on tree layer composition recorded during field survey were complemented with available forestry data from the vegetation map.

For the third zone (100–2000 m), a vegetation map was created using the data compiled from several sources: forest management plans in stand resolution (http://geoportal.uhul.cz/LhpoMap/?MapComposition=spta), Natura 2000 habitat mapping (http://mapmaker.nature.cz/wmsconnector/com.esri.wms.Esrimap/aopk_biotopy_wms) and floristic inventories. For those areas and habitats with few data available (mostly human-made habitats distant from our sampling sites), the map was corrected by expert judgement of a botanist from the local Protected Landscape Area Administration. The information from the resulting vegetation map was extracted for consecutive, 100 m wide rings (Figure A.1c, available online).

Plant species were assigned to pollen types according to Table A.1, available online.

Calculation of RPP based on local data

Extended R-value (ERV; Prentice and Parsons, 1983) model analysis with both pollen count and vegetation ring data was applied using the ERV Analysis v 2.5 programme (Sugita, unpublished). To compare the performance of different dispersal models (such as in Mariani et al., 2016; Theuerkauf et al., 2016), Lagrangian stochastic model (Kuparinen et al., 2007) and Gaussian plume model (Prentice, 1985), hereafter as LSM and GPM, were applied as taxon-specific distance weighting to vegetation ring data. In the GPM wind speed was set to 3 m·s^–1 using neutral atmospheric conditions. Basin radii were set to 0.5 m and species-specific terminal velocities were used for each pollen taxon (Table 1; Broström et al., 2004; Eisenhut, 1961; Sugita et al., 1999). We estimated the RPP based on local data (RPPL) and background pollen for each dispersal model using three sub-models (ERV 1, 2 and 3; Parsons and Prentice, 1981; Sugita, 1994). In ERV 1 and 2, ring vegetation data were input as proportions of selected taxa; in ERV 3, ring vegetation data were converted to proportions within the ring area including bare land. All sub-models produced likelihood function scores and based on that, RSAP was estimated using moving window of 300 m. Once RSAP was determined, both RPPL and background pollen could be identified at the estimated RSAP. We used Poaceae as the reference taxon with RPP set as 1 as it was well represented in both pollen and vegetation at all sites.

Table 1.

List of pollen taxa with terminal velocities and calculated RPPL with error estimates and RPPR.

Taxon	Terminal velocity (m·s^–1)	RPPL (LSM ERV 1)	SD	RPPR (LSM)	SD
Abies	0.12			1.08	0.99
Acer	0.056	0.07	0.01	1.25	0.75
Alnus	0.021			2.44	0.73
Anthemis arvensis-type	0.031	0.68	0.06	0.94	0.43
Asteraceae subf. Cichorioideae	0.051	0.65	0.06	1.04	0.64
Betula	0.024			2.53	0.91
Brassicaceae	0.022	0.07	0.04	0.09	0.03
Carpinus betulus	0.042	0.1	0.01	1.36	0.36
Corylus	0.025	0.3	0.04	2.31	1.13
Cyperaceae	0.035	0.77	0.05	2.10	0.88
Fagus	0.057	0.06	0.002	0.88	0.25
Fraxinus	0.022			0.79	0.37
Chenopodiaceae	0.019			1.58	0.74
Picea	0.056	0.36	0.02	2.39	0.93
Pinus	0.031			1.55	0.44
Plantago lanceolata-type	0.029	9.84	0.24	2.24	0.71
Poaceae	0.035	1	0	1.00	0.00
Quercus	0.035	1.7	0.03	2.08	0.46
Ranunculus acris-type	0.014	0.59	0.09	1.38	1.13
Rubiaceae	0.019	0.76	0.05	1.03	0.74
Rumex acetosa-Typ	0.018	0.11	0.19	1.94	1.35
Salix	0.022			1.43	0.62
Senecio-Typ	0.017	0.28	0.04	1.49	1.37
Tilia	0.032	0.45	0.02	2.30	1.24
Ulmus	0.032			0.96	0.77
Urtica	0.007			1.16	0.52

RPPL: relative pollen productivity based on local data; RPPR: relative pollen productivity based on regional data; LSM: Lagrangian stochastic model; ERV: extended R-value; SD: standard deviation.

For the first run, we selected taxa that were present at more than half of the sites (19 taxa) in both pollen and surrounding vegetation within the radius 100 m. For the second run, we excluded taxa that exhibited short range of distance-weighted vegetation abundance and pollen abundance relationships.

Calculation of RPP based on regional data

According to the REVEALS model, we assumed that combining all pollen assemblages provides a realistic estimate of the regional vegetation (Mazier et al., 2012; Sugita, 2007). The optimal RPP based on regional data (RPPR) were found using the DEoptim package in R (Mullen et al., 2011) by obtaining the minimum distance between REVEALS estimates and regional vegetation composition. Dispersal models for the REVEALS calculation were applied using the DISQOVER package (Theuerkauf et al., 2016) for the LSM, and using an R script from Abraham et al. (2014) for the GPM. As the distance measure for the REVEALS and regional vegetation comparison, sum of squared differences per taxa between REVEALS vegetation and vegetation data were selected and divided by vegetation data per taxa plus one.

