Earth and Planetary Science Letters 284 (2009) 544–552

Contents lists available at ScienceDirect

Earth and Planetary Science Letters

j ourna l homepage: www.e lsev ie r.com/ locate /eps l

Present-day climatic equivalents of European Cenozoic climates

Torsten Utescher a,⁎, Volker Mosbrugger b, Dimiter Ivanov c, David L. Dilcher d

a Steinmann Institute, Geological Department, Bonn University, Nussallee 8, Bonn, 53115, Germanyb Senckenberg Research Institute and Natural History Museum, Senckenberganlage 25, Frankfurt/M., 60325, Germanyc Institute of Botany, Bulgarian Academy of Sciences, Acad. G. Bonchev Str., 23, BG-1113 Sofia, Bulgariad Museum of Natural History, University of Florida, Gainesville, FL 32611-7800, USA

Coexistence approach

⁎ Corresponding author. Tel.: +49 228 739773; fax: +E-mail addresses: utescher@geo.uni-bonn.de (T. Ute

volker.mosbrugger@senckenberg.de (V. Mosbrugger), ddilcher@flmnh.ufl.edu (D.L. Dilcher).

0012-821X/$ – see front matter © 2009 Elsevier B.V. Adoi:10.1016/j.epsl.2009.05.021

a b s t r a c t

a r t i c l e i n f o

Article history:Received 17 September 2008Received in revised form 15 April 2009Accepted 14 May 2009Available online 16 June 2009

Editor: P. DeMenocal

Keywords:PalaeoclimatePalaeobotanyCenozoicNorth Germany

Recently, continental climate evolution in Central Europe over the last 45 Ma has been reconstructed fromthe palaeobotanical record using a Nearest Living Relative methodology (Coexistence Approach; CA). Thereconstructed climate curves document in detail the transition from almost tropical conditions in the Mid-Eocene to a temperate climate at the Pliocene/Pleistocene transition. The observed climatic shifts areprimarily expressed as non-proportional changes of the different variables taken into account. In the presentstudy a published palaeoclimate data set for a total of 42 macrofloras complemented by new calculations isused as base to analyse the climatic space in which a fossil flora existed. To define these spaces CA intervalscalculated for 3 temperature (mean annual temperature, cold and warm month mean) and 3 precipitationvariables (mean annual precipitation, mean monthly precipitation of the driest and of the wettest month)are combined. Using a global gridded climatology (10' resolution), this climate space is then utilized toidentify Recent climate analogues with respect to the variables regarded. For 18 macrofloras climaticanalogue regions with respect to 6 variables are identified on the globe. For 16 macrofloras, analogues existwhen three temperature parameters and mean annual precipitation are regarded. No Recent equivalents arefound in 8 cases. This corroborates the assumption of the temporary existence of non-analogue climates inthe Cenozoic. As shown by multivariate statistics the observed anomalies with respect to present-dayconditions basically refer to high winter temperatures.Deploying a GIS, the Recent climate analogues can be presented as sets of grid cells for each flora that can bemapped on a globe. Once identified, these regions can be merged with adequate thematic layers to assessadditional proxy data for the palaeofloras. To exemplify the procedure Koeppen climate type, numbers ofdays with ground frost, as well as Matthews biome classes are reconstructed. The resulting Koeppen types,ranging from A to C, are largely consistent with data previously published. The ground frost record showsalmost frost-free conditions for the Mid-Eocene greenhouse and for the Mid-Miocene Climatic Optimum. Forthe Chattian, up to 7 days with ground frost result. During the Late Miocene Cooling the number of days withground frost significantly increased. The inferred Matthews biomes reveal changing patterns of evergreenand deciduous forest cover.When Recent climate analogues are found for a fossil flora, present climate can be used to calibrate theoriginal CA data. It is shown that calibration considerably improves the resolution of the continental climaterecords.

© 2009 Elsevier B.V. All rights reserved.

