Effects of the sampling design and selection of parameter values on pollen-based quantitative reconstructions of regional vegetation: a case study in southern Sweden using the REVEALS model

The need for quantification of land cover from pollen data has led to the development of a Landscape Reconstruction Algorithm (LRA). The LRA includes several models of which the REVEALS model estimates regional vegetation abundance using pollen assemblages from large sites (lakes or bogs). In this paper we explore the effects of selection and number of pollen samples, and choice of pollen productivity estimates on the REVEALS results. The effect of the size of vegetation surveys is also tested. The results suggest that the differences between two sizes of vegetation surveys have little effect on the model validation. The “characteristic radius” of regional vegetation in southern Sweden was estimated as 200 km. However, the vegetation composition in a 100 × 100 km² square matches well with that estimated by REVEALS. Whether 25, 20 (outliers excluded) or 4 pollen samples are used does not change the REVEALS reconstructions much although the error estimates are larger when outliers are included, and very large when only four samples are used. Therefore validation of the REVEALS model and REVEALS reconstructions of past vegetation can be performed using a limited number of pollen samples, although with caution. The use of many pollen samples from multiple sites is always better whenever possible. REVEALS reconstructions are closer to the actual vegetation when the Danish Pollen Productivity Estimates (PPEs) are used instead of the Swedish PPEs for Cereals, Rumex acetosa/acetosella, Plantago lanceolata and Calluna, indicating that the Danish PPEs are more reliable than the Swedish ones for those taxa. It is recommended to test more than one set of PPEs in validation and applications of the REVEALS model for a better evaluation of the results.     

Pollen productivity estimates of key European plant taxa for quantitative reconstruction of past vegetation: a review

Information on the spatial distribution of past vegetation on local, regional and global scales is increasingly used within climate modelling, nature conservancy and archaeology. It is possible to obtain such information from fossil pollen records in lakes and bogs using the landscape reconstruction algorithm (LRA) and its two models, REVEALS and LOVE. These models assume that reliable pollen productivity estimates (PPEs) are available for the plant taxa involved in the quantitative reconstructions of past vegetation, and that PPEs are constant through time. This paper presents and discusses the PPEs for 15 tree and 18 herb taxa obtained in nine study areas of Europe. Observed differences in PPEs between regions may be explained by methodological issues and environmental variables, of which climate and related factors such as reproduction strategies and growth forms appear to be the most important. An evaluation of the PPEs at hand so far suggests that they can be used in modelling applications and quantitative reconstructions of past vegetation, provided that consideration of past environmental variability within the region is used to inform selection of PPEs, and bearing in mind that PPEs might have changed through time as a response to climate change. Application of a range of possible PPEs will allow a better evaluation of the results.     

Reassessment of Holocene vegetation change on the upper Tibetan Plateau using the pollen-based REVEALS model

Previous studies based on fossil pollen data have reported significant changes in vegetation on the alpine Tibetan Plateau during the Holocene. However, since the relative proportions of fossil pollen taxa are largely influenced by individual pollen productivities and the dispersal characteristics, such inferences on vegetation have the potential to be considerably biased. We therefore examined the modern pollen–vegetation relationships for four common pollen species on the Tibetan Plateau, using Extended R-value (ERV) models. Assuming an average radius of 100 m for the sampled lakes, we estimated the relevant source area of pollen (RSAP) to be 2200 m (which represents the distance from the lake). Using Poaceae as the reference taxa (Pollen Productivity Estimate, PPE = 1), ERV Submodel 2 derived relative high PPEs for the steppe and desert taxa: 2.079 ± 0.432 for Artemisia and 5.379 ± 1.077 for Chenopodiaceae. Low PPEs were estimated for the Cyperaceae (1.036 ± 0.012), whose plants are characteristic of the alpine Kobresia meadows. Applying these PPEs to four fossil pollen sequences since the Late Glacial, the plant abundances on the central and north-eastern Tibetan Plateau were quantified using the “Regional Estimates of Vegetation Abundance from Large Sites” (REVEALS) model. The proportions of Artemisia and Chenopodiaceae were greatly reduced compared to their original pollen percentages in the reconstructed vegetation, owing to their high productivities and their dispersal characteristics, while Cyperaceae showed a relative increase in the vegetation reconstruction. The reconstructed vegetation assemblages of the four pollen sequence sites always yielded smaller compositional species turnovers than suggested by the pollen spectra, as revealed by Detrended Canonical Correspondence Analyses (DCCA) of the Holocene sections. The strength of the previously reported vegetation changes may therefore have been overestimated, which indicates the importance of taking into account pollen–vegetation relationships when discussing the potential drivers (such as climate, land use, atmospheric CO₂ concentrations) and implications (such as for land surface–climate feedbacks, carbon storage, and biodiversity) of vegetation change.     

