Elsevier

Global and Planetary Change

Volume 175, April 2019, Pages 27-35
Global and Planetary Change

Enabling possibilities to quantify past climate from fossil assemblages at a global scale

https://doi.org/10.1016/j.gloplacha.2019.01.016Get rights and content

Highlights

  • A new, globally applicable probabilistic tool to quantitively reconstruct past climates.

  • The calibration dataset contains the distributions of multiple palaeoclimatic bio-indicators.

  • Understudied regions can now be explored using the CREST/GBIF framework.

  • A dedicated software exists to facilitate the use of the CREST/GBIF framework.

Abstract

The field of quantitative palaeoclimatology has made significant progress in the past decades. However, this progress has been spatially heterogeneous and strong discrepancies – both in terms of quality and density – exist between Europe and North America and the rest of the world. The need to balance this distribution of quantified records has never been stronger, and improving our understanding of past global climate is urgent in order to better evaluate and employ the predictions of climate models. In this paper, it is argued that this gap can be reduced by applying the climate reconstruction method CREST (Climate REconstruction SofTware) calibrated using the open-access GBIF (Global Biodiversity Information Facility) database, which contains hundreds of thousands modern distributions of numerous bio-indicator proxies (e.g. pollen, chironomids, foraminifers, etc.). Using the taxonomical diversity of the GBIF database, CREST can be used to reconstruct various climate and/or environmental parameters from assemblage variety of different records. Independent from the usual reconstruction techniques using surface samples for calibration, the CREST/GBIF framework can also be used as an independent tool that can be easily and efficiently applied (1) to complete the global coverage of climate reconstructions, as exemplified with a precipitation reconstruction from Lake Van, Turkey, (2) to revisit existing data sets to obtain new quantitative reconstructions, and (3) to evaluate and/or refine reconstructions based on other methods. The application of this tool promises to foster advances in our understanding of the past climate variability of the Earth system at a global scale.

Introduction

Since the seminal papers of Andersson, 1903, Andersson, 1909, Andersson, 1910 and Iversen (1944), a variety of statistical techniques has been developed to produce quantified estimates of past climatic conditions from fossil botanical data. Valuable in its own right as a way to explore how past climates have shaped modern environments (e.g. the historical perspective presented in Birks and Seppä, 2010), the quantification of climatic variables is now of fundamental importance for evaluating and refining climate models (e.g.Harrison et al., 2015). As such, accurate palaeoclimatic reconstructions are an increasingly important element in global efforts to understand climate change and mitigate its impacts.

Of the methods that have been developed over time (for a full review of their relative strengths and weaknesses, the reader is refered for instance to the syntheses of Birks et al. (2010), Brewer et al. (2013), Guiot and de Vernal (2007), and Juggins and Birks (2012)), two remain most commonly used: the modern analogue technique (Guiot and Pons, 1986; Overpeck et al., 1985) and the regression techniques WA (Weighted Averaging) and WA-PLS (Weighted Averaging-Partial Least Square, Birks et al., 1990; ter Braak and Juggins, 1993; ter Braak and Looman, 1986; ter Braak and van Dame, 1989). The statistical backgrounds of these two techniques are very different but nevertheless rely on the same type of datasets for calibration: extensive collections of modern assemblage data covering diverse climatic and environmental gradients. These modern pollen datasets are, unfortunately, only commonly available from North America (Whitmore et al., 2005), parts of Eurasia (Binney et al., 2017; Davis et al., 2013; Marinova et al., 2017) or Africa (Gajewski et al., 2002) and, more recently, China (Cao et al., 2014; Cao et al., 2017; Zheng et al., 2014). While additional regional databases may eventually be developed, it will require many years of intensive and collective sampling efforts. It should also be considered that time is not the only limiting factor. Some environments, such as drylands, are not favourable for the preservation of pollen data in surface sediments or traps. In such environments reliable surface samples will never become available, and quantification techniques that rely on modern assemblage samples will remain inapplicable.