1. Introduction

Recently, continental climate records over the last 45 Ma havebeen reconstructed from the macrofloral record of Central andNorthwest Europe (Mosbrugger et al., 2005). The technique appliedis the Coexistence Approach (CA; Mosbrugger and Utescher, 1997), aquantitative method for calculation of various climate variables. This

49 228 739037.scher),imiter@bio.bas.bg (D. Ivanov),

ll rights reserved.

approach has proven a useful tool for quantitative palaeoclimateanalysis in the Cenozoic (e.g., Uhl et al., 2003; Roth-Nebelsick et al.,2004; Bruch et al., 2007; Bozukov et al., 2009). The temperaturerecords obtained reveal continental cooling from almost tropicalconditions in the middle Eocene to temperate climate at thePliocene/Pleistocene transition. Major global climatic eventsknown from marine data archives (e.g., Zachos et al., 2001; Zachoset al., 2005) such as the Late Eocene Cooling and the Mid-MioceneClimatic Optimum are mirrored in the terrestrial curves. It was alsoshown that climate variability reflects non-proportional changes ofdifferent variables. The most prominent changes are observed inwinter temperatures, while summer temperatures and mean annual

545T. Utescher et al. / Earth and Planetary Science Letters 284 (2009) 544–552

precipitation stayed more or less stable until the Pliocene. In Centraland Northwest Europe, cooling during the Late Miocene is connectedto an increase in seasonality of climate.

Although these records document detailed time series, there isevidence from palaeoclimate studies of various Cenozoic time slicesthat spatial climate patterns could considerably differ from today (e.g.,Bruch et al., 2004; 2006). Also, model studies carried out for variousCenozoic scenarios do not fully explain the observed proxy-data basedpatterns (e.g., Steppuhn et al., 2009;Micheels et al., 2007). Thus, manyaspects of the functioning of the climate system in the Cenozoic arenot well understood.

The CA provides quantitative data for basic climate parameters,such as mean annual temperature, mean temperatures of the coldestand the warmest months as well as annual and monthly precipitation.Each variable is calculated independently using climatic requirementsof all Nearest Living Relatives (NLRs) known for a fossil flora(Mosbrugger and Utescher, 1997). Here we use published palaeocli-mate data and records for the North German Cenozoic (Utescher et al.,2000;Mosbrugger et al., 2005) complemented by new calculation.Wecombine 6 basic climate variables to define a climatic space in whichthe fossil flora existed to address the following questions:

• Are there present-day climate equivalents for the inferred palaeo-conditions?

• Can these be used for interpreting Cenozoic climates?• How usual/unusual are the inferred palaeoclimate conditions in thecontext of present-day climates and what anomalies are found?

• Can present-day climate equivalents, once identified, be used tocalibrate the results obtained from the CA? Are they useful to gainaccess to additional proxies for a fossil flora?

Fig. 1. Palaeogeographical map of North Germany and neighbouring areas showing the O

2. Materials and methods

2.1. Floral record

In the present study, climate data obtained for a total of 42macrofloras floras are studied. The macrofloras (fruits and seed, leaffloras) originate from the North German Cenozoic (Fig. 1; Table 1) andare the same that have been used to reconstruct the climate recordspublished in Utescher et al. (2000) and Mosbrugger et al. (2005). Forfurther details on the floras and the complete set of palaeoclimate datathe reader is referred to the above publications. For 23 of the florasadditional precipitation data had to be calculated. The completepalaeoclimate data sets for the floras are given in the supplementaryonline materials (1), together with the new results obtained in thisstudy.

2.2. Definition of the climatic space — identification of present-dayequivalents

As stated above all the palaeoclimate data used in the followingoriginate from the application of the Coexistence Approach (CA;Mosbrugger and Utescher, 1997) on the palaeobotanical record. TheCA employs climatic requirements of the Nearest Living Relatives(NLRs) of fossil taxa to calculate for a specified climate variable, theintervals in which a maximum number of the NLRs known for thefossil flora may coexist. In the CA procedure, these ranges are iden-tified separately for each climate parameter.

To define the climatic space in which the fossil flora existed, weselect six climate parameters frequently employed when describingpalaeoclimate conditions, i.e., mean annual temperature (MAT), cold

ligocene configuration. Study area: LRB: Lower Rhine Basin; WB: Weisselster Basin.

Table 1Macrofloras analysed and stratigraphical ages.