We inverted the REVEALS model and modified our previous script (Abraham et al., 2014) for the calculation of RPPRs. To obtain error estimates, we iterated calculation of RPPRs for 12 times. In each run, we selected 39 sites with repetitions leaving one-third of the sites out. Optimization of the RPPRs for each selection starts with random set of RPPR values and each iteration searches for better set of RPPR values. We set 500 iterations, which showed to be sufficient to decrease the variation between each selection. The initial range of possible pollen productivity values was set to 0.1–100. The best values with the lowest distance were subsequently recalculated to the reference taxon Poaceae. Mean values and standard deviations were determined based on all runs of RPPR calculations.

Regional vegetation composition was set as one ring from 0.5 m to 60 km following previous investigations (Abraham et al., 2014). Practically, the area of regional vegetation for all aggregated pollen sites was delimited by merging of 60-km rings around all sites. The vegetation abundances of tree taxa within the 60-km radius were obtained from the forest inventory statistics in Czechia for municipalities of III grade (http://www.uhul.cz/ke-stazeni/informace-o-lese/slhp) and Slovakia for districts (http://gis.nlcsk.org/lgis/); abundance of Brassicaceae was obtained from the Czech agricultural statistics (https://vdb.czso.cz/vdbvo2/faces/en/shortUrl?su=2b61f31b); and abundances of other herb taxa were extrapolated from available phytosociological releves into a CORINE land cover (hereafter CLC; European Environmental Agency, 2006; https://www.eea.europa.eu/ds_resolveuid/DAT-109-en).

In addition, we analysed whether our field sampling within the region was representative. We compared vegetation cover within a 500-m radius around 39 sampled sites (local-scale data) and vegetation within a 500-m radius around a set of 39 randomly generated sampling points (regional scale data). Using random sampling in the rest of the region (in 2–60 km radius around our field sites), we were able to evaluate the representativeness of our field sampling. In both datasets, within a 500-m radius, we observed vegetation composition based on various sources as described above and landscape characteristics from CLC map. Besides the representation of CLC classes, we analysed landscape patchiness, measured as the total length of the CLC borders around all 39 sites.

Results

Pollen assemblages

At 39 sites, altogether 23,257 pollen grains were identified (Figure 2). Altogether 128 pollen types were identified in all samples.

Figure 2.

Modern pollen assemblages diagram reflected in percentage values of total terrestrial pollen. The horizontal line divides forest sites (above) and non-forest sites (below). Numbers refer to sites in the map (Figure 1).

Vegetation data

Vegetation composition from our field ring data at the local scale (Figure 3a) shows higher abundances of Poaceae, Quercus, Tilia and Carpinus at the expense of Fagus and Pinus, which are more abundant in randomly selected sites at the regional scale (Figure 3b). The proportion of forest is similar at both spatial levels. Arable fields and artificial surfaces yielding large proportions of the regional vegetation were avoided in our sampling, instead we focused on pastures and heterogeneous agricultural areas. Landscape patchiness derived from the total length of CLC borders is 4.5 ± 1.8 km at our field sampling and 3.1 ± 2.4 km at random points.

Figure 3.

CLC and vegetation composition in the 500-m radius around sites of (a) data at real sites from field survey, (b) data at randomly placed sites. CLC classes: 112 and 142, artificial surfaces; 211, arable land; 221, vineyards; 231, pastures; 242 and 243, heterogeneous agricultural areas; 311, forests; and 324, transitional woodland-shrub.

RSAP

A moving window average analysis combined with a visual assessment suggests that LFS reaches asymptote at 450 m for ERV 1 and ERV 2, while at 350 m for ERV 3 (Figure 4). Therefore, we consider these distances as their RSAPs, respectively. LFS for ERV 1 indicates the best iterative solutions (lowest LFS values) obtained; we, therefore, consider this sub-model as the best for calculating the RPPL. LFS results calculated for the GPM dispersal model are shown in Appendix A (Figure A.2), available online.

Figure 4.

Likelihood function scores for all ERV sub-models 1, 2, and 3 using the LSM dispersal model. The triangles indicate RSAP for particular sub-model. ERV 1 (450 m) has been finally selected for RPPL calculation.

Pollen-vegetation relationships and background pollen

Original data consisting of pollen proportions and LSM-dispersal-weighted vegetation abundance still show scattered relationship (Figure 5a). The linearity of the pollen-vegetation relationship is achieved in the ERV-adjusted data for Quercus, Plantago lanceolata-type, Poaceae and Tilia. Relationships of other taxa remain more scattered, suffer from many low values and few outlying high values, making the relationship weaker (Figure 5a and b; please refer to Table A.2 and Figures A.3–A.7, available online, for other sub-models and dispersal models).

Figure 5.

