skip to main content

Title: Machine-learning based reconstructions of primary and secondary climate variables from North American and European fossil pollen data

We test several quantitative algorithms as palaeoclimate reconstruction tools for North American and European fossil pollen data, using both classical methods and newer machine-learning approaches based on regression tree ensembles and artificial neural networks. We focus on the reconstruction of secondary climate variables (here, January temperature and annual water balance), as their comparatively small ecological influence compared to the primary variable (July temperature) presents special challenges to palaeo-reconstructions. We test the pollen–climate models using a novel and comprehensive cross-validation approach, running a series ofh-block cross-validations usinghvalues of 100–1500 km. Our study illustrates major benefits of this variableh-block cross-validation scheme, as the effect of spatial autocorrelation is minimized, while the cross-validations with increasinghvalues can reveal instabilities in the calibration model and approximate challenges faced in palaeo-reconstructions with poor modern analogues. We achieve well-performing calibration models for both primary and secondary climate variables, with boosted regression trees providing the overall most robust performance, while the palaeoclimate reconstructions from fossil datasets show major independent features for the primary and secondary variables. Our results suggest that with careful variable selection and consideration of ecological processes, robust reconstruction of both primary and secondary climate variables is possible.

more » « less
Author(s) / Creator(s):
; ; ;
Publisher / Repository:
Nature Publishing Group
Date Published:
Journal Name:
Scientific Reports
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract Aim

    We analysed a dataset composed of multiple palaeoclimate and lake‐sediment pollen records from New England to explore how postglacial changes in the composition and spatial patterns of vegetation were controlled by regional‐scale climate change, a subregional environmental gradient, and landscape‐scale variations in soil characteristics.


    The 120,000‐km2study area includes parts of Vermont and New Hampshire in the north, where sites are 150–200 km from the Atlantic Ocean, and spans the coastline from southeastern New York to Cape Cod and the adjacent islands, including Block Island, the Elizabeth Islands, Nantucket, and Martha's Vineyard.


    We analysed pollen records from 29 study sites, using multivariate cluster analysis to visualize changes in the composition and spatial patterns of vegetation during the last 14,000 years. The pollen data were compared with temperature and precipitation reconstructions.


    Boreal forest featuringPiceaandPinus banksianawas present across the region when conditions were cool and dry 14,000–12,000 calibrated14C years before present (ybp).Pinus strobusbecame regionally dominant as temperatures increased between 12,000 and 10,000 ybp. The composition of forests in inland and coastal areas diverged in response to further warming after 10,000 ybp, whenQuercusandPinus rigidaexpanded across southern New England, whereas conditions remained cool enough in inland areas to maintainPinus strobus. Increasing precipitation allowedTsuga canadensis,Fagus grandifolia, andBetulato replacePinus strobusin inland areas during 9,000–8,000 ybp, and also led to the expansion ofCaryaacross the coastal part of the region beginning at 7,000–6,000 ybp. Abrupt cooling at 5,500–5,000 ybp caused sharp declines inTsugain inland areas andQuercusat some coastal sites, and the populations of those taxa remained low until they recovered around 3,000 ybp in response to rising precipitation. Throughout most of the Holocene, sites underlain by sandy glacial deposits were occupied byPinus rigidaandQuercus.

    Main conclusions

    Postglacial changes in the composition and spatial pattern of New England forests were controlled by long‐term trends and abrupt shifts in temperature and precipitation, as well as by the environmental gradient between coastal and inland parts of the region. Substrate and soil moisture shaped landscape‐scale variations in forest composition.

    more » « less
  2. Abstract Aim

    This study examines how climate shapedMicrotus californicus(Rodentia: Arvicolinae) ecomorphology throughout the Quaternary. It tests three hypotheses: (a) climate corresponds with consistent shape variation inM. californicusdentition; (b) Quaternary warming and drying trends causedM. californicusmorphotypes to predictably shift in range through time and (c) Quaternary warming and drying led to predictable changes in tooth morphological variation. Finally, we discuss how shifts in climate‐linked morphological variation may affect the potential ofM. californicusto react to future climate change.


    Western United States.


    Microtus californicus(Peale, 1848).


    Geometric morphometrics and partial least squares analyses were used to discern how climate contributes to consistent variation in the shapes of theM. californicuslower first molar (m1), validated for the full toothrow. We further corroborate this relationship, reconstructing precipitation at fossil localities using m1 morphology and comparing those values to palaeoclimate‐model‐derived precipitations. Disparity analyses and a MANOVA were performed to examine changes in variation and whether a shift in tooth shape occurred through time.


    Microtus californicusm1 and toothrow shapes are narrower and more curved in cooler, wetter climates. Morphology‐based palaeoclimate reconstructions align with model‐based palaeoclimate estimations. When time averaging is accounted for,M. californicusdemonstrates a 12% reduction in variation from fossil to present‐day specimens, and these changes in tooth shape correspond with climate‐related morphotypes.