Macrofloras analysed

Code Locality Referencesa Stratigraphic age

Y Tegelen 1 Late Pliocene/Pleistocene (Tiglian)P Fortuna Garsdorf

Mine/11 horizon2 Piacenzian, MN 16a/b

O Hambach Mine/10 horizon 2 Early Piacenziann Hambach Mine/9C horizon 2 Late Zanclean (Brunssumian)m Hambach Mine/9B horizon 2 Zanclean (Brunssumian)l Frechen Mine/9A horizon 2 Early Zancleank1 Garzweiler Mine/upper

Hauptkies Fm.2 Messinian

k Hambach Mine/lowerHauptkies Fm.

2 Messinian

j Hambach Mine/7F horizon,top

2 Late Tortonian

i Hambach Mine/7F horizon,base

2 Late Tortonian

h1 Frechen Mine/7D horizon 2 Later Tortonianh Zukunft Mine/7A horizon 2 Early Tortoniang Zukunft Mine/6E horizon 2 Serravallian, NN8/NN9 transitionf Zukunft Mine/6D horizon 2 Langhian/Serravallian transitione Hambach Mine/6C horizon 2 Langhian, late MN 5d Zukunft Mine/6A horizon 2 Late Burdigaliand1 Theresia Mine, base of

Main Seam2 Late Burdigalian

c Fortuna Garsdorf Mine/5D 2 Burdigalianb Rott 2 Latest Chattian, MP 30a Enspel 2 Chattian, 25.8 Ma (radiometric)X Klettwitz 12 1 Earliest Tortonian (?)W Bröthen, Upper Seam 1 Late SerravallianV Klettwitz 3 1 Late Langhian/early SerravallianU Klettwitz 5 1 Late Langhian, NN5T Kleinleipisch Mine 1 Langhian (?), older than code US Seese Mine 1 Early/middle Miocene transitionR Schlabendorf well 1 Early/Middle Miocene transition,

older than code SQ Wiesa, basal part 1 Early/Middle Miocene transition,

older than code RP Brandis 1 Hemmorian, NN2 (?)O Bitterfeld, clay 1 Early Miocene, NN1 (?)N Bitterfeld, Main Seam 1 Early Miocene, NN1 (?), older than

code OM Bitterfeld, Lower Seam 1 Early Miocene, NN1 (?), older than

code NL Mockrhena 1 1 Oligocene/Miocene transition (?)K Kleinsaubernitz, base 1 Late ChattianJ Bockwitz Mine 1 Chattian, older than code KH Nerchau 1 RupelianG Boehlen 1 Late PriabonianF Haselbach, Zeitz Sand 1 Priabonian, older than code GE Knau gravel pit 1 Early Priabonian, older than code FC Profen 1 1 BartonianB Scheiplitz 1 Early Bartonian, older than code CA Geiseltal, base of section 1 Lutetian, MP12-13

a Palaeoclimate data published in (1) Mosbrugger et al. (2005); (2) Utescher et al.(2000).

546 T. Utescher et al. / Earth and Planetary Science Letters 284 (2009) 544–552

month mean (CMM), warm month mean (WMM), mean annualprecipitation (MAP), mean monthly precipitation of the driest(MPdry) and of the wettest month (MPwet). To avoid unnaturalrestrictions caused by uncertainties inherent to the CA methodologyitself, minimum widths Wmin are set for the width of a single CAinterval (WCA) with 1 °C for temperature variables,10mm for monthlyand 100 mm for annual precipitation. These values represent theclimatic resolution of the CA (Mosbrugger and Utescher, 1997). If WCA

falls below the above thresholds new intervals are then determinedusing Wmin together with the CA interval mean.

In a next step, present regions with equivalent climatic conditionsregarding the 6 variables are identified. This is done using a 10' griddedclimatology (New et al., 2002) and a GIS (ArcView). Grid points withaltitudes >1000 m are excluded from the analysis. Mountain climateshave their special characteristics and are therefore considered as in-

adequate for lowland sites as analysed in this study. As a result, a data setcontaining grid points and corresponding climate variables is obtainedfor each flora, interpreted as present-day analogue for the fossil climate.This analogy certainly does not mean true climatic equivalence withrespect to the entire complexity of the climate system, because only 6variables are considered here. In addition, gridded data only representan approximation to “real” measured values and have a heterogeneousreliability depending on spatial density and quality of measurements atthe available observation points (cf. New et al., 2002).