Scatter plots showing the relationship between pollen and plant abundances within 450 m. (a) Scatter plots of distance-weighted plant abundances with the LSM and pollen proportions used for model calibration. (b) Scatter plots of adjusted vegetation abundances after being processed with the ERV 1 model and pollen proportions. Line depicts pollen productivity and shading represents background pollen component.

Estimated background pollen loading at RSAP (Figure 5b, hatched area) is highest for Cyperaceae, Rumex acetosa-Typ, Picea and Brassicaceae, while Acer, Plantago lanceolata-type, Rubiaceae, Senecio-Typ and Tilia yield the lowest values.

We were not able to estimate RPPL for some regionally important taxa including Pinus and Fraxinus because of their very poor representation in the sampled vegetation.

RPP estimates

The ERV-based RPP estimates calculated at the local scale (RPPL) identified most of the taxa as low pollen producers (Figure 5b and Table 1). Low producers are as follows (lower than 0.4; see Figure 6): Acer, Brassicaceae, Carpinus betulus, Corylus, Fagus, Picea, Rumex acetosa-Typ and Senecio-Typ; low-to-medium producers are as follows (0.4–1): Anthemis arvensis-type, Asteraceae subf. Cichorioideae, Cyperaceae, Ranunculus acris-type, Rubiaceae and Tilia. Quercus was identified as medium-to-high pollen producer (1–2.5) and Plantago lanceolata-type, the only taxon with significantly higher values than Poaceae, was identified as a high pollen producer.

Figure 6.

Comparison of calculated values for the RPPL and RPPR for the LSM dispersal model with error estimates. Vertical dashed lines delimit categories as described in the text: low, low-to-medium, medium-to-high and high pollen producers.

The REVEALS-based RPP estimates calculated at the regional scale using the LSM (RPPR) relative to Poaceae identified the following taxa as low pollen producer (lower than 0.4): Brassicaceae; low-to-medium (0.4–1): Anthemis arvensis-type, Fagus, Fraxinus and Ulmus; medium-to-high pollen producers (1–2.5): Abies, Acer, Alnus, Asteraceae subf. Cichorioideae, Carpinus betulus, Corylus, Cyperaceae, Chenopodiaceae, Picea, Pinus, Plantago lanceolata-type, Quercus, Ranunculus acris-type, Rubiaceae, Rumex acetosa-Typ, Salix, Senecio-Typ, Tilia and Urtica; and high pollen producer (higher than 2.5; Table 1): Betula.

Comparison of RPPs calculated with ERV and REVEALS (Figure 6) shows that the following taxa yielded similar productivity values (including error estimates): Anthemis arvensis-type, Asteraceae subf. Cichorioideae, Brassicaceae, Plantago lanceolata-type, Quercus, Ranunculus acris-type, Rubiaceae and Senecio-Typ. The other taxa yielded significantly different values, particularly striking being the difference for Carpinus, Corylus, Picea and Rumex acetosa-Typ which all were identified as low pollen producers according to the ERV model, while as medium-to-high pollen producer according to the REVEALS model.

In our calculations using the LSM dispersal model, all three ERV sub-models produce similar results (see Appendix A). We therefore discuss only ERV 1, as it yielded the lowest likelihood function scores when estimating RSAP (Figure 4). We also further only evaluate results obtained by the LSM dispersal model, as the GPM has been shown to portrait pollen dispersal and deposition unrealistically; in our study landscape, this concerns mainly trees such as Abies, Fagus and Quercus (Abraham et al., 2014; Theuerkauf et al., 2013). When comparing RPPs for LSM and GPM (see Table A.2, available online), we find overestimated values for Fagus, Quercus and Tilia when GPM was applied.

Discussion

Our study provides the first calculated RPP estimates based on field observations from a region of warm temperate vegetation situated at the western periphery of the forest-steppe biome in central-eastern Europe. In eastern Europe, such estimates are essential for creating reliable quantitative land-cover reconstructions (Feurdean et al., 2015; Kuneš et al., 2015). Thus, this study provides a new set of parameters that opens new opportunities in answering ecological questions dependent on past vegetation reconstructions.

Performance of locally derived RPPs in a patchy landscape

Our study provides the first estimates of RSAP using the LSM. RPPL were estimated at comparable RSAP as in other studies using similar vegetation mapping methods (Baker et al., 2016; Mazier et al., 2008; Yuecong et al., 2017). There are also studies estimating much larger RSAP originating from generally treeless vegetation types (Räsänen et al., 2007; Xu et al., 2014), but also from some forested landscapes (Li et al., 2015), or patchy cultural landscape (Abraham and Kozáková, 2012). Most studies attribute larger than expected RSAP to the uneven distribution of rare vegetation types (Li et al., 2015; Nielsen and Sugita, 2005). In case of the previous Czech RPP study conducted in similar habitat conditions, the higher RSAP estimates may be attributed not only to an uneven distribution of different vegetation types, but also to a lower effort invested in vegetation mapping (Abraham and Kozáková, 2012)

Much smaller RSAP than in our study region was obtained in semi-cultural landscapes of Scandinavia and eastern China (Broström et al., 2004; Bunting and Hjelle, 2010; Li et al., 2017). Small RSAP is usually connected with fine-grain homogeneous vegetation. However, patchy semi-cultural landscapes in eastern China still provided low RSAP (145 m), perhaps due to coarse-grain satellite-generated vegetation maps. RSAP was found to be only few metres based on estimating the RPP of a few herb taxa in Norway (Bunting and Hjelle, 2010).

