Our understanding of climate and vegetation changes throughout the Holocene is hampered by representativeness in sedimentary archives. Potential biases such as production and preservation of the markers are identified by comparing these proxies with modern environments. It is important to conduct multi-proxy studies and robust calibrations on each terrestrial biome. These calibrations use large databases dominated by forest samples. Therefore, including data from steppe and desert–steppe sites becomes necessary to better calibrate arid environments. The Mongolian Plateau, ranging from the Baikal area to the Gobi desert, is especially characterized by low annual precipitation and continental annual air temperature. The characterization of the climate system of this area is crucial for the understanding of Holocene monsoon oscillations. This study focuses on the calibration of proxy–climate relationships for pollen and glycerol dialkyl glycerol tetraethers (GDGTs) by comparing large Eurasian calibrations with a set of 49 new surface samples (moss polster, soil and mud from temporary dry ponds). These calibrations are then cross-validated by an independent dataset of top-core samples and applied to four Late Holocene paleosequences (two brGDGT and two pollen records) surrounding the Mongolian Plateau: in the Altai mountains, the Baikal area and the Qaidam basin, to test the accuracy of local and global calibrations. We show that (1) preserved pollen assemblages are clearly imprinted on the extremities of the ecosystem range but mitigated and unclear on the ecotones; (2) for both proxies, inferred relationships depend on the geographical range covered by the calibration database as well as on the nature of samples; (3) even if local calibrations suffer from reduced amplitude of climatic parameters due to local homogeneity, they better reflect actual climate than the global ones by reducing the limits for saturation impact; (4) a bias in climatic reconstructions is induced by the over-parameterization of the models by the addition of artificial correlation; and (5) paleoclimate values reconstructed here are consistent with Mongolia–China Late Holocene climate trends and validate the application of local calibrations for both pollen and GDGTs (closest fit to actual values and realistic paleoclimate amplitude). We encourage the application of this surface calibration method to reconstruct paleoclimate and especially consolidate our understanding of the Holocene climate and environment variations in arid central Asia.
Since the understanding of the interactions between the paleoclimate proxies, such as pollen or biomarker abundances, and general circulation model outputs became a major issue in future climate change modeling, resolving the issue of climate proxy calibration is crucial
Eurasian map of all the pollen surface samples included in the database. The color code refers to the biome pollen inferred for each site. The biomes are WAMX, warm mixed forest; WAST, warm steppe; TEDE, temperate deciduous forest; XERO, xerophytic shrubland; COMX, cool mixed forest; HODE, hot desert; CLMX, cold mixed forest; PION, pioneer forest; TAIG, taiga forest; COST, cold steppe; COCO, cold conifer forest; TUND, tundra; ANTH, anthropic environment; CLDE, cold deciduous forest; CODE, cold desert. The thickest points underline the COST samples selected for this study to operate the transfer function method among all COST sites (shown by a lozenge on map). The arrows indicate the main climatic system driving the Mongolian climate: in orange the Westerlies arriving from the North Atlantic ocean and in blue the Indian Summer Monsoon (ISM) and the East Asian Summer Monsoon (EASM). The dashed line represents the EASM limit following
Lake sediment archives are commonly used to infer past variations of these
climate and environmental systems associated with vegetation and human land
use
Among new promising proxies and from the three last decades, biomarkers such
as the glycerol dialkyl glycerol tetraethers (GDGTs) have provided new
perspectives on continental temperature reconstructions
Since multi-proxy studies become more and more accurate in both temperature
and precipitation reconstruction, local to regional calibrations have been
proposed for dry areas such as the arid central Asian (ACA) area: pollen
semi-quantitative climate reconstruction collection of a new set of modern surface samples for Mongolia with homogeneous characterization of their bioclimate environment followed by pollen and GDGT pattern characterization; evaluation of the match between actual bioclimate environments and the associated pollen rain and biomarker assemblages based on mathematical criterion without eco-physiological considerations; creation of local Mongolian Plateau (MP) climate calibrations for pollen and GDGTs and comparison of local and global calibrations in the Mongolian case study; a posteriori validation of the inferred relationships between proxies and ecological likelihood based on the currently developed evidences of brGDGT and pollen rain ecological significance; discussion of the implications of the calibration mismatches in terms of climatic reconstructions in arid and cold environments; testing of the new calibrations (pollen and brGDGTs) through their application on four surrounding Late Holocene records: two pollen records, Dulikha bog
The study area lies from 52
To test the reliability of our modern calibrations, we have finally selected
four paleosequences within or close to the MP used as test benches of the
calibrations. For the pollen analysis, the cores of D3L6 from
The central part of the Mongolian Plateau (MP) is characterized by a dry and
cold flat mosaic of steppes and deserts with a 1220
The distribution of vegetation and biomes follows a latitudinal belt
organization: in the north the boreal forest presents a mosaic of light taiga
dominated by
In the central steppe–forest biome, the vegetation is marked by an ecotone
with short grassland controlled by grazing in the valley and larches on the
slopes. The forest is gathered in patches constituting between 10
Different chemical processes were performed on the samples: bryophytic parts of
the moss samples were deflocculated by potassium hydroxide (KOH) and filtered by 250 and
10
Among all of the pollen-inferred climate methods, the MAT and the WAPLS were
applied in this study on four different modern pollen datasets, and on the
D3L6 and Dulikha fossil pollen sequence to test the accuracy of these
calibrations
Simplified surface pollen diagram, bioclimatically sorted, of the Siberian–Mongolian transect. The pollen taxa are expressed in % TP. The ecosystem units were determined with a CONISS analysis. The left-hand colored dots represent the ecosystem for each sample from light-taiga–riparian forest (deep blue), light/dark taiga–birch sub-taiga, steppe–forest, alpine meadow, steppe, steppe–desert and desert (deep red). The color scale is presented in Fig.