    Main conclusions

    As California became drier and hotter since the late Pleistocene,M. californicusdental morphology generally tracked these changes by adapting to the consumption of rougher vegetation in drier environments. This resulted in the loss of some high‐precipitation morphotypes, indicating that ecomorphology, often observed at the species and community levels, translates to intraspecific variation and dynamically changes in response to changing climates. The loss of climate‐linked morphological variation since the late Pleistocene may limit the ability ofM. californicusto respond to future changes in climate. These findings portend that other species may have experienced similar losses in adaptability.

    more » « less
  3. In north-western North America, the so-called divergence problem (DP) is expressed in tree ring width (RW) as an unstable temperature signal in recent decades. Maximum latewood density (MXD), from the same region, shows minimal evidence of DP. While MXD is a superior proxy for summer temperatures, there are very few long MXD records from North America. Latewood blue intensity (LWB) measures similar wood properties as MXD, expresses a similar climate response, is much cheaper to generate and thereby could provide the means to profoundly expand the extant network of temperature sensitive tree-ring (TR) chronologies in North America. In this study, LWB is measured from 17 white spruce sites ( Picea glauca) in south-western Yukon to test whether LWB is immune to the temporal calibration instabilities observed in RW. A number of detrending methodologies are examined. The strongest calibration results for both RW and LWB are consistently returned using age-dependent spline (ADS) detrending within the signal-free (SF) framework. RW data calibrate best with June–July maximum temperatures (Tmax), explaining up to 28% variance, but all models fail validation and residual analysis. In comparison, LWB calibrates strongly (explaining 43–51% of May–August Tmax) and validates well. The reconstruction extends to 1337 CE, but uncertainties increase substantially before the early 17th century because of low replication. RW-, MXD- and LWB-based summer temperature reconstructions from the Gulf of Alaska, the Wrangell Mountains and Northern Alaska display good agreement at multi-decadal and higher frequencies, but the Yukon LWB reconstruction appears potentially limited in its expression of centennial-scale variation. While LWB improves dendroclimatic calibration, future work must focus on suitably preserved sub-fossil material to increase replication prior to 1650 CE. 
    more » « less
  4. Abstract

    The oxygen isotopic composition of planktic foraminiferal calcite () is one of the most prevalent proxies used in the paleoceanographic community. The relationship between, temperature, and seawater oxygen isotopic composition () is firmly rooted in thermodynamics, and experimental constraints are commonly used for sea surface temperature (SST) reconstructions. However, in marine sedimentary applications, additional sources of uncertainty emerge, and these uncertainty constraints have not as of yet been included in global calibration models. Here, we compile a global data set of over 2,600 marine sediment core top samples for five planktic species:Globigerinoides ruber,Trilobatus sacculifer,Globigerina bulloides,Neogloboquadrina incompta, andNeogloboquadrina pachyderma. We developed a suite of Bayesian regression models to calibrate the relationship betweenand SST. Spanning SSTs from 0.0 to 29.5 °C, our annual model with species pooled together has a mean standard error of approximately 0.54‰. Accounting for seasonality and species‐specific differences improves model validation, reducing the mean standard error to 0.47‰. Example applications spanning the Late Quaternary show good agreement with independent alkenone‐based estimates. Our pooled calibration model may also be used for reconstruction in the deeper geological past, using modern planktic foraminifera as an analog for non‐extant species. Our core top‐based models provide a robust assessment of uncertainty in thepaleothermometer that can be used in statistical assessments of interproxy and model‐proxy comparisons. The suite of models is publicly available as the Open Source software librarybayfox, for Python, R, and MATLAB/Octave.

    more » « less
  5. Abstract

    Selection pressures along climate gradients give rise to predictable variation in plant functional traits of individual species suggestive of local adaptation. Species whose ranges include winter rainfall, Mediterranean climates, or other strongly seasonal climates, may be exposed to divergent selection pressures at different ends of seasonality gradients.

    Here, we evaluate how rainfall seasonality in conjunction with other key climatic variables impacts patterns of trait variation inPelargonium scabrum, a woody shrub from the Greater Cape Floristic Region of South Africa. This biodiversity hotspot encompasses a Mediterranean climate (wet winters and hot, dry summers) and displays steep gradients in temperature and water availability.

    We used Bayesian regression models to evaluate leaf trait–trait and trait–climate relationships among 26 populations. Models included rainfall seasonality and its interaction with other climate variables (mean annual temperature, mean annual precipitation and potential evapotranspiration) as predictors to test for the impact of climate variation on three leaf traits: size, dissection and leaf mass per area (LMA). We evaluated model explanatory power by calculating BayesianR2values, and predictive power via leave‐one‐out cross‐validation.

    Trait–trait associations were modulated by rainfall seasonality, including a reversal in the relationship between leaf size and dissection depending on the proportion of rain received in winter. Trait–climate models were improved by including rainfall seasonality as a predictor for both explanatory and predictive power. For leaf dissection and LMA, we detected significant interactions between rainfall seasonality and other environmental variables, leading to reversals in the relationships between these traits and the three environmental variables depending on the proportion of winter rainfall.

    Differences in the timing of rainfall, coupled with strong differences in the covariation of climate variables, impose divergent selection pressures onP. scabrumpopulations resulting in divergence of trait values, trait integration and responses to climate gradients. These patterns are consistent with local adaptation ofP. scabrumpopulations mediated by the interactions between temperature and the amount and timing of rainfall. Species arrayed along broad climate gradients represent an excellent opportunity for investigating patterns of trait variation and abundances and distributions of species in relation to future changes in climate.

    A freePlain Language Summarycan be found within the Supporting Information of this article.

    more » « less