Enabling possibilities to quantify past climate from fossil assemblages 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)
- et al.
Vegetation of Eurasia from the last glacial maximum to present: key biogeographic patterns
Quat. Sci. Rev.
(2017) - et al.
Refining mutual climatic range (MCR) quantitative estimates of palaeotemperature using ubiquity analysis
Quat. Sci. Rev.
(2006) - et al.
Pollen data as climate proxies
- et al.
A modern pollen–climate dataset from China and Mongolia: assessing its potential for climate reconstruction
Rev. Palaeobot. Palynol.
(2014) - et al.
Impacts of the spatial extent of pollen-climate calibration-set on the absolute values, range and trends of reconstructed Holocene precipitation
Quat. Sci. Rev.
(2017) - et al.
Critical evaluation of climate syntheses to benchmark CMIP6/PMIP4 127 ka last interglacial simulations in the high-latitude regions
Quat. Sci. Rev.
(2017) - et al.
Evolving southwest African response to abrupt deglacial North Atlantic climate change events
Quat. Sci. Rev.
(2015) - et al.
Influence of tropical easterlies in the southwestern Cape of Africa during the Holocene
Quat. Sci. Rev.
(2015) - et al.
The dynamic relationship between temperate and tropical circulation systems across South Africa since the last glacial maximum
Quat. Sci. Rev.
(2017) - et al.
Southeast African records reveal a coherent shift from high- to low-latitude forcing mechanisms along the east African margin across last glacial–interglacial transition
Quat. Sci. Rev.
(2015)
Late Pleistocene-Holocene vegetation and climate change in the Middle Kalahari, Lake Ngami, Botswana
Quat. Sci. Rev.
A late Pleistocene long pollen record from Lake Urmia, NW Iran
Quat. Res.
Modern climate-vegetation-pollen relations in Africa and adjacent areas
Quat. Sci. Rev.
Chapter thirteen transfer functions: methods for quantitative paleoceanography based on microfossils
Dev. Mar. Geol.
Is spatial autocorrelation introducing biases in the apparent accuracy of paleoclimatic reconstructions?
Quat. Sci. Rev.
Predictive habitat distribution models in ecology
Ecol. Appl.
Holocene temperature changes in northern Fennoscandia reconstructed from chironomids using Bayesian modelling
Quat. Sci. Rev.
Probability density functions as botanical-climatological transfer functions for climate reconstruction
Quat. Res.
Dynamics of the last four glacial terminations recorded in Lake Van, Turkey
Quat. Sci. Rev.
Dating Late Glacial abrupt climate changes in the 14,570 yr long continuous varve record of Lake Van, Turkey
Palaeogeogr. Palaeoclimatol. Palaeoecol.
Vegetation and climate changes in the South Eastern Mediterranean during the Last Glacial-Interglacial cycle (86 ka): new marine pollen record
Quat. Sci. Rev.
50,000 years of climate in the Namib Desert, Pella, South Africa
Palaeogeogr. Palaeoclimatol. Palaeoecol.
A 600,000 year long continental pollen record from Lake Van, eastern Anatolia (Turkey)
Quat. Sci. Rev.
The climate of Europe during the Holocene: a gridded pollen-based reconstruction and its multi-proxy evaluation
Quat. Sci. Rev.
Quantitative interpretation of fossil pollen spectra: dissimilarity coefficients and the method of modern analogs
Quat. Res.
A flexible approach to assessing synchroneity of past events using Bayesian reconstructions of sedimentation history
Quat. Sci. Rev.
Vegetation and environmental changes during the last interglacial in eastern Anatolia (Turkey): a new high-resolution pollen record from Lake Van
Palaeogeogr. Palaeoclimatol. Palaeoecol.
Chronostratigraphy of the 600,000 year old continental record of Lake Van (Turkey)
Quat. Sci. Rev.
The secret assumption of transfer functions: problems with spatial autocorrelation in evaluating model performance
Quat. Sci. Rev.
Evaluation of transfer functions in spatially structured environments
Quat. Sci. Rev.
Piecing together the past: statistical insights into paleoclimatic reconstructions
Quat. Sci. Rev.
Interglacial vegetation succession: a view from southern Europe
Quat. Sci. Rev.
Modern pollen data from North America and Greenland for multi-scale paleoenvironmental applications
Quat. Sci. Rev.
Etopo1 1 Arc-Minute Global Relief Model: Procedures, Data Sources and Analysis. NOAA Tech. Memo. NESDIS NGDC-24
Klimatet i Sverige efter istiden
Nord. Tidskr. H
The Climate of Sweden in the Late-Quaternary Period, Facts and Theories. Sveriges Geol. Undersöknings Årsbok, Ser. C 3
Die Veränderungen des Klimas seit dem Maximum der letzten Eiszeit
Pollen-based continental climate reconstructions at 6 and 21 ka: a global synthesis
Clim. Dyn.
Late-Quaternary palaeoclimatic research in Fennoscandia – a historical review
Boreas
Diatoms and pH reconstruction
Philos. Trans. R. Soc. Lond. Ser. B Biol. Sci.
Strengths and weaknesses of quantitative climate reconstructions based on Late-Quaternary biological proxies
Open Ecol. J.
Flexible paleoclimate age-depth models using an autoregressive gamma process
Bayesian Anal.
Cited by (16)
Middle Miocene evolution of East Asian summer monsoon precipitation in the northeast part of the Tibetan Plateau based on a quantitative analysis of palynological records
2023, Palaeogeography, Palaeoclimatology, PalaeoecologyMultiple forcing on Late Miocene East Asian Summer Monsoon Precipitation Variability in NE Tibetan Plateau
2023, CatenaCitation 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, PalaeoecologyCitation 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 ReviewsCitation 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).
Extreme hydroclimate response gradients within the western Cape Floristic region of South Africa since the Last Glacial Maximum
2019, Quaternary Science ReviewsA 25,000 year record of climate and vegetation change from the southwestern Cape coast, South Africa
2022, Quaternary Research (United States)