Because Mongolia and Siberia have relatively few weather stations
(Fig.
For consistency with the sampling process and the modeling methodologies
developed for pollen analysis, soil parts of the moss polsters, soil samples
and pond mud were treated for GDGT analysis. After freeze drying, about 0.6 g of material was sub-sampled. The total lipid extract (TLE) was
microwave extracted (MARS 6 CEM) with dichloromethane (DCM)
Each compound was identified and manually integrated according to its
Fractional abundances of
To infer temperatures from brGDGT abundances, two types of model were applied:
linear relationships between temperature and MBT–CBT indexes, and multiple
regression (mr) models between one climate parameter and a proportion of
multiple brGDGT fractional abundances. For the simple linear regression model,
a correlation matrix between climate parameters and indexes was calculated
using the corrplot Rcran library. For mr models, we developed in the
R environment a stepwise selection model
GDGTs and pollen data were analyzed with a principal component analysis (PCA)
to determine the axes explaining the variance within the samples. The biotic
values (pollen and GDGTs) were also compared to abiotic parameters (climate,
elevation, location and soil features) by the way of a redundancy analysis
(RDA). The regression models were run with the
The pollen rain (Fig.
Multivariate statistics for the proxies clustered by ecosystems:
The pollen rain trends follow similar variations to bioclimate parameters
in MAP, MAAT and elevation (Fig.
To reconstruct climate parameters from pollen data, MAT and WAPLS methods were
applied on the four scales, modern pollen datasets and the 10 climate
parameters (Table
Statistical results of the MAT and WAPLS methods applied to four surface pollen datasets and 10 climate parameters
In the MMNT5C12 sediments, isoGDGTs are dominated by GDGT-0 and crenarchaeol
(74.6
The average
The sediment samples from the lake MMNT5C12, used for past sequence comparison,
are more homogeneous than the surface samples, especially when compared with
the moss polsters that present a wide variability
(Fig.
The brGDGT/climate RDA shows that the brGDGT variance is dominated by the MAP
as the first component (Fig.
The stepwise selection model (SSM) for climate–brGDGT modeling was applied
only on the 5- and 6-methyl, because 7-methyl brGDGTs show weak significance
in the variance explanation (PCA; Fig.
Statistical values and equations of the best brGDGT
Validation of brGDGT-climate models on the study sites: reconstructed values for literature MAAT
According to
The topographic fence in Mongolia also affects the pollen and brGDGT
distributions by itself, as seen in both RDA analyses (Fig.
Both GDGT and pollen calibrations show that the precipitation calibrations are
more reliable than temperature ones (Tables
Statistical values plotted against the number of parameters of the different mr–GDGT models: the
To reduce the signal
The soils of the Gobi desert also have a high salinity level which is also a
parameter of control on brGDGT fractional abundances
Finally, the saturation effect of the proxies when they reach the limits of
their range of appliance is also to be taken into consideration. Since both
pollen and brGDGT signals are analyzed in fractional abundance (i.e., percentage of the
total count of concentration), these proxies evolve in a
Relationships between the eight major pollen taxa (% TP)
and MAP (
Among the possible methods, statistical values help to select the most reliable paleoclimate reconstruction. However, the correlation
(
The cross-values of the nine best
For both
Whatever proxy is used, when reconstructing temperatures and precipitation
from past records in a given location, there is the issue of basing
reconstructions on calibrations based on local or global datasets
Similarly, for pollen transfer functions, the geographic range of the surface
samples on which the calibration relies is a relevant parameter to take into
account for the reliability of the paleoclimate reconstructions. The choice of
the maximum value of this geographic range has been discussed previously for
vegetation modeling, for example, the Relevant Source Area of Pollen
Concerning transfer functions, WAPLS performs better for the local database
than for the COST and EAP databases
(Table
To test the reliability of our local calibrations, the pollen transfer
function and the brGDGT mr models have been applied on four
paleosequences. Because there is still no available core analyzed for both
pollen and brGDGTs either in ACA area or in the MP, the Dulikha bog
ACA climate reconstruction for the 5000
For the pollen transfer function (Fig.