RPPLs of the most important tree taxa in our study area, Quercus, Tilia, Carpinus and Corylus, are strikingly different from values from Białowieża forest in Poland (Baker et al., 2016), where RPPs reached fairly high values for all tree taxa. The authors attribute this to differences in vegetation structure, arguing that their landscape reflects mostly a closed canopy forest similar to natural conditions in the past, while other studies come from fragmented cultural landscapes (Abraham and Kozáková, 2012; Broström et al., 2004; Li et al., 2017). However, in our case, one would also expect high RPPL values for trees given the much higher predicted pollen production of solitary trees (Aaby, 1988; Figure 7) than those growing in closed canopy forest (Feeser and Dörfler, 2014). The fact that the opposite was found can be explained by compensation of herb taxa yielding much higher productivity in semi-open landscape. The most similar value for Quercus was obtained by another Czech study conducted in a cultural landscape with similar climatic conditions (Abraham and Kozáková, 2012). However, Tilia yielded slightly lower values in this study than in Central Bohemia, while values for the rest of the trees cannot be compared due to missing values in both regions.

Figure 7.

Patchy vegetation structure of the White Carpathians with solitary trees.

Among herbs, the most striking RPP value in our study was found for Plantago lanceolata (9.84). It has quite a small pollen grain which can easily disperse in a semi-open or open landscape such as our study. From our observations, only a small number of plants contribute to its very high pollen representation in the pollen spectra (Figure 2). In case of the site with highest pollen proportion, we might have recorded dispersion from a nearby plant individual, which is beyond sensitivity of the dispersal models used. These characteristics also make it the best anthropogenic indicator during the first deforestation phases in the Neolithic (Kuneš and Abraham, 2017). Such a high productivity of P. lanceolata has been found only in southern Sweden (Broström et al., 2004); in the other study from the Czech Republic, the value was lower (Abraham and Kozáková, 2012), but it is still characterizing P. lanceolata as high pollen producer.

Validation of RPPs through regional vegetation maps

Regional vegetation abundances allowed us to evaluate RPPL by another method and calculate RPPs for an additional 10 pollen taxa. Using such a large study area enabled us to involve taxa (see below) that could not be included because of their low local representation in the ERV model. One of the requirements for the ERV model is a sufficient site-to-site variability in pollen proportions, that is, sufficient site-to-site variation in the local vegetation (Sugita, 1994). Alnus and Betula pollen had low variations (1–12%) in the moss cushion dataset (Figure 2). Pinus pollen percentages had a higher variation (6–47%), but actual representation in the local vegetation was much lower than in the broader region (Figure 3). We attempted to include Pinus in an initial ERV 1 model run, which provided an unreliable RPPL and a high background component of Pinus. We interpreted the variation in the moss cushion pollen data of Alnus, Betula and Pinus as stochasticity of the background component produced by populations beyond the considered radius of 500 m. Pollen percentages of Picea also showed some stochastic variation, that is, two forest sites (id: 9, 18) had similar pollen percentage (12–18%) as three non-forest sites (id: 20, 33, 12; Figure 2), but at the same time, it was recorded representatively at the local scale (10–70% in vegetation at seven sites; Figure 5a) and overall vegetation proportion at both local and regional scales was similar (Figure 3). The ERV demonstrates its main strength when the site-to-site variation of pollen corresponds to the vegetation pattern within the local vegetation survey. Using only local vegetation data to estimate RPP, ERV is able to subtract the regional component from the calculation. The REVEALS model, however, uses the regional vegetation data for calculating pollen-vegetation relationship requiring one large site or many small sites to overcome the local effect. Thus, we find RPPL and RPPR methodologically analogous and mutually comparable. Moreover, numerous studies use parameters obtained through the ERV model to estimate regional vegetation (e.g. Hellman et al., 2008; Soepboer et al., 2010).

RPPR of the taxa discussed above (Alnus, Betula and Pinus) ranks them as medium pollen producers, which is very similar to the values obtained in northern Germany (Theuerkauf et al., 2013). Abies also belongs among taxa present only at the regional scale, but its RPPR is hardly comparable with any previous results, for example, very high RPPR in Abraham et al. (2014). As a big and heavy pollen grain, such discrepancy is because of the use of the GPM that biases dispersal of large pollen grains such as Abies (Abraham et al., 2014; Mariani et al., 2016; Theuerkauf et al., 2013).

