Novel methods of estimating relative pollen productivity: A key parameter for reconstruction of past land cover from pollen records

Yiman Fang,M Jane Bunting,Chunmei Ma

doi:10.1177/0309133319861808

Abstract

Reconstructing land cover from pollen data using mathematical models of the relationship between them has the potential to translate the many thousand pollen records produced over the last 100 years (over 2300 radiocarbon-dated pollen records exist for the UK alone) into formats relevant to ecologists, archaeologists and climate scientists. However, the reliability of these reconstructions depends on model parameters. A key parameter is Relative Pollen Productivity (RPP), usually estimated from empirical data using ‘Extended R Value analysis’ (ERV analysis). Lack of RPP estimates for many regions is currently a major limitation on reconstructing global land cover. We present two alternatives to ERV analysis, the Modified Davis Method and an iteration method, which use the same underlying model of the relationship between pollen and vegetation to estimate RPP from empirical data, but with different assumptions. We test them in simulation against ERV analysis, and use a case study of a problematic empirical dataset to determine whether they have the potential to increase the speed and geographic range of RPP estimation. The two alternative methods are shown to perform at least as well as ERV analysis in simulation. We also present new RPP estimates from southeastern sub-tropical China for nine taxa estimated using the Modified Davis Method. Adding these two methods to the ‘toolkit’ for land cover reconstruction from pollen records opens up the possibility to estimate a key parameter from existing datasets with less field time than using current methods. This can both speed up the inclusion of more of the globe in past land cover mapping exercises such as the PAGES Landcover6k working group and improve our understanding of how this parameter varies within a single taxon and the factors control that variation.

Highlights

Land cover is a fundamental earth system variable, with well understood impacts on regional and global climate and hydrological cycles
The errors reported by the Extended R-value method (ERV) analysis software are one standard deviation (SD) estimated by propagation, while the errors presented for the two new methods are the standard deviations of the full set of estimated Relative Pollen Productivity (RPP) values, it is possible to make finer discrimination of RPP rank
The results of the three methods we applied differ from each other substantially, but given the small dataset and individual site effects identified through using the Modified Davis Method (MDM) results, we argue that the MDM results with major outliers removed are the most appropriate for use in future land cover reconstruction

Summary

Introduction

Land cover is a fundamental earth system variable, with well understood impacts on regional and global climate and hydrological cycles. Reconstructing past land cover is an important step to improve testing of regional, continental and global scale climate models against known past climate conditions, and recent developments coordinated by the PAGES Working Group LandCover6k (http://www.pastglobalchanges.org/ini/wg/landcover6k/intro) using models of pollen dispersal and deposition combined with data from pollen records preserved in lake and mire sedimentary systems offer a substantial step forward in producing global land cover maps (Hellman et al, 2008a, b; Soepboer et al, 2010; Trondman et al, 2015; Bunting et al, 2018) Extending this approach from core researched areas (e.g. Northwest Europe, Mazier et al, 2012; temperate China, Li, 2016) to the rest of the globe is limited by the need to parameterise models for the main plant taxa present in each region, which is currently achieved using the “Extended R value” approach (hereafter ERV approach). These known variations are treated as “noise” in terms of the pollen productivity measure needed to interpret sedimentary pollen records, where individual pollen assemblages come from samples which amalgamate multiple years, and reconstructions typically cover “time windows” of 100-500 years and are based on multiple pollen samples within each window

Methods

Results

Discussion

Conclusion