In Fig.
Even if the pollen-based and brGDGT-based climate reconstructions were not
conducted on the same core, the D3L6 and NRX records (Fig.
Figure
The paleoenvironmental and paleoclimatic signals may present several
uncertainties (differential production, preservation, etc.) which can
misguide the interpretation of past variations. This study shows how both a
multi-proxy approach and an accurate calibration are important in preventing
these biases. We propose a new calibration for mean annual precipitation (MAP)
and mean annual air temperature (MAAT) from brGDGTs as well as a new pollen
surface database available for transfer functions. The correlations between
pollen rain and climate on the one hand and brGDGT soil production and climate on
the other are visible but are still mitigated by the complex climate system of
arid central Asia and the diversity of soils and ecosystems. Precisely, each
of our proxies seems to be more narrowly linked to precipitation (MAP) than
temperature (MAAT) counter to the majority of calibrations in the
literature. This is validated on both modern and past sequences for pollen and
brGDGTs. The nature of the samples considered (soil, moss polster and mud from
temporary dry ponds) also greatly affected these correlations. The calibration
attempt for the extreme bioclimates of the MP is difficult because of the low
range of climate values, despite the climate diversity ranging from cold and
slightly wet (north) to the arid and warm (south) conditions. Even if global
and regional calibrations could be applied in such a setting, local
calibrations provide enhanced accuracy and specificity. The MAAT and MAP
values do not remarkably spread in the vectorial space, which makes it harder to
distinguish the linear correlation against variance noise. Moreover, this
range of values is close to the lower saturation limit of the proxies, which
makes the accurate local calibration tricky but necessary. The local
calibrations also suffer from the reduced size and small geographic extent of
the dataset. The vegetation cover, extending from a high-cover taiga forest to
bare-soil desert cover, also buffers the climate signal and the GDGT/pollen
response. The correlations between climate parameters and GDGT/pollen
proportion are therefore lower than they could be at global
scale. Nonetheless, and despite the lower correlation of the local
calibration, these local approaches appear to be more accurate to fit the
actual climate parameters than the global ones: both for pollen transfer
functions and brGDGT multiple regression models. It is especially the case
during the short Late Holocene period which is not suffering from abrupt
ecosystem changes. These positive model results have to be considered in
light of over-parameterization limits. Too many parameters in mr–brGDGT
models or in pollen MAT or WAPLS transfer function can add artificially to the
linear relation between climate and proxies and lead to misinterpretation of
paleoclimate records. Akaike's information criterion combined with RMSE and
The data produced and analyzed in this study will be available at Pangaea.
The supplement related to this article is available online at:
LD conducted the analytical work. LD, SJ, OP, GM designed the study. All the authors contributed to the scientific reflection as well as to the preparation of the paper.
The authors declare that they have no conflict of interest.
We want to thank the editor and the three referees for their relevant recommendations. We also thank all the direct and indirect contributors to the global surface pollen dataset as well as the Laboratory of Ecological and Evolutionary Synthesis of the National University of Mongolia for its support during the field trip. We also express gratitude to Laure Paradis for her GIS advice, Marc Dugerdil for the help with Python fixing, Jérôme Magail and the Monaco–Mongolia joint mission for their technical and financial support in providing top cores and sediment samples from Arkhangai, Salomé Ansanay-Alex for her spectrometer expertise, and La Tendresse and Le BIB for their support. This is ISEM publication no. ISEM 2021-075.
This research has been supported by the French Centre National de la Recherche Scientifique (CNRS) and the ISEM team DECG. Lucas Dugerdil's salary has been supported by the Ecole Normale Supérieure de Lyon. For the analytical work completed at LGLTPE-ENS de Lyon and the paper registration fees, this research was funded by Institut Universitaire de France funds to Guillemette Ménot.
This paper was edited by Nathalie Combourieu Nebout and reviewed by Natalia Rudaya and two anonymous referees.