Despite Brassicaceae and Cyperaceae yield still quite high background component according to the ERV model (Figure 5b), their pollen-vegetation relationship already allows RPP estimation according to both methods, that is, Brassicaceae at five sites gain 3–10% and Cyperaceae at 10 sites 5–15% in the local vegetation (Figure 5a). Similar overall vegetation abundance at both spatial scales provided similar RPPL and RPPR, Brassicaceae resulted as low and Cyperaceae as medium pollen producer. The rest of herbs, including Anthemis-type, Rubiaceae, Asteraceae subf. Cichorioideae, Ranunculus acris-type and Senecio-Typ, resulted in an analogous pattern having similar vegetation abundances and RPPs at both scales.

Plantago lanceolata-type is the only herb ranked as a high pollen producer according to both methods and other studies (see above). Its high RPPL could be explained by an outlier value of pollen proportion > 0.45 which significantly influences the slope estimate in the ERV model, but averaging in the REVEALS model might reduce its effect. However, exclusion of this datapoint led to similar RPPL value 5–20 depending on the distance weighting. The difference between RPPR and RPPL must be a result of size of our rings (too coarse) and scale of pollen dispersion. Plantago lanceolata grows directly in 70% of the surveyed non-forest vegetation plots (within first 0.5 m).

Corylus, Acer, Tilia and Carpinus have higher vegetation abundances at the local than the regional scale (Figure 3b). This discrepancy is followed by a higher RPPR than locally calculated RPPL, which is likely a consequence of our sampling. Pollen assemblages reflect the local vegetation pattern, which is relatively exceptional compared with the rest of the region (Figure 3a), and thus are not representative for the REVEALS-based RPPR estimate. This finding has certain importance on the selection of the sites and taxa for the RPPR estimate. REVEALS model, but also LRA and ERV models, respectively (Sugita, 1994, 2007), assume that the species are distributed within the vegetation mosaic randomly or representatively in order that their overall proportion around sites is similar to the proportion in the region (see our herb taxa). However, fossil pollen sites are usually surrounded by specific habitats, which can host some target species from the vegetation mosaic, for example, Alnus, Pinus and Cyperaceae. This skewed representation towards the local scale produces pollen signal of the taxa concerned, which may lead to biased results using algorithms based on the above mentioned assumption. Corylus, Acer, Tilia and Carpinus in this study have low RPPLs and high RPPRs. REVEALS models base the calculation on the high pollen signal and scarce regional vegetation. The result shows that the unfulfilled requirement of random distribution matters in the REVEALS model. RPPRs are biased correspondingly to the difference between local and regional vegetation. The effect of the vegetation gradients on the RPPLs is smaller since pollen assemblage from small sites always corresponds to local vegetation data. Taxa constantly lacking in the wetland habitats, like Pinus, Alnus or Betula in this study (see above), do not fulfil the random pattern; however, they are suitable for the REVEALS model, because their regional component is not shaded by the local signal.

Quercus represents an important dominant in our local and regional vegetation and, therefore, it is absolutely vital to understand its pollen-vegetation relationship. It is determined as medium pollen producer in our local study, while the regional estimates put it rather among high pollen producers (the absolute difference is not too big). Using similar large-scale vegetation dataset, Quercus was identified as strong pollen producer in Germany (Theuerkauf et al., 2013). Similarly, Quercus was found to be strong at local-scale studies in Poland (Baker et al., 2016), Sweden (Broström et al., 2004), England (Bunting et al., 2005) and two regions in China (Li et al., 2015, 2017).

Vegetation structure is another factor that might influence RPP estimates. Forests grow in patchiness in a traditional cultural landscape with grasslands, which influences pollen production of trees. Pollen production of Poaceae can be influenced by management of grasslands, hay making by scythe around our sampling sites.

Conclusion

Our study identified Plantago lanceolata-type (3–10) as a high pollen producer, Quercus (1.5–2) as medium-to-high, Asteraceae subf. Cichorioideae, Anthemis-type, Ranunculus acris-type and Rubiaceae (ca. 0.5) as low-to-medium, and Brassicaceae (ca. 0.1–0.2) as low pollen producers. We compared these values using two approaches of calculation at different spatial scales, locally in the traditionally managed semi-natural landscape and regionally in the intensively managed landscape.

We consider these RPPL values as robust and suitable for the use in Holocene vegetation reconstruction even at regional scale. RPPRs of the tree taxa (Acer, Carpinus, Corylus, Fagus and Tilia) are influenced by their higher vegetation abundance at the local scale. Comparison of RPPL and RPPR emphasizes the importance of the same overall representation in vegetation at both local and regional scales. RPPL can be used in the REVEALS model; however, when multiple small sites are used instead of one large site, their local vegetation in average should represent the regional and no patterning should appear with increasing distance.