In order to enable quantitative climate reconstructions in these ‘quantification deserts’, alternative methods that are independent from surface samples have been developed. Foremost among these are techniques based on indicator species. These use modern distributions of bio-indicators (e.g. plant species) for their calibration instead of surface samples. This family of techniques includes very basic approaches, such as the mono-specific indicator species approach of Andersson (1909) – who estimated Holocene summer temperature in Sweden based on the sole observation of macroremains of Corylus (hazel) – and more complex Bayesian techniques, such as that formalized by Kühl et al. (2002), which uses conditional probability density functions (pdfs) to represent the climate dependencies of different taxa. The recently developed CREST (Climate REconstruction SofTware) method (Chevalier et al., 2014) is derived from the work of Kühl et al. (2002), but CREST has reduced the number of assumptions to expand the applicability of the approach. Originally developed to reconstruct the palaeoclimates of the southern African drylands – a region formerly considered as a ‘quantification desert’ – from fossil pollen data, CREST has proven to be a reliable technique for reconstructing both modern (Chevalier et al., 2014) and past climates (Chase et al., 2015a; Chase et al., 2015b; Chevalier and Chase, 2015; Chevalier and Chase, 2016; Cordova et al., 2017; Lim et al., 2016) from fossil pollen data.

Until recently, the need to have access to extensive databases of modern distributions of plants has limited the application of pdf techniques such as CREST. In this paper, it is proposed to use a curated version of the GBIF (Global Biodiversity Information Facility) database to overcome this limitation. Open-access, the GBIF database contains >920,000,000 georeferenced presence records (last access: June 2018) of a variety of living organisms commonly used a palaeo-indicators (animals, plants, bacteria, etc.) from both the marine and terrestrial realms. The combination of CREST and GBIF will enable the reconstruction of various climate variables from fossil pollen records across the globe (Fig. 1), and will also open possibilities to adapt pdf-based techniques with a large variety of non-pollen palaeo-proxies with stable, semi-permanent states from both the terrestrial (beetles and chironomids) and oceanic realms (foraminifers and diatoms).

Section snippets

The CREST method

The CREST method (Chevalier et al., 2014) is related to a Bayesian approach that combines presence-only occurrence data and modern climatologies to estimate the conditional response of a given taxon to a variable of interest. Taking the form of probability density functions (pdfs), these links are fitted in one or two steps based on the nature of the proxy being studied. In simple cases, where fossils can be identified at species level (e.g. plant macrofossils), the pdfs are defined by unimodal

Regional setting

Lake Van is the fourth largest terminal lake in the world (38.6°N, 42.8°E, volume 607 km3, area 3570 km2, maximum water depth 460 m), extending for 130 km on the eastern Anatolian high plateau, Turkey (Litt et al., 2014; Pickarski et al., 2015b). With an elevation of 1646 m.a.s.l., Lake Van is located in a zone of complex tectonic movements, associated with the collision of Afro/Arabian plate from the south and the Eurasian plate from the north, and is surrounded by high mountain ranges. The

Discussion/perspective

The case study from Lake Van highlights the potential of using CREST with GBIF. Lake Van being located near the southeastern edge of the European Modern Pollen Database (Davis et al., 2013), the performance of MAT and WAPLS would be strongly limited and the full range of past variability could be missed. Hence, eastern Turkey is part of a “quantification desert”. With GBIF and CREST, more potential climates have been included in the calibration dataset, which now encompasses Central and Eastern

Conclusion

By greatly expanding the regions where quantified palaeoclimatic reconstructions can be obtained, the combination of the CREST method with the global GBIF database enables quantified climate reconstructions from some of the most remote and under-studied locations on earth. If applied broadly, this methodology could fill the gaps in the global data coverage and, open new possibilities for multi-scalar data-model comparisons. Obtaining reconstructions from the ‘quantification deserts’ will also

Competing interests

The author declares that he has no conflict of interest.

Acknowledgements

I particularly want to thank Brian Chase for all his help, support and repeated encouragements during the elaboration of this manuscript.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability

The CREST-formatted GBIF database can be downloaded from https://figshare.com/s/cf9fd5074af921d17c2b (Chevalier, 2018). Lake Van precipitation reconstructions are available from https://chevaliermanuel.wixsite.com/webpage/softwares-datasets and on PANGAEA.