If a present-day analogue is found the climatic space, once iden-tified for a fossil flora, can be used to calibrate the palaeoclimate dataoriginally obtained from the Coexistence Approach. The calibratedcoexistence intervals (CAcal) for MAT; CMM;WMM;MAP; MPdry; andMPwet are obtained as 1-dimensional projections of the climaticspace. The resulting calibrated CA ranges are interpreted as the mostprobable solution for the fossil climate with respect to present-dayconditions.

Interestingly, it is shown that, for part of the palaeofloras, nomodernclimates are found that reflect the palaeo-conditions as calculated bythe CA. The use of a 4-dimentional climatic space defined byMAT, CMM,WMMandMAP then allows calibration inmost of these cases but undera reduced significance. If this fails a meaningful calibration cannot beperformed. To study the anomaly of fossil climates without presentanalogues principal component analysis is used (PCA; PAST program).

2.3. Access to additional proxies

Once identified, present-day climatic analogues can be used togain access to a variety of proxy data that today correspond to pa-laeoclimate conditions the inferred to a fossil flora. This is achievedby merging their coordinates with any thematic layer using a GIS(ArcView). To test the ability of the approach Koeppen climate typesare determined for the floras studied using a data set by Kottek et al.(2006). Also a record for ground frost duration (data from New et al.,2002) is generated for the North German Cenozoic. The clusters ofgrid cells representing the modern analogues provide easy access tovegetational proxies as well. In the same way we extract potentialvegetation using the Matthews classification system (Matthews,1983; data set: UNEP grid, 2008).

3. Results and discussion

3.1. Identification of modern climate equivalents — calibration of climaterecords based upon macrofloras

The climate records of the North German Cenozoic are shownin Figs. 2 and 3 for MAT, CMM,WMM, andMAP (Utescher et al., 2000;Mosbrugger et al., 2005). Among the 42 macrofloras analysed(Table 1), present-day climatic equivalents with respect to the 6variables are found in 18 cases. For 16 floras Recent equivalents existin a four-dimensional climatic space (MAT, CMM, WMM and MAP).For 8 floras no corresponding Recent climate space is found andhence calibration procedure cannot be performed.

The probability under which the algorithm can detect equivalentconditions certainly depends on the widths of CA intervals. Thisrelation is illustrated on Fig. 4(A) where standardized 4-dimensional(MAT, CMM, WMM and MAP) variable spaces obtained for the fossilfloras are plotted against the number of grid cells with correspondingvalues found in the Recent climatology. The relation is linear for largeCA intervals, for small CA intervals and variable spaces the scatter hasa more random aspect.

Fig. 4(B) shows the number of cases when climatic equivalents areidentified, related to the 4-dimensional variable spaces describedabove. The spaces are grouped into 3 equally sized classes (small,medium, large). It is shown that small spaces as resulting from narrowCA ranges, close to the resolution limits of the method, are commonly

Fig. 2. Climate evolution in the Cenozoic: Continental cold month mean record for North Germany and global oxygen isotope record from Zachos et al. (2001). The original recordfrom Mosbrugger et al. (2005) is shown in grey, calibrated data are in colour. For code numbers cf. Table 1. WB: Weisselster Basin; LRB: Lower Rhine Basin.

547T. Utescher et al. / Earth and Planetary Science Letters 284 (2009) 544–552

too restrictive to calibrate in a 6-dimensional space (10% of the cases).Here, mostly analogues with respect to the four variables (MAT, CMM,WMM, MAP) are found (60% of cases). For small spaces no analoguesare found in 30% of cases, decreasing to ca.15% when variable spaces

Fig. 3. Climate evolution in the Cenozoic: North German continental records for mean annuafurther explanations cf. Fig. 2.

are medium sized. As expected, 6-dimensional calibration works wellfor wide CA ranges, but under a lower climatic resolution.