This can be hardly fulfilled by taxa, which are distributed systematically around the small fossil sites and at the same time, they are important for the terrestrial ecosystem (Pinus, Picea, Betula and Poaceae). This kind of uneven vegetation pattern hinders potential application of quantitative vegetation reconstruction at small sites assuming two pollen components from local and regional sources. Any pollen-based quantitative past vegetation reconstructions in the studied region must face this obstacle since large lakes are not available here.

Increasing availability of large-scale vegetation data in the future will help to more profoundly understand the importance of regional pollen component and means of past vegetation reconstruction in such sub-optimal settings.

Supplemental Material

AppendixB – Supplemental material for Relative pollen productivity estimates for vegetation reconstruction in central-eastern Europe inferred at local and regional scales

Supplemental material, AppendixB for Relative pollen productivity estimates for vegetation reconstruction in central-eastern Europe inferred at local and regional scales by Petr Kuneš, Vojtěch Abraham, Barbora Werchan, Zuzana Plesková, Karel Fajmon, Eva Jamrichová and Jan Roleček in The Holocene

Supplemental Material

AppendixC – Supplemental material for Relative pollen productivity estimates for vegetation reconstruction in central-eastern Europe inferred at local and regional scales

Supplemental material, AppendixC for Relative pollen productivity estimates for vegetation reconstruction in central-eastern Europe inferred at local and regional scales by Petr Kuneš, Vojtěch Abraham, Barbora Werchan, Zuzana Plesková, Karel Fajmon, Eva Jamrichová and Jan Roleček in The Holocene

Supplemental Material

RPP-AppendixA-revision – Supplemental material for Relative pollen productivity estimates for vegetation reconstruction in central-eastern Europe inferred at local and regional scales

Supplemental material, RPP-AppendixA-revision for Relative pollen productivity estimates for vegetation reconstruction in central-eastern Europe inferred at local and regional scales by Petr Kuneš, Vojtěch Abraham, Barbora Werchan, Zuzana Plesková, Karel Fajmon, Eva Jamrichová and Jan Roleček in The Holocene

Footnotes

Acknowledgements

The authors thank Martin Theuerkauf for the help with the setting of the DEoptim function. They also thank Shinya Sugita for providing them a new version of the ERV programme. The authors are grateful to the following colleagues who kindly helped during any stage of the fieldwork: Přemysl Bobek, Markéta Chudomelová, Pavel Daněk, Pavel Dřevojan, Michelle Farrell, Radim Hédl, Radka Kozáková, Pavel Novák and Helena Prokešová. They thank the Nature Conservation Agency of the Czech Republic to be able to use Natura 2000 habitat mapping data (http://mapmaker.nature.cz/wmsconnector/com.esri.wms.Esrimap/aopk_biotopy_wms). Forest Management Institute is acknowledged for using forest management plans (), and PLA administration Bílé Karpaty is acknowledged for allowing this work in the area.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship and/or publication of this article. This work was supported by the Czech Science Foundation (Grant No. 16-10100 S). Jan Roleček was partly supported by the long-term developmental project of the Czech Academy of Sciences (RVO 67985939).

ORCID iDs

Petr Kuneš

Vojtěch Abraham

Supplemental material

Supplemental material for this article is available online.

References

Aaby

(1988) The cultural landscape as reflected in percentage and influx pollen diagrams from two Danish ombrotrophic mires. In: Birks

Birks

HJB

Kaland

, et al. (eds) The Cultural Landscape – Past, Present and Future. Cambridge: Cambridge University Press, pp. 209–228.

Abraham

Kozáková

(2012) Relative pollen productivity estimates in the modern agricultural landscape of Central Bohemia (Czech Republic). Review of Palaeobotany and Palynology 179: 1–12.

Abraham

Oušková

Kuneš

(2014) Present-day vegetation helps quantifying past land cover in selected regions of the Czech Republic. PLoS ONE 9(6): e100117.

Baker

Zimny

Keczyński

, et al. (2016) Pollen productivity estimates from old-growth forest strongly differ from those obtained in cultural landscapes: Evidence from the Białowieża National Park, Poland. The Holocene 26(1): 80–92.

Beug

(2004) Leitfaden der Pollenbestimmung für Mitteleuropa und angrenzende Gebiete. München: Verlag Dr. Friedrich Pfeil.

Broström

Nielsen

Gaillard

M-J

, et al. (2008) Pollen productivity estimates of key European plant taxa for quantitative reconstruction of past vegetation: A review. Vegetation History and Archaeobotany 17(5): 461–478.

Broström

Sugita

Gaillard

(2004) Pollen productivity estimates for the reconstruction of past vegetation cover in the cultural landscape of southern Sweden. The Holocene 14(3): 368–381.

Bunting

Armitage

Binney

, et al. (2005) Estimates of ‘relative pollen productivity’ and ‘relevant source area of pollen’ for major tree taxa in two Norfolk (UK) woodlands. The Holocene 15(3): 459–465.

Bunting

Hjelle

(2010) Effect of vegetation data collection strategies on estimates of relevant source area of pollen (RSAP) and relative pollen productivity estimates (relative PPE) for non-arboreal taxa. Vegetation History and Archaeobotany 19(4): 365–374.

