Relative pollen productivity estimates in the modern agricultural landscape of Central Bohemia (Czech Republic)

Abstract

We estimated relative pollen productivity estimates (PPE), key parameters for the quantitative interpretation of pollen data, for 13 taxa using modern pollen assemblages from 54 sites and recent vegetation data. Vegetation mapping in the area covered a minimum radius of 2 km around each sampling site. Vegetation data were weighted by the Prentice model, i.e. weighting by distance and by the dispersal–deposition parameters of different pollen types. PPE values were calculated by three submodels of the Extended R-value model. ERV 1 produced the best goodness of fit. The PPEs for Urtica and Sambucus nigra are published here for the first time, and the PPE for the Chenopodiaceae represents the first estimate for Europe. Values for the other ten taxa (Poaceae, Pinus, Salix, Fraxinus, Quercus, Tilia, Artemisia, Plantago lanceolata, Alnus and Cerealia) are comparable with or fall within the ranges of values published in previous studies. Herb taxa produce ca 3–11 times more pollen than the Poaceae. Herbs produce even more pollen than trees, whose production is 1–6 times higher than that of the Poaceae. The lowest pollen producers are the Cerealia, producing 20 times less pollen than the Poaceae. Our estimate of the relevant source area of pollen (RSAP) of 1050 m is relatively high compared to other studies in semi-open landscapes. This is possibly caused by the uneven pattern of some taxa in the vegetation mosaic (Pinus, P. lanceolata, Salix and Alnus). The distance of 1100 m, at which all taxa are present around each site, is similar to the RSAP distance (1050 m).

Introduction

The fossil pollen record can trace different environmental factors that caused vegetation changes in the past such as humans or the climate (Pokorný, 2005, Skrzypek et al., 2009). When the goal of palynological research is to know past vegetation abundances, it is critical to understand modern pollen–vegetation relationships that can be used as a basis for quantitative vegetation reconstruction. One of the key parameters in this relationship is pollen productivity. Only percentage data are available for a majority of Czech pollen profiles (Kuneš et al., 2009), so we had to deal with relative pollen productivity, which is expressed as “relative pollen productivity estimates”, hereafter referred to as PPE. The theoretical framework of its calculation was established with the introduction of the Extended R-value model (further referred to as the ERV model; Parsons and Prentice, 1981, Sugita, 1994), in combination with the maximum likelihood method (Prentice and Parsons, 1983, Bunting et al., 2004), can be used to estimate the relevant source area of pollen (RSAP; Sugita, 1994). This is a breakthrough in pollen analysis because it makes estimates of pollen productivity much more appropriate and allows for the study of the effects of vegetation structure on pollen deposition (e.g. Hellman et al., 2009a). Hence, pollen data become the temporal and spatial proxy for the vegetation.

The complexity of the relationship between vegetation and the pollen assemblage is influenced not only by the surrounding vegetation and pollen productivity but also, for example, by the taphonomy and the dispersive characteristics of individual pollen types or atmospheric conditions. Dispersion and deposition of pollen can be described using these factors as parameters in a mechanistic model. In our study, we used the Prentice model (Prentice, 1985), which is applied as distance weighting for all taxa on vegetation data. This means that not only distance (the highest importance is given to the vegetation growing close to a site) but also taxon-specific pollen dispersal properties (e.g. fall speeds) are taken into account.

Both the ERV model and the Prentice model were theoretically developed around the 1980s, but their use has risen mainly in the last decade. PPEs have now been calculated for many regions of Europe (Broström et al., 2008) as well as for North America (e.g. Calcote, 1995), Africa (Duffin and Bunting, 2007) and Asia (Li et al., 2011). In combination with the Landscape Reconstruction Algorithm (Sugita, 2007a, Sugita, 2007b) or the MultiScenario Approach (Bunting and Middleton, 2009), vegetation abundances have been estimated from the fossil pollen record in the Swiss Plateau (Soepboer et al., 2010), Southern Scandinavia (Nielsen and Odgaard, 2010, Kuneš et al., 2011) and northern England (Bunting et al., 2008).

The aim of the present study is to explore the relationship between pollen and vegetation in Central Bohemia and to produce reliable PPEs. Besides an early pioneer study in Bohemia (Križo, 1958), the modern pollen–vegetation relationship has been studied only in the forested mountain ranges along the state border using pollen traps (Pidek et al., 2010). Our interest, however, lies rather in quantifying past human–environment interactions in the central part of the country. This region has the longest human occupation history in all of Bohemia, which is also the reason why we focused on taxa classified as anthropogenic indicators (Behre, 1981). In Central Bohemia, archaeological exploration coupled with many pollen cores from abandoned meanders of the river Labe. All those studies offer a unique opportunity to combine archaeological knowledge with palaeoecological evidence of human impacts (Dreslerová and Pokorný, 2004).

A robust estimation of the relevant source area of pollen (Sugita, 1994) is critical for obtaining correct PPEs. However, the RSAP for sites of the same basin size under fixed atmospheric conditions is not controlled by different pollen productivity or the fall speed of individual taxa (Bunting et al., 2004), as one could expect, but by the spatial structure of the vegetation mosaic, as has been shown by previous simulations (Sugita, 1994, Sugita et al., 1999, Bunting et al., 2004, Broström et al., 2005, Gaillard et al., 2008, Hellman et al., 2009a, Hellman et al., 2009b). Since the spatial pattern and structure of vegetation affects the estimates of the RSAP using the ERV models, the two following assumptions have to be considered carefully in order to obtain reliable results. First, the ERV model requires vegetation heterogeneity to produce site-to-site variability in pollen loading, which is necessary for parameter estimation by the maximum likelihood method. The difference between heterogeneity and homogeneity is defined by the size of the patches surrounding a sedimentation basin — their size should be larger than the size of the basin. Second, the ERV model requires similar overall proportions and patch sizes of major plant taxa among the regions of the sites included into the study (Sugita, 1994). If pollen is sampled from moss polsters in the real landscape, then the first assumption can be met easily. The second, however, is not easy to meet even in a single study region, where the composition of taxa can differ among patches. Thus, the second aim of the present paper is to examine the properties of the vegetation structure in the real landscape in consideration of their possible influence on the ERV model. Some of these properties (e.g. patch size, position of the sample within the patch) have already appeared as parameters in some past simulations (Sugita, 1994, Sugita et al., 1999, Bunting et al., 2004, Broström et al., 2005, Gaillard et al., 2008, Hellman et al., 2009a, Hellman et al., 2009b), so we aim to discuss the results from simulations and the real landscape in order to find the key parameter of the vegetation mosaic which controls the RSAP. We can summarize the aims of this paper as follows: (i) to calculate PPEs for the agricultural landscape of Central Bohemia, (ii) to explore the distance of the RSAP and find the main factor influencing it, and (iii) to discuss the relationship between the vegetation mosaic and the sampling strategy and its possible effect on the ERV model.