Principal component analysis (PCA, PAST program) is used in orderto visualize the spatial arrangement of the fossil data with respect to

l temperature, warm month mean and mean annual precipitation. For colour codes and

Fig. 4. A) Standardized 4-dimensional variable spaces roughly approximated as product of MAT, CMM,WMM andMAP (using the original CA intervals for the 42 floras), in each casenormalized by the total range of the variable, plotted against the number of climatically equivalent grid cells found in the Newet al. (2002) data set. B) Variable space classes obtainedfrom A and number of cases where climatic equivalents are identified regarding a 4-dimensional or 6-dimensional space, respectively.

548 T. Utescher et al. / Earth and Planetary Science Letters 284 (2009) 544–552

Present. For the palaeoclimate datameans of CA intervals are used. TheRecent climatology (New et al., 2002) is upscaled to a 30' resolutionfor data processing reasons. As evident from Fig. 5 the palaeo-sites plotamong Recent data. The fossil climates all are positioned in the warm,wet part of the datafield (negative loadingswith respect to component2) and mostly have higher positive loadings with respect to compo-nent 3 (high precipitation in the wettest month).

Most of the floras with present-day equivalents in a 6-dimensionalspace originate from the middle Eocene, the time-span of the LateOligocene to the Burdigalian and the Messinian. For Tortonian andPliocene floras, modern reference climates aremostly identified in a 4-dimensional space, although a closer similarity of the climate systemto present conditions could be expected (Fig. 2). As the primary causethe narrow CA ranges obtained for the floras can be cited (see above).

Fig. 5. Principal Component Analysis (PCA, PAST program) using the correlation matrixof the variables MAT, CMM,WMM,MAP, MPdry, andMPwet. The scatter diagram showsthe Recent climate field (New et al. (2002), upscaled to a 30' resolution) together withthe 42 palaeo-sites (red) as well as projections of the original axes (variables). Highnegative loadings regarding component 2: warm and wet; high positive loadingsregarding component 3: high precipitation in the wettest month.

The anomalies of fossil climates without Recent equivalents becameobvious when regarding the temperature space (MAT, CMM, WMM;Fig. 6). 8 of the floras have very high negative loadings of component 3,correlated with CMM, and clearly plot outside the scatter of the Recentdata. CMM obviously is a crucial factor when no Recent analogues aredetected, thus pointing to the temporary existence of non-analogueconditions in the past. Palaeoclimates without present climate equiva-lents occur in exceptionally warm periods of the Miocene, the Mid-Miocene Climatic Optimum (e.g., Zachos et al., 2001) and the warmingevent in the latest Tortonian (Mosbrugger et al., 2005). These obviouslyare time-spans with very mild winters and more equable temperaturesthroughout the year when compared to today. The existence of non-analogue climate conditions in the Cenozoic has been assumed by

Fig. 6. Principal Component Analysis (PCA, PAST program) using the correlation matrixof the variables MAT, CMM, and WMM. The scatter diagram shows the Recent climatefield (New et al. (2002), upscaled to a 30' resolution) together with the 42 palaeo-sites(red) as well as projections of the original axes (variables). High positive loadingsregarding component 1: warm conditions; high negative loadings regarding compo-nent 3: high temperatures in the coldest month. Codes given for some of the fossil datapoints refer to Table 1.

Fig. 7. Present-day climatic analogues for the North German Cenozoic palaeoclimates regarding 6 climate variables. Grid points are identified on the basis of the CA ranges obtainedfor MAT, CMM, WMM, MAT, MPdry and MPwet using the New et al. (2002) climatology. For a better visualization clusters of single grid cells are replaced by (bigger) solid circles.

549T. Utescher et al. / Earth and Planetary Science Letters 284 (2009) 544–552

various authors (e.g., Mosbrugger et al., 2005; Williams and Jackson,2007; Boehme et al., 2008).

The calibrated climate data obtained for the macrofloras, as well asthe improvement as regards the climatic resolution, are given thesupplementary online materials (1 and 2), next to the original CA dataas published in Utescher et al. (2000) and Mosbrugger et al. (2005).

Using 6 climate variables for calibration the width of MAT rangesis reduced by 72.9% compared to the primary CA ranges. As shownby high standard deviations (cf. Supplementary materials 2) widthreduction is highly variable. The improvement of the climatic res-olution is best for MAT (72.9% at a mean) while for WMM approx-imately 50% are reached.

