Spatial scaling of pollen-based alpha and beta diversity within forest and open landscapes of Central Europe

Pollen is an abundant fossil and the most common proxy for plant diversity during the Holocene. Based on datasets in open, forest, and mixed habitats, we used the spatial distribution of floristic diversity to estimate the source area of pollen diversity and identify factors influencing the significance of this relationship. Our study areas are Bohemian-Moravian Highlands and White Carpathians (the Czech Republic and the Slovak Republic). Sampling 60 sites in forest and open habitats in two study regions with contrasting floristic diversity, we calculated taxonomic richness (alpha diversity) and total spatial variance (beta diversity) for pollen and floristic data along two transects, each 1 km long. Following this, we calculated the correlation between floristic and pollen diversity. We also assessed the consistency of the relationship in different habitats. Finally, we regressed local contributions of individual sites to the beta diversity of pollen and floristic data in each of the regions. There was a positive correlation between pollen and floristic richness in both habitats in both regions; open and mixed datasets were significant. The highest correlation (adjusted R2) mostly occurred within the first tens of metres (1.5–70) and then within the first hundreds of metres (250–550). Variances of pollen data significantly correlated with variances of floristic data between 100 and 250 m. Local contributions to beta diversity of pollen and plants significantly correlated in the forest and one of the mixed datasets. Floristic richness at the pollen site and position of the site within the landscape structure determine the sequence of the appearing species in the increasing distance. The number of species sets the source area of pollen richness and dissimilarity of appearing species controls the source area of pollen variance. These findings, linking pollen and floristic diversity, provide an essential stepping-stone for the reconstruction of historic plant diversity.


3 8
The south-western White Carpathian Mts. are situated on the periphery of a forest-steppe without the contribution of wind dispersal (Sugita 1994). We tried to avoid overlapping of 1 5 5 sampling sites and simultaneously kept the sampled area compact and homogeneous in terms  We collected pollen samples from a moss cushion of at least 50 cm 2 in the central point of same year as the pollen data (Table A2). The effort of vegetation sampling was spread into 1 6 0 three zones. Within the first 10 m, we recorded complete species lists in both regions; 1 6 1 however, in WCM, 21 additional plots of 1 m 2 were sampled following a modified help of aerial photographs. The occurrence of additional species not present in the first 10 m was recorded for each of the mapped polygons. Up to 1,000 m, we recorded additional plant 1 6 6 5 species, and vegetation types were mapped along two 20-metre-wide linear transects.

6 7
Directions of the transects were chosen based on the aerial map to cover the highest possible 1 6 8 habitat diversity. At the same time, the two transects had a minimum angular distance of 90°.

7 2
Based on the collected data, we compiled six datasets: two "uniform datasets" with forest and  Moss polsters were prepared for pollen analysis using standard procedures (Faegri et al.,  The pollen concentrate was stored in glycerine or silicone. Pollen slides were counted under 1 8 0 the light microscope at 400× magnification; for selected taxa at 1,000× magnification. The  Due to the varying pollen sum across samples (between 943 to ca. 4,000 grains), we unified  The number of pollen taxa (pollen richness) was regressed against floristic richness. Although scale as an alpha diversity. The concept of beta diversity in ecology is less equivocal, and 1 9 0 there are many definitions and corresponding ways as to how to calculate beta diversity table. We used the Jaccard index on presence-absence data as its measure. The relative 1 9 5 character of BDtotal ranging from 0 to 1 allows for different numbers of sites, thus also 1 9 6 enabling comparison between mixed and uniform datasets. Pollen BDtotal values calculated 1 9 7 for our six datasets were regressed against six floristic BDtotals at different distances from beta diversity (hereafter also "local contribution") and its significance. We explored the   In both regions, we found 169 pollen types (95 in BMH and 151 in WCM) and 1,323 plant 2 1 0 species (799 in BMH and 1,098 in WCM). Mean pollen richness per sample varied between 2 1 1 50 pollen types in WCM meadows, 42 pollen types in WCM forest, 38 pollen types in BMH 2 1 2 meadows, and 31 pollen types in BMH forest (Fig. 2). Floristic richness followed the same 2 1 3 order as pollen richness at a distance between 600 and 1,000 m: in WCM, forest species 2 1 4 gradually increased along the whole transect, with the lowest richness between 40 and 200 m. In other datasets, however, the increase was more irregular, with more than half of the species 2 1 6 appearing already within the first 100 m (Fig. 3). BDtotal in mixed datasets of pollen and plants was always higher than in uniform datasets.  Pollen and floristic richness values were lower in forests than in meadows; however, BDtotals 2 2 5 were higher in the forest than in meadow datasets. All datasets showed a positive correlation of pollen and floristic richness for at least some  (Table A3). All WCM). Mixed datasets had higher adjusted R 2 than their uniform subsets, except for BMH 2 3 6 regions between 40 and 200 m (Fig. 4a). The average distance of maximum adjusted R 2 for 2 3 7 six compared datasets was 286 m. Adjusted R 2 between pollen and floristic richness showed two general ranges of distances 2 3 9 where the correlation was high. The first, within tens of metres from the central points,  (Fig. 4a). At this distance, most species naturally appear for the first time (Fig. 5). The showed the best fit at 70 m, where species confined to forest roads frequently appear (Fig. 5). The second range of maximum adjusted R 2 values appeared between 400 and 550 in BMH  BMH forest largely originated from human-made habitats (forest roads between 10 and 100 2 5 5 m, and built-up areas usually at a distance above 500 m), whereas WCM forests are enriched 2 5 6 by meadows and other semi-natural habitats, usually at a distance above 200 m (Fig. 5).