Tree line identification from pollen data: beyond the limit?

Aim The boreal tree line is a prominent biogeographic feature, the position of which reflects climatic conditions. Pollen is the key sensor used to reconstruct past tree line patterns. Our aims in this study were to investigate pollen–vegetation relationships at the boreal tree line and to assess the success of a modified version of the biomization method that incorporates pollen productivity and dispersal in distinguishing the tree line. Location Northern Canada (307 sites) and Alaska (316 sites). Methods The REVEALS method for estimating regional vegetation composition from pollen data was simplified to provide correction factors to account for differential production and dispersal of pollen among taxa. The REVEALS‐based correction factors were used to adapt the biomization method and applied as a set of experiments to pollen data from lake sediments and moss polsters from the boreal tree line. Proportions of forest and tundra predicted from modern pollen samples along two longitudinal transects were compared with those derived from a vegetation map by: (1) a tally of ‘correct’ versus ‘incorrect’ assignments using vegetation in the relevant map pixels, and (2) a comparison of the shape and position of north–south forest‐cover curves generated from all transect pixels and from pollen data. Possible causes of bias in the misclassifications were assessed. Results Correcting for pollen productivity alone gave fewest misclassifications and the closest estimate of the modern mapped tree line position (Canada, + 300 km; Alaska, + 10 km). In Canada success rates were c. 40–70% and all experiments over‐predicted forest cover. Most corrections improved results over uncorrected biomization; using only lakes improved success rates to c. 80%. In Alaska success rates were 70–80% and classification errors were more evenly distributed; there was little improvement over uncorrected biomization. Main conclusions Corrected biomization should improve broad‐scale reconstructions of spatial patterns in forest/non‐forest vegetation mosaics and across climate‐sensitive ecotones. The Canadian example shows this is particularly the case in regions affected by taxa with extremely high pollen productivity (such as Pinus). Improved representation of actual vegetation distribution is most likely if pollen data from lake sediments are used because the REVEALS algorithm is based on the pollen dynamics of lake‐based systems.     

Seed dispersal in heterogeneous environments: bridging the gap between mechanistic dispersal and forest dynamics models

Seed dispersal is an important determinant of vegetation composition. We present a mechanistic model of seed dispersal by wind that incorporates heterogeneous vegetation structure. Vegetation affects wind speeds, a primary determinant of dispersal distance. Existing models combine wind speed and fall velocity of seeds. We expand on them by allowing vegetation, and thus wind profiles, to vary along seed trajectories, making the model applicable to any wind-dispersed plant in any community. Using seed trap data on seeds dispersing from forests into adjacent sites of two distinct vegetation structures, we show that our model was unbiased and accurate, even though dispersal patterns differed greatly between the two structures. Our spatially heterogeneous model performed better than models that assumed homogeneous vegetation for the same system. Its sensitivity to vegetation structure and ability to predict seed arrival when vegetation structure was incorporated demonstrates the model's utility for providing realistic estimates of seed arrival in realistic landscapes. Thus, we begin to bridge mechanistic seed dispersal and forest dynamics models. We discuss the merits of our model for incorporation into forest simulators, applications where such incorporation has been or is likely to be especially fruitful, and future model refinements to increase understanding of seed dispersal by wind.     

Evaluating the effect of flowering age and forest structure on pollen productivity estimates

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.     