As shown by the resulting records (Figs. 2 and 3), the intervals ofthe climate variables obtained from calibration are not only reduced inwidth but often have a pointed asymmetrical position within theoriginal CA ranges. Thus, considering probabilities in the context ofRecent climate, additional aspects in the interpretation of the recordsare revealed. In the time-span from the Mid-Eocene to Rupelian, thecalibration procedure primarily selects probability ranges within thewarm and wet ends of the CA intervals (exception: Scheiplitz flora,code B) while in the Chattian, the cool and dry ends are selected (cf.Figs. 2 and 3). As indicated by the data cooling in the later Chattianwasfirst in CMM (Bockwitz flora, code J) followed by a pointed decrease inMAT and WMM (Kleinsaubernitz flora, code K). Calibrated data of theWeisselster record then showa short-term temperature shift of almost4 °C (CMM) right after theMi-1 glaciation. Calibration reveals regionaldifferentiation between the records. In the Burdigalian and Langhian,MAP tends to be higher in the WB when compared to the LRB. Duringthe Mid-Miocene Climatic Optimum, WMM in the LRB tends to behigher than in the Weisselster Basin (cf. floras from the Rhenish MainSeam, code e; f). In both regions, winter temperatures did not exceed12 °C at that time and were well below the Eocene level. Concerningthe late Neogene part of the climate record, calibration reveals furtherdetail. The temperaturedecrease at the beginningof theMessinianwasmost pronounced with CMM dropping by over 4 °C (lower Hauptkiesflora, code k) followed by a significant decrease in MAP by around

200mm(upper Hauptkies flora, code k1). In the earliest Pliocene,MATand CMM most probably did not reach the Miocene level, stayingbelow 15 °C and 5 °C, respectively.

3.2. Access to other proxies

3.2.1. Present regions with reference climates and Koeppen climate typeUsing a global data set of Koeppen climate type (Kottek et al., 2006)

together with the climatic space identified, the climate type repre-sented by a fossil flora can be identified. This is done for the 18 floraswhere reference climates are identified for all 6 variables used in theanalysis. The results obtained are summarized in the supplementaryonline materials (3). In the Lutetian, the warmest time interval of therecord, climatically equivalent regions presently exist in Cuba and NMexico,while for the late Eocene Scheiplitz andBoehlenfloras, presentconditions in SE China correspond. In the cool phase of the LateOligocene, reference climates are found Southern Europe, along theAdriatic coast and the south coast of the Black Sea, as well in theSoutheast USA (Fig. 7). Themiddle Eocene climate as reconstructed forthe Geiseltal flora (code A) is of a (“para”) tropical type (Af; AwKoeppen type), the late Eocene floras (code B; G) existed under Cfa/Cwa conditions. The late Oligocene Bockwitzflora (code J) correspondsto a Cfa type climate, for the slightly youngerRottflora (codeb) also theCsa type is possible indicating hot, drier summers. The comparativelywide climatic range of the latest Oligocene Kleinsaubernitz flora (codeK) also includes the oceanic Cfb type. Koeppen types inferred to 10Neogene floras with equivalents in a 6-dimensional climatic space(code M: N; O; P; R; S; V; f; h1; k) all belong to the humid Cfa typecharacterized by uniform precipitation distribution. For the base of theRhenishMain Seamand the laterMessinian (floras TheresiaMineflora,code d1GarzweilerMine, upper part of theHauptkies Fm., code k1) theCwa type is also possible. Early to Mid-Miocene climates have theirrecent counterparts in the SE USA, in coastal areas of E China and inSouth China (Guangxi) (Fig. 7), for the Messinian floras grid cells in SEHubei (China) are specified. With the Exception of the Geiseltal florapreviously reported to represent a Cwa type climate, there is a broad

550 T. Utescher et al. / Earth and Planetary Science Letters 284 (2009) 544–552

agreement between the present data and results from literature byvarious authors (cf. compilation of climate type data in Mai, 1995;Supplementary materials 3).