10.

Bunting

Farrell

Broström

, et al. (2013) Palynological perspectives on vegetation survey: A critical step for model-based reconstruction of Quaternary land cover. Quaternary Science Reviews 82: 41–55.

11.

Chytrý

(2012) Vegetation of the Czech Republic: Diversity, ecology, history and dynamics. Preslia 84: 427–504.

12.

Eisenhut

(1961) Untersuchungen über die Morphologie und Ökologie der Pollenkörner heimischer und fremdländischer Waldbäume (translated into English by Jackson

Jaumann

, 1989). Hamburg: Paul Parey.

13.

Faegri

Kaland

Krzywinski

(1989) Textbook of Pollen Analysis. 4th Edition. Caldwell, NJ: The Blackburn Press.

14.

Feeser

Dörfler

(2014) The glade effect: Vegetation openness and structure and their influences on arboreal pollen production and the reconstruction of anthropogenic forest opening. Anthropocene 8: 92–100.

15.

Feurdean

Marinova

Nielsen

, et al. (2015) Origin of the forest steppe and exceptional grassland diversity in Transylvania (central-eastern Europe). Journal of Biogeography 42(5): 951–963.

16.

Filipova-Marinova

Kvavadze

Connor

, et al. (2010) Estimating absolute pollen productivity for some European Tertiary-relict taxa. Vegetation History and Archaeobotany 19(4): 351–364.

17.

Gaillard

M-J

Sugita

Mazier

, et al. (2010) Holocene land-cover reconstructions for studies on land cover-climate feedbacks. Climate of the Past 6(4): 483–499.

18.

Hájková

Jamrichová

Petr

, et al. (2018) Persistence of a vegetation mosaic in a peripheral region: Could turbulent medieval history disrupt Holocene continuity of extremely species-rich grasslands? Vegetation History and Archaeobotany 27(4): 591–610.

19.

Hellman

SEV

Gaillard

M-J

Broström

, et al. (2008) The REVEALS model, a new tool to estimate past regional plant abundance from pollen data in large lakes: Validation in southern Sweden. Journal of Quaternary Science 23(1): 21–42.

20.

Hennekens

Schaminée

JHJ

(2001) TURBOVEG, a comprehensive data base management system for vegetation data. Journal of Vegetation Science 12(4): 589–591.

21.

Jongepierová

(2008) Louky Bílých Karpat [Grasslands of the White Carpathian Mountains]. Veselí nad Moravou: ZO ČSOP Bílé Karpaty.

22.

Kuneš

Abraham

(2017) History of Czech vegetation since the late Pleistocene. In: Chytrý

Danihelka

Kaplan

, et al. (eds) Flora and Vegetation of the Czech Republic. Cham: Springer, pp. 193–227.

23.

Kuneš

Svobodová-Svitavská

Kolář

, et al. (2015) The origin of grasslands in the temperate forest zone of east-central Europe: Long-term legacy of climate and human impact. Quaternary Science Reviews 116: 15–27.

24.

Kuparinen

Markkanen

Riikonen

, et al. (2007) Modeling air-mediated dispersal of spores, pollen and seeds in forested areas. Ecological Modelling 208: 177–188.

25.

Gaillard

M-J

Sugita

, et al. (2017) Relative pollen productivity estimates for major plant taxa of cultural landscapes in central eastern China. Vegetation History and Archaeobotany 26(6): 587–605.

26.

Nielsen

Zhao

, et al. (2015) Pollen production estimates (PPEs) and fall speeds for major tree taxa and relevant source areas of pollen (RSAP) in Changbai Mountain, northeastern China. Review of Palaeobotany and Palynology 216: 92–100.

27.

Mariani

Connor

Theuerkauf

, et al. (2016) Testing quantitative pollen dispersal models in animal-pollinated vegetation mosaics: An example from temperate Tasmania, Australia. Quaternary Science Reviews 154: 214–225.

28.

Marquer

Gaillard

M-J

Sugita

, et al. (2017) Quantifying the effects of land use and climate on Holocene vegetation in Europe. Quaternary Science Reviews 171: 20–37.

29.

Mazier

Broström

Gaillard

M-J

, et al. (2008) Pollen productivity estimates and relevant source area of pollen for selected plant taxa in a pasture woodland landscape of the Jura Mountains (Switzerland). Vegetation History and Archaeobotany 17(5): 479–495.

30.

Mazier

Gaillard

M-J

Kuneš

, et al. (2012) Testing the effect of site selection and parameter setting on REVEALS-model estimates of plant abundance using the Czech Quaternary Palynological Database. Review of Palaeobotany and Palynology 187: 38–49.

31.

Mullen

Ardia

Gil

, et al. (2011) DEoptim: An R package for global optimization by differential evolution. Journal of Statistical Software 40: 1–26.

32.

Nielsen

Sugita

(2005) Estimating relevant source area of pollen for small Danish lakes around AD 1800. The Holocene 15(7): 1006–1020.