Soil phosphorus as a control of productivity and openness in temperate interglacial forest ecosystems

Aim Observations of long chronosequences in forest ecosystems show that, after some millennia of build‐up, biomass declines in relation to the slow depletion of soil phosphorus. Plants that dominate during this period of soil impoverishment have specialized strategies for P acquisition, including ectomycorrhiza or root clusters. We use quantitative, pollen‐based reconstructions of regional vegetation in four Quaternary warm stages (Holocene, Eemian, Holsteinian, Harreskovian) to test whether inferred forest cover and productivity changes are consistent with long‐term modification of soil nutrient pools. Location Southern Scandinavia (Denmark, southern Sweden). Methods The REVEALS model was used to estimate regional vegetation abundances of 25 pollen‐type‐equivalent taxa from pollen records of large sedimentary basins in southernmost Scandinavia. Based on the estimated regional vegetation, we then calculated time‐series of Ellenberg indicator values for L (light), R (soil reaction) and N (a productivity proxy). We classified the vegetation records into distinct phases and compared these phases and the samples using hierarchical clustering and ordination. Results All three interglacials developed coniferous or mixed forests. However, pure deciduous forests were never reached during the Holsteinian, while pure coniferous forests never developed in the Holocene. Above‐ground productivity was inferred to be low initially, peaking in the first third of the warm stages and then slowly declining (except during the Holocene). Dominant trees of the post‐peak phases all had ectomycorrhiza as a strategy for P acquisition, indicating that easily accessible P pools had become depleted. Increases in fire regimes may have amplified the inferred final drop in productivity. Mid/late Holocene productivity changes were much influenced by agricultural activities. Main conclusions REVEALS vegetation estimates combined with Ellenberg indicator values suggest a consistent pattern in warm stages of initially rising productivity, followed by a long and slow decline. The P‐acquisition strategies of dominant trees indicate that the decline reflects increasing P depletion of soils.     

Pollen productivity estimates and relevant source area of pollen for selected plant taxa in a pasture woodland landscape of the Jura Mountains (Switzerland)

Relevant source area of pollen (RSAP) and pollen productivity for 11 key taxa characteristic of the pasture woodland landscape of the Jura Mountains, Switzerland, were estimated using pollen assemblages from moss polsters at 20 sites. To obtain robust pollen productivity estimates (PPEs), we used vegetation survey data at a fine spatial-resolution (1 × 1 m²) and randomized locations for sampling sites, techniques rarely used in palynology. Three Extended R value (ERV) submodels and three distance-weighting methods for plant abundance calculation were applied. Different combinations of the submodels and distance-weighting methods provide slightly different estimates of RSAP and PPEs. Although ERV submodel 1 using 1/d (d = distance in meters) best fits the dataset, PPE values for heavy pollen types (e.g. Abies) were sensitive to the method used for distance-weighting. Taxon-specific distance-weighting methods, such as Prentice’s model, emphasize the intertaxonomic differences in pollen dispersal and deposition, and are thus theoretically sound. For the dataset obtained in this project, Prentice’s model was more appropriate than other distance-weighting methods to estimate PPEs. Most of the taxa have PPEs equal to (Fagus, Plantago media and Potentilla-type), or higher (Abies, Picea, Rubiaceae and Trollius europaeus) than Poaceae (PPE = 1). Acer, Cyperaceae, and Plantago montana-type are low pollen producers. This set of PPEs will be useful for reconstructing heterogeneous, mountainous pasture woodland landscapes from fossil pollen records. The RSAP for moss polsters in this semi-open landscape region is ca. 300 m.     

A practical approximation of pollen dispersal–deposition for calculation of pollen productivity and quantification of vegetation

We compared simple mathematical pollen dispersal–deposition models with Gaussian plume models. The simple mathematical models proved equal or better approximations of real world pollen dispersal–deposition. We concluded that for most standard applications, such as estimating pollen productivity or quantitative vegetation reconstructions, simple mathematical models would perform satisfactory. Such easy-to-use models may lower the threshold to employ quantitative tools to palaeoecological questions.     