3.2.2. Additional climate parameters— number of days with ground frostRecent reference climates for the Palaeo-sites regarding at least

four climate variables (MAT, CMM, WMM, MAP) are used toreconstruct the likely number of days with ground frost (Fig. 8). Asexpected, major trends of continental climate evolution in the studyarea are reflected. Two obviously frost-free time-periods result for thecontinental North German Cenozoic: The middle Eocene, wherealmost tropical conditions are reconstructed and no ice-sheets existed,and the Mid-Miocene Climatic Optimum during the Langhian, thelatter, however, only in the Lower Rhine Basin, maybe due to itswesterly exposure. According to the present data, the number of dayswith ground frost increased from the early Oligocene on, peaking in theChattian. For the Bockwitz flora (code J), up to 7 days are inferred. Alsothere is evidence from foliar physiognomic studies for a rare presenceof days with ground frost (0–36 days) from the Chattian localityMonod–Rivaz (Traiser et al., 2004). At the time of the Late OligoceneWarming known from the marine part (Zachos et al., 2001) thecontinental ground frost record shows a decreasing trend (codeK). Thesubsequent increase (codes L; M) peaking in the early Aquitanian ispossibly related to the Mi-1 Glaciation Event.

During the Late Miocene Cooling, the probability for ground frostrapidly increased, attaining a range of 9 to 11 days near the Pliocene/Pleistocene boundary (Tiglian flora). A decreasing trend at thebeginning of the Pliocene can be correlated to the warm phase in theearly Zanclean. Also, leaf physiognomic data point out a similar in-crease (Erdobenye Kovago-oldal flora, late Miocene, Hungary: ca.30 days with frost; Traiser et al., 2004). Vertebrate faunas indicate that

Fig. 8. Number of days with ground frost in the North German Cenozoic inferred from climatcalibrated using MAT, CMM, WMM, MAP, MPdry, MPwet; yellow bars: CA intervals calibrat

periodic frosts occurred in Central Europe from the Serravallian on(Boehme, 2003).

3.2.3. Present reference biomesVegetation proxies obtained from this approach are meaningful

because they can be crosschecked with qualitative descriptions ofthe palaeo-vegetation available for most of the fossil floras and thusprovide an opportunity to verify the validity of the whole climatereconstruction.

For 18 floras, present reference biomes with respect to the 6climate variables presently considered are presented in Fig. 9. Thebars in each case represent percentages of grid cells within the dis-tribution of the different biome type (Matthews classification;Matthews, 1983). According to the inferred palaeoclimate data, trop-ical evergreen broadleaved forest would have existed in the middleEocene of North Germany. This coincides with the characterization ofthe Messel–Selsey Flora Complex reported as evergreen “paratropi-cal” broadleaved forest (Mai,1995). Climate conditions reconstructedfor the early late Eocene Scheiplitz flora infer a cooler variety of theevergreen broadleaved vegetation with seasonal character, thetropical/subtropical evergreen seasonal broadleaved forest biomeaccording to Matthews. According to Mai and Walther (2000) theScheiplitz flora represents an evergreen, physiognomically differ-entiated, vegetation. Interestingly, there is evidence for seasonaldrought by the occurrence of notophyllous taxa.With the subsequentcooling towards the end of the Paleogene, climate conditionsincreasingly favour deciduous forests with evergreens. For theBoehlen, Kleinsaubernitz and Rott floras Matthews biome 10 (colddeciduous forest with evergreens) is obtained. For the Bockwitz flora,a deciduous forest formation, even without evergreens, correspondsto the climate data, in addition the 23 Matthews biome indicating a

ically equivalent using the present climatology (New et al., 2002). Red bars: CA intervalsed using MAT, CMM, WMM, MAP.

Fig. 9. Cold month mean record (cf. Fig. 2) and Matthews biomes presently corresponding with the inferred palaeoclimate conditions. Percentages given in the bar chart refer torelative frequencies of grid cells obtained when overlaying climatic reference regions of the single floras and the gridded biome data.