33.

Parsons

Prentice

(1981) Statistical approaches to R-values and the pollen–vegetation relationship. Review of Palaeobotany and Palynology 32(2–3): 127–152.

34.

Prentice

(1985) Pollen representation, source area, and basin size: Toward a unified theory of pollen analysis. Quaternary Research 23(1): 76–86.

35.

Prentice

Parsons

(1983) Maximum likelihood linear calibration of pollen spectra in terms of forest composition. Biometrics 39(4): 1051–1057.

36.

Punt

Blackmore

Hoen

, et al. (eds) (1976–2009) The Northwest European Pollen Flora. Volumes 1–9. Amsterdam: Elsevier.

37.

R Core Team (2017) R: A Language and Environment for Statistical Computing. Vienna: R Foundation for Statistical Computing.

38.

Räsänen

Suutari

Nielsen

(2007) A step further towards quantitative reconstruction of past vegetation in Fennoscandian boreal forests: Pollen productivity estimates for six dominant taxa. Review of Palaeobotany and Palynology 146(1–4): 208–220.

39.

Rasser

Harzhauser

(2008) Palaeogene and Neogene. In: McCann

(ed.) The Geology of Central Europe. London: The Geological Society, pp. 1031–1139.

40.

Reille

(1998 (1992)) Pollen et spores d’Europe et d’Afrique du Nord. Marseille: Laboratoire de Botanique historique et Palynologie.

41.

Roberts

Fyfe

Woodbridge

, et al. (2018) Europe’s lost forests: A pollen-based synthesis for the last 11,000 years. Scientific Reports 8(1): 716.

42.

Roleček

Čornej

Tokarjuk

(2014) Understanding the extreme species richness of semi-dry grasslands in east-central Europe: A comparative approach. Preslia 86(1): 13–34.

43.

Seddon

AWR

Mackay

Baker

, et al. (2014) Looking forward through the past: Identification of 50 priority research questions in palaeoecology. Journal of Ecology 102(1): 256–267.

44.

Sjögren

(2013) A pilot study on absolute pollen productivity estimates from mires in Norway. The Holocene 23(10): 1487–1492.

45.

Soepboer

Sugita

Lotter

(2010) Regional vegetation-cover changes on the Swiss Plateau during the past two millennia: A pollen-based reconstruction using the REVEALS model. Quaternary Science Reviews 29(3–4): 472–483.

46.

Stockmarr

(1971) Tablets with spores used in absolute pollen analysis. Pollen et spores 13(4): 615–621.

47.

Sugita

(1994) Pollen representation of vegetation in Quaternary Sediments: Theory and method in patchy vegetation. Journal of Ecology 82(4): 881–897.

48.

Sugita

(2007) Theory of quantitative reconstruction of vegetation I: Pollen from large sites REVEALS regional vegetation composition. The Holocene 17(2): 229–241.

49.

Sugita

Gaillard

Broström

(1999) Landscape openness and pollen records: a simulation approach. Holocene 9(4): 409–421.

50.

Sugita

Hicks

Sormunen

(2010) Absolute pollen productivity and pollen-vegetation relationships in northern Finland. Journal of Quaternary Science 25(5): 724–736.

51.

Theuerkauf

Couwenberg

Kuparinen

, et al. (2016) A matter of dispersal: REVEALS in R introduces state-of-the-art dispersal models to quantitative vegetation reconstruction. Vegetation History and Archaeobotany 25(6): 541–553.

52.

Theuerkauf

Kuparinen

Joosten

(2013) Pollen productivity estimates strongly depend on assumed pollen dispersal. The Holocene 23: 14–24.

53.

Tolasz

Míková

Valeriánová

, et al. (2007) Atlas podnebí Česka [Climate Atlas of Czechia]. Praha: ČHMÚ, UPOL.

54.

Trondman

A-K

Gaillard

M-J

Mazier

, et al. (2015) Pollen-based quantitative reconstructions of Holocene regional vegetation cover (plant-functional types and land-cover types) in Europe suitable for climate modelling. Global Change Biology 21(2): 676–697.

55.

van der Knaap

van Leeuwen

JFN

Svitavská-Svobodová

, et al. (2010) Annual pollen traps reveal the complexity of climatic control on pollen productivity in Europe and the Caucasus. Vegetation History and Archaeobotany 19(4): 285–307.

56.

Wilson

Peet

Dengler

, et al. (2012) Plant species richness: The world records. Journal of Vegetation Science 23(4): 796–802.

57.

Cao

Tian

, et al. (2014) Relative pollen productivities of typical steppe species in northern China and their potential in past vegetation reconstruction. Science China Earth Sciences 57(6): 1254–1266.

58.

Yuecong

Yawen

Bunting

, et al. (2017) Relative pollen productivities and relevant source area of pollen in the forest–steppe ecotone of northern China. Review of Palaeobotany and Palynology 244: 1–12.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.01 MB

0.79 MB

0.55 MB