References (85)

  • C.E. Cordova et al.

    Late Pleistocene-Holocene vegetation and climate change in the Middle Kalahari, Lake Ngami, Botswana

    Quat. Sci. Rev.

    (2017)
  • M. Djamali et al.

    A late Pleistocene long pollen record from Lake Urmia, NW Iran

    Quat. Res.

    (2008)
  • K. Gajewski et al.

    Modern climate-vegetation-pollen relations in Africa and adjacent areas

    Quat. Sci. Rev.

    (2002)
  • J. Guiot et al.

    Chapter thirteen transfer functions: methods for quantitative paleoceanography based on microfossils

    Dev. Mar. Geol.

    (2007)
  • J. Guiot et al.

    Is spatial autocorrelation introducing biases in the apparent accuracy of paleoclimatic reconstructions?

    Quat. Sci. Rev.

    (2011)
  • A. Guisan et al.

    Predictive habitat distribution models in ecology

    Ecol. Appl.

    (2000)
  • A. Korhola et al.

    Holocene temperature changes in northern Fennoscandia reconstructed from chironomids using Bayesian modelling

    Quat. Sci. Rev.

    (2002)
  • N. Kühl et al.

    Probability density functions as botanical-climatological transfer functions for climate reconstruction

    Quat. Res.

    (2002)
  • O. Kwiecien et al.

    Dynamics of the last four glacial terminations recorded in Lake Van, Turkey

    Quat. Sci. Rev.

    (2014)
  • G. Landmann et al.

    Dating Late Glacial abrupt climate changes in the 14,570 yr long continuous varve record of Lake Van, Turkey

    Palaeogeogr. Palaeoclimatol. Palaeoecol.

    (1996)
  • D. Langgut et al.

    Vegetation and climate changes in the South Eastern Mediterranean during the Last Glacial-Interglacial cycle (86 ka): new marine pollen record

    Quat. Sci. Rev.

    (2011)
  • S. Lim et al.

    50,000 years of climate in the Namib Desert, Pella, South Africa

    Palaeogeogr. Palaeoclimatol. Palaeoecol.

    (2016)
  • T. Litt et al.

    A 600,000 year long continental pollen record from Lake Van, eastern Anatolia (Turkey)

    Quat. Sci. Rev.

    (2014)
  • A. Mauri et al.

    The climate of Europe during the Holocene: a gridded pollen-based reconstruction and its multi-proxy evaluation

    Quat. Sci. Rev.

    (2015)
  • J.T. Overpeck et al.

    Quantitative interpretation of fossil pollen spectra: dissimilarity coefficients and the method of modern analogs

    Quat. Res.

    (1985)
  • A.C. Parnell et al.

    A flexible approach to assessing synchroneity of past events using Bayesian reconstructions of sedimentation history

    Quat. Sci. Rev.

    (2008)
  • N. Pickarski et al.

    Vegetation and environmental changes during the last interglacial in eastern Anatolia (Turkey): a new high-resolution pollen record from Lake Van

    Palaeogeogr. Palaeoclimatol. Palaeoecol.

    (2015)
  • M. Stockhecke et al.

    Chronostratigraphy of the 600,000 year old continental record of Lake Van (Turkey)

    Quat. Sci. Rev.

    (2014)
  • R.J. Telford et al.

    The secret assumption of transfer functions: problems with spatial autocorrelation in evaluating model performance

    Quat. Sci. Rev.

    (2005)
  • R.J. Telford et al.

    Evaluation of transfer functions in spatially structured environments

    Quat. Sci. Rev.

    (2009)
  • M.P. Tingley et al.

    Piecing together the past: statistical insights into paleoclimatic reconstructions

    Quat. Sci. Rev.

    (2012)
  • P.C. Tzedakis et al.

    Interglacial vegetation succession: a view from southern Europe

    Quat. Sci. Rev.