Are pollen records from small sites appropriate for REVEALS model-based quantitative reconstructions of past regional vegetation? An empirical test in southern Sweden

In this paper we test the performance of the Regional Estimates of VEgetation Abundance from Large Sites (REVEALS) model using pollen records from multiple small sites. We use Holocene pollen records from large and small sites in southern Sweden to identify what is/are the most significant variable(s) affecting the REVEALS-based reconstructions, i.e. type of site (lakes and/or bogs), number of sites, site size, site location in relation to vegetation zones, and/or distance between small sites and large sites. To achieve this objective we grouped the small sites according to (i) the two major modern vegetation zones of the study region, and (ii) the distance between the small sites and large lakes, i.e. small sites within 50, 100, 150, or 200 km of the large lakes. The REVEALS-based reconstructions were performed using 24 pollen taxa. Redundancy analysis was performed on the results from all REVEALS-model runs using the groups within (i) and (ii) separately, and on the results from all runs using the groups within (ii) together. The explanatory power and significance of the variables were identified using forward selection and Monte Carlo permutation tests. The results show that (a) although the REVEALS model was designed for pollen data from large lakes, it also performs well with pollen data from multiple small sites in reconstructing the percentage cover of groups of plant taxa (e.g. open land taxa, summer-green trees, evergreen trees) or individual plant taxa; however, in the case of this study area, the reconstruction of the percentage cover of Calluna vulgaris, Cyperaceae, and Betula may be problematic when using small bogs; (b) standard errors of multiple small-site REVEALS estimates will generally be larger than those obtained using pollen records from large lakes, and they will decrease with increasing size of pollen counts and increasing number of small sites; (c) small lakes are better to use than small bogs if the total number of small sites is low; and (d) the size of small sites and the distance between them do not play a major role, but the distance between the small sites and landscape/vegetation boundaries is a determinant factor for the accuracy of the vegetation reconstructions.     

Modelling pollen sedimentation in Danish lakes at c.ad 1800: an attempt to validate the POLLSCAPE model

Aims To validate the POLLSCAPE simulation model of pollen dispersal and deposition, and evaluate the effect of factors such as pollen productivity, wind speed and regional plant abundance, using a data set of ad 1800 pollen assemblages and historical land cover data. Location Denmark. Methods ad 1800 land cover from historical maps is digitized for 2000 m radii around 30 Danish lakes (3.5–33 ha). The simulation model POLLSCAPE is used to predict sedimentary pollen assemblages in the lakes from the plant abundance data inferred from these maps, with different model parameter settings for wind speed, pollen productivity, regional pollen loading, etc. The model predictions are compared with observed ad 1800 pollen assemblages from the lake sediment records. Furthermore, pollen productivity is estimated from the ad 1800 pollen and vegetation data using the Extended R‐value model. Results Generally the model reproduces the patterns in the observed pollen assemblages, and for most pollen types there are significant correlations between observed and predicted pollen proportions. The pollen proportions predicted by the POLLSCAPE model are sensitive to the pollen productivity estimates used, the regional background pollen loading and average wind speed. There is a difference in background pollen loading between eastern and western Denmark, especially of Calluna pollen. The fit between predicted and observed pollen assemblages is best at wind speeds around 2.5 m s^?1, and decreases rapidly at lower wind speeds. The pollen productivity estimates from the ad 1800 data set are comparable with estimates from moss polsters in modern analogues of traditional cultural landscapes in Sweden and Norway. Main conclusions The POLLSCAPE model reproduces the patterns in the observed pollen assemblages from the lakes well, considering the uncertainty in the historical plant abundance data. This study indicates that the simulation model can be a useful tool for investigating relationships between vegetation and pollen composition, but also that the simulated pollen proportions are sensitive to the pollen productivity estimates, the regional background and to wind speed.     

Using models of pollen dispersal and deposition in hilly landscapes: Some possible approaches

Recent developments in the modelling of pollen dispersal and deposition have led to the development of user-friendly computer software for simulating vegetation mosaics and pollen assemblages at specified points within those mosaics. In this paper we discuss the possible application of these approaches to modelling the pollen deposition in mountain areas. First, we demonstrate the use of the Multiple Scenario Approach by reconstructing mid-Holocene tree line position in the southern Dark Peak area of the Peak District in northern England from a published pollen record. However, the underlying model of pollen dispersal and deposition assumes a flat landscape, which makes extending the approach to mountainous areas problematic. Therefore we also present simulation experiments exploring the effects of modifying aspects of the existing model to better simulate the montane situation (ecotone structure, changes in the wind rose to mimic the effects of topography on aerial pollen transport). We suggest that, as a first approximation, topographic effects can be incorporated in the models by varying the wind rose at the sampling point to reflect the impact of topography on air flow.     