5 7
Relationship between pollen and floristic variance, calibration of beta diversity 2 5 8 The highest adjusted R 2 between pollen and floristic BDtotal was identified at 150 m. The 2 5 9 significant correlation appeared between 100 and 250 m, a remarkable high, but the 2 6 0 insignificant correlation appeared between 300 and 600 m (Fig. 6a). The floristic BDtotal of 2 6 1 WCM meadow at 150 m was lower than the floristic BDtotal of the rest of the datasets 2 6 2 concerning the linear relationship to the pollen BDtotal (Fig. 6b). A distance of 150 m 2 6 3 followed the steep decrease of floristic BDtotal at 10-100 m (Fig. 3), when most of the taxa 2 6 4 appeared, and fell between both ranges of maximum adjusted R 2 of the richness regression. forest datasets at 100-400 m and in the mixed dataset from the BMH region at 900-1,000 m. datasets from the WCM region did not show any relationship (Fig. 4b).  The source area of pollen richness, measured as adjusted R 2 between pollen and floristic 2 7 9 richness, is determined by numbers of new plant species appearing with increasing distance. The position of the pollen site within the landscape structure affects the order of the habitats 2 8 1 and the sequence of appearing species. The grain size of the landscape structure is smaller in 2 8 2 the WCM region than the BMH region; thus, the WCM region resembles a more even mosaic  despite the conceptual differences mentioned above: while RSAP calculation was based on 2 9 6 pollen/vegetation proportions of 17 taxa, here we deal with incidences of the whole spectra. Moreover, vegetation structure for the range between 10 and 1,000 m was recorded 2 9 8 independently in the two studies. We suggest that this robustness indicates the significant 2 9 9 effect of landscape structure on the pollen-vegetation relationship.

0 0
A critical characteristic of the landscape mosaic, affecting the source area of pollen richness, 3 0 1 appears to be the number of species per patch rather than its area (Fig. A4). The appearance of Matthias et al. (2015) and Meltsov et al. (2013) to the more general "species-rich patches".  The strong effect of high pollen richness in WCM meadows is also visible in the comparison variance and the pattern of the other datasets lay at 6 m ( Fig. 6b). Again, this may be caused 3 2 7 by the high fine-scale diversity of the meadows, which include most pollen types present in  implies that the resulting distance of 100-250 m represents all datasets. Though they differ in 3 3 5 species richness, openness, and habitats of origin, the relationship between variances is fairly linear. The only exception is the biodiversity hotspot of WCM meadows mentioned above. It shows that the spatial scale at which the pollen variance corresponds to the floristic variance  The mechanism of establishing the source area of pollen variance was similar to that high, yet insignificant relationship of the variances at the distance 250-600 m (Fig. 6a) 3 4 6 corresponds to the distance of the second range of fit between richness (Fig. 4a).

4 7
The link between pollen and total floristic variance is underlined by the relationship between 3 4 8 the amount of variance contributed by individual sites and the total variance. Indeed, these 3 4 9 amounts in pollen and floristic data are significantly correlated. Distances of the high 3 5 0 correlation of local contribution to total variance are related to the source area of pollen 3 5 1 richness of individual datasets. In WCM forest and BMH forest, the increase of correlation of 3 5 2 local contribution to beta diversity usually follows (or precedes) the richness correlation by a 3 5 3 single ring (Fig. 4).

5 4
Beta diversity understood as directional turnover (temporal or spatial) belongs to more pollen-based turnover correlates with forest-inventory-based turnover. We extend this finding 3 5 9 from woody taxa to all species and from directional turnover to non-directional variance. Moreover, forest sites with high contributions to pollen beta diversity also show a high 3 6 1 contribution to floristic beta diversity (Fig. 4b).