551T. Utescher et al. / Earth and Planetary Science Letters 284 (2009) 544–552

woody cover below 40%. For the Chattian Rott–Thierbach FloralComplex (Bockwitz, Kleinsaubernitz, and Rott floras in the record),dominance of summergreen taxa, is described for the first time invegetation history of in Central Europe (e.g., Mai, 1995). Besidesazonal communities the Rott flora evidences a diverse mixed meso-phytic upland vegetation with deciduous and laurophyllous taxa(Mosbrugger, 1989). The lower Miocene climate conditions firstfavour deciduous vegetation with evergreens (Bitterfeld 1 and 2floras), towards the Mid-Miocene also tropical/subtropical and tem-perate evergreen seasonal broadleaved forest (Bitterfeld clay,Brandis, Theresia Mine, Zukunft 6D, Klettwitz 3). According to pre-vious results, the different Bitterfeld levels yield a diverse Arctoter-tiary flora with admixed laurophyllous elements (Mai, 1995). Asknown from palynological studies of the Rhenish Main Seam (cf.Theresia Mine flora), the flora was dominantly evergreen, althoughazonal, with percentages of Palaeotropical elements around 40%(Ashraf et al., 1997).

For the later Tortonian and the Messinian floras (codes h1; k; k1),deciduous or evergreen forest vegetation result. This is in contra-diction to the palaeobotanical record where a deciduous forest isdescribed (Belz and Mosbrugger, 1994). However, the vegetation des-cribed there represents a riparian forest typically composed of decid-uous taxa and does not allow for conclusions as regards the zonalvegetation. Palynological records, in contrast, show still comparativelyhigh percentages of evergreen angiosperms in the Tortonian. In thelatest Miocene, however, their record is scanty (Ashraf et al., 1997).

4. Conclusions

Present day climate equivalents for the Cenozoic of North Germanyas reconstructed by the Coexistence Approach are found in 34 out of42 cases. The probability of identification of such Recent analoguesstrongly depends on the widths of the original CA intervals calcu-lated for a fossil flora. Analogue conditions commonly are detected forPalaeogene floras where mostly wide CA intervals are obtained. When

analysing Neogene floras, CA intervals tend to be narrower, becausethe taxonomic resolution continuously improves with decreasing age(Mosbrugger and Utescher, 1997). This may explain the fact that formany of the Neogene floras analysed, present-day climate equivalentsare only found in a 4-dimensional climatic space. In some cases, highlyresolved CA data are too restrictive in the procedure and no equiv-alents are found, even when using a high resolving climatology. Inseveral cases, no climatic equivalents are identified, although the CAranges are reasonably wide. This may point to the existence of unusualor unique continental climate conditions. As shown by the analysisthese non-analogue climates were, however, comparatively close toRecent conditions.

When identified, present-day climatic equivalents of Cenozoicclimates first of all provide Recent reference regions for the fossilconditions. Using GIS provides an easy access to a variety of additionaldata such as bioclimatic and vegetational proxies presently correspond-ing to the inferred palaeo-conditions. These data can help to interpretpast climates and can partly be crosschecked with results obtained byother approaches.

Using present reference climates CA intervals can be calibrated.Regarding a 6-dimensional space calibration can considerably improvethe climatic resolution of the CA by up to 73%. The resulting CAcalib

intervals are interpreted as the most probable solution of variable com-bine. Another advantage of the procedure is the fact that – as in the CA –

climatic tolerances of plants are considered only with respect to singlevariables. Thus, the results obtained do not depend on shapes andoverlappingof presentplant distribution areas,which are largely a resultof the perturbations caused by numerous strong glacial events duringlatest Earth History.

The present approach can be useful when interpreting CA dataobtained from the analysis of microfloras. Here commonly wider CAintervals are obtained when compared to the analysis of themacrobotanical record. Thus, an improvement of the climatic resolu-tion will help to detect short-term events and to quantify climatevariability.Work is currently in progress to computerise the algorithm.

552 T. Utescher et al. / Earth and Planetary Science Letters 284 (2009) 544–552

Acknowledgements

We thank our colleagues A. Micheels and A. Dreist for support fortheir support in processing of climatological data. We are grateful tomany members of the NECLIME group for fruitful discussions. Also weare indebted to our reviewers who greatly helped to improve themanuscript by their comments and corrections. The funding obtainedfrom the German Research Foundation is gratefully acknowledged.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.epsl.2009.05.021.

References

Ashraf, A.R