The development of composite dispersal functions for estimating absolute pollen productivity in the Swiss Alps

Considering the complexity of real-world pollen dispersal, a single set of parameters may be inadequate to model pollen dispersal, especially as dispersal occurs on both local and regional scales. Here we combine more than one dispersal function into a composite dispersal function (CDF). The function incorporates multiple parameters and different modes of pollen transportation, and thus has the potential to better simulate the relationship between deposited pollen and the surrounding vegetation than would otherwise be possible. CDFs based on different dispersal functions and combinations of dispersal functions were evaluated using a pollen-trap dataset from the Swiss Alps. Absolute pollen productivity (APP) was estimated at 7,700 ± 2,000 grains cm⁻² year⁻¹ for Larix decidua, 13,500 ± 1,900 grains cm⁻² year⁻¹ for Picea abies and 95,600 ± 17,700 grains cm⁻² year⁻¹ for Pinus cembra (with 95% confidence level). The results are consistent with previous APP estimates made from the same dataset using different methods.     

Pollen dispersal in spatially aggregated populations

We perform a theoretical study of effective pollen dispersal within plant populations exhibiting intraspecific spatial aggregation. We simulate nonuniform distributions of individuals by means of a Poisson cluster process and use an individual-based spatially explicit model of pollen dispersal to assess the effects of different aggregation patterns on the effective pollen pool size (N(ep)) and the axial variance of pollen dispersal (sigma (p)). Results show clear interactions between clumping and both N(ep) and sigma (p), whose precise form and intensity depend on the relative spatial scale of aggregation to pollen dispersal range. If clump size is small relative to dispersal range, clumping results in lower N(ep) and sigma (p) than in randomly distributed populations. Interestingly, by contrast, aggregation may actually enlarge N(ep) and has minimum impact on sigma (p) if clump size is near or above the scale of dispersal. High intraclump to global density ratios enhance the sensitivity of both N(ep) and sigma (p) to clumping, while leptokurtic pollen dispersal generates sharper reductions of both N(ep) and sigma (p) for small clump sizes and stronger increments of N(ep) for larger clump sizes. Overall, our results indicate that isolation-by-distance models in plants should not ignore the effects of intraspecific spatial aggregation on effective dispersal.     

Fire‐induced population reduction and landscape opening increases gene flow via pollen dispersal in Pinus halepensis

Population reduction and disturbances may alter dispersal, mating patterns and gene flow. Rather than taking the common approach of comparing different populations or sites, here we studied gene flow via wind‐mediated effective pollen dispersal on the same plant individuals before and after a fire‐induced population drop, in a natural stand of Pinus halepensis. The fire killed 96% of the pine trees in the stand and cleared the vegetation in the area. Thirteen trees survived in two groups separated by ~80 m, and seven of these trees had serotinous (closed) prefire cones that did not open despite the fire. We analysed pollen from closed pre and postfire cones using microsatellites. The two groups of surviving trees were highly genetically differentiated, and the pollen they produced also showed strong among‐group differentiation and very high kinship both before and after the fire, indicating limited and very local pollen dispersal. The pollen not produced by the survivors also showed significant prefire spatial genetic structure and high kinship, indicating mainly within‐population origin and limited gene flow from outside, but became spatially homogeneous with random kinship after the fire. We suggest that postfire gene flow via wind‐mediated pollen dispersal increased by two putative mechanisms: (i) a drastic reduction in local pollen production due to population thinning, effectively increasing pollen immigration through reduced dilution effect; (ii) an increase in wind speeds in the vegetation‐free postfire landscape. This research shows that dispersal can alleviate negative genetic effects of population size reduction and that disturbances might enhance gene flow, rather than reduce it.     