    (1995)
  • J. Whitmore et al.

    Modern pollen data from North America and Greenland for multi-scale paleoenvironmental applications

    Quat. Sci. Rev.

    (2005)
  • C. Amante et al.

    Etopo1 1 Arc-Minute Global Relief Model: Procedures, Data Sources and Analysis. NOAA Tech. Memo. NESDIS NGDC-24

    (2009)
  • G. Andersson

    Klimatet i Sverige efter istiden

    Nord. Tidskr. H

    (1903)
  • G. Andersson

    The Climate of Sweden in the Late-Quaternary Period, Facts and Theories. Sveriges Geol. Undersöknings Årsbok, Ser. C 3

    (1909)
  • G. Andersson

    Die Veränderungen des Klimas seit dem Maximum der letzten Eiszeit

  • P.J. Bartlein et al.

    Pollen-based continental climate reconstructions at 6 and 21 ka: a global synthesis

    Clim. Dyn.

    (2011)
  • H.J.B. Birks et al.

    Late-Quaternary palaeoclimatic research in Fennoscandia – a historical review

    Boreas

    (2010)
  • H.J.B. Birks et al.

    Diatoms and pH reconstruction

    Philos. Trans. R. Soc. Lond. Ser. B Biol. Sci.

    (1990)
  • H.J.B. Birks et al.

    Strengths and weaknesses of quantitative climate reconstructions based on Late-Quaternary biological proxies

    Open Ecol. J.

    (2010)
  • M. Blaauw et al.

    Flexible paleoclimate age-depth models using an autoregressive gamma process

    Bayesian Anal.

    (2011)
  • Cited by (16)

    • Multiple forcing on Late Miocene East Asian Summer Monsoon Precipitation Variability in NE Tibetan Plateau

      2023, Catena
      Citation Excerpt :

      To investigate the EASM precipitation variations and its driving forces over the orbital timescale, the previously published late Miocene pollen record (∼10.83–6.3 Ma, 55 samples) from the YW section (Fig. 1B; Hui et al., 2021) was also integrated into this study. The CREST method is a statistical approach that combines presence-only plant occurrence data with modern climate data to estimate the probability density functions (PDFs) of plant taxa to specific climate variables (Chevalier, 2019; Chevalier et al., 2014). The PDFs represent the statistical relationship between a taxon (e.g. Betula) and a specific climate variable (e.g. PWetQ).

    • Miocene East Asia summer monsoon precipitation variability and its possible driving forces

      2021, Palaeogeography, Palaeoclimatology, Palaeoecology
      Citation Excerpt :

      We refer the readers to Hui et al. (2011) for more detailed information of the vegetation changes. To obtain the quantitative records of EASM from the YW section pollen data, the CREST method and software described in Chevalier et al. (2014) and the CREST-formatted GBIF database developed by Chevalier (2019) were used to reconstruct PWetQ. The CREST method is a Bayesian approach that uses modern plant distributions and corresponding climate data to define probability density functions (PDFs) describing the relationship between a unique plant taxon and specific climate variables, such as PWetQ (Chevalier et al., 2014, 2020).

    • Pollen-based climate reconstruction techniques for late Quaternary studies

      2020, Earth-Science Reviews
      Citation Excerpt :

      Supporting further field campaigns, especially in quantification deserts, and making these data publicly accessible (e.g. by adding them to relevant regional Constituent Databases in Neotoma, or any other repository, such as Pangaea www.pangaea.de or NOAA www.ncdc.noaa.gov) is critical to improve the global coverage of modern pollen assemblages and, by extension, the applicability of climate reconstruction methods. To our knowledge, only the Global Biodiversity Information Facility (GBIF) provides a set of globally dense distributions of plants that can be directly used with the pdf techniques (Chevalier, 2019; Chevalier, 2018). Other high-resolution datasets of the distribution of most European trees (Mauri et al., 2017) and North American (Biodiversity Information Serving Our Nation (BISON) and US Forest Inventory and Analysis (FIA)) vegetation are also available (see Table 2).

    View all citing articles on Scopus
    View full text