Limited seed dispersal shapes fine-scale spatial genetic structure in a Neotropical dioecious large-seeded palm

Seed and pollen dispersal contribute to gene flow and shape the genetic patterns of plants over fine spatial scales. We inferred fine-scale spatial genetic structure (FSGS) and estimated realized dispersal distances in Phytelephas aequatorialis, a Neotropical dioecious large-seeded palm. We aimed to explore how seed and pollen dispersal shape this genetic pattern in a focal population. For this purpose, we genotyped 138 seedlings and 99 adults with 20 newly developed microsatellite markers. We tested if rodent-mediated seed dispersal has a stronger influence than insect-mediated pollen dispersal in shaping FSGS. We also tested if pollen dispersal was influenced by the density of male palms around mother palms in order to further explore this ecological process in large-seeded plants. Rodent-mediated dispersal of these large seeds occurred mostly over short distances (mean 34.76 ± 34.06 m) while pollen dispersal distances were two times higher (mean 67.91 ± 38.29 m). The spatial extent of FSGS up to 35 m and the fact that seed dispersal did not increase the distance at which male alleles disperse suggest that spatially limited seed dispersal is the main factor shaping FSGS and contributes only marginally to gene flow within the population. Pollen dispersal distances depended on the density of male palms, decreasing when individuals show a clumped distribution and increasing when they are scattered. Our results show that limited seed dispersal mediated by rodents shapes FSGS in P. aequatorialis, while more extensive pollen dispersal accounts for a larger contribution to gene flow and may maintain high genetic diversity. Abstract in Spanish is available with online material.     

Selective interactions between short-distance pollen and seed dispersal in self-compatible species

In plants, genes may disperse through both pollen and seeds. Here we provide a first theoretical study of the mechanisms and consequences of the joint evolution of pollen and seed dispersal. We focus on hermaphroditic self-compatible species distributed in structured populations, assuming island dispersal of pollen and seeds among small patches of plants within large populations. Three traits are studied: the rate of among-patch seed dispersal, the rate of among-patch pollen dispersal, and the rate of within-patch pollen movement. We first analytically derive the evolutionary equilibrium state of each trait, dissect the pairwise selective interactions, and describe the joint three-trait evolutionary equilibrium under the cost of dispersal and kin competition. These results are then analytically and numerically extended to the case when selfed seeds suffer from depressed competitiveness (inbreeding depression, no heterosis). Finally individual-based simulations are used to account for a more realistic model of inbreeding load. Pollen movement is shown to generate opposite selection pressures on seed dispersal depending on spatial scale: within-patch pollen movement favors seed dispersal, whereas among-patch pollen dispersal inhibits seed dispersal. Seed dispersal selects for short-distance movements of pollen and it selects against long-distance dispersal. These interactions shape the joint evolution of these traits. Kin competition favors among-patch seed dispersal over among-patch pollen dispersal for low costs of within-patch pollen movement (and vice versa for significant costs of within-patch pollen movement). Inbreeding depression favors allogamy through high rates of within- and among-patch pollen movement. Surprisingly, it may select either for or against seed dispersal depending on the cost of among-patch pollen dispersal. Heterosis favors increased among-patch dispersal through pollen and seeds. But because these two stages inhibit each other, their joint evolution might lead to decreased seed dispersal in the presence of heterosis. Of crucial importance are the costs of dispersal.     

Quantitative representation of local forest composition in forest‐floor pollen assemblages

1 We compared modern pollen assemblages from 60 moss polster sites in northern New York with forest composition data within 20–120 m of the sites using extended R ‐value (ERV) models, which correct for non‐linearities arising from use of pollen percentage data. Our sites were concentrated in two regions, one dominated by Tsuga and hardwood ( Acer , Betula , Fagus ) forests, and the other by Tsuga , Pinus , Betula , Acer and Quercus forests.
2 Our results confirm that forest‐floor pollen assemblages are dominated by pollen originating from trees growing more than 20 m from the site of deposition. However, our results suggest that background pollen percentages were overestimated by Jackson & Wong in 1994, owing to unusually high Pinus pollen production in the year of their sampling.
3 Expansion of our vegetation sampling radius from 20 to 120 m resulted in modest but consistent improvement in model fit and a decrease in background pollen percentages.
4 ERV model parameters (slope and background) differed substantially between the two study regions, primarily owing to differences in background pollen productivity and dispersal from regional sources.
5 High background pollen percentages may lead to poor estimation of calibration parameters in regions of complex vegetation patterns. Expansion of the vegetation sampling radius to reduce the background component may lead to better parameter estimates.
6 Calibration of pollen–vegetation relationships requires definition of the vegetation term so that it approximates the vegetation sampled by the pollen assemblages. Critical challenges are to define better the appropriate vegetation sampling area and distance‐weighting functions for application to pollen–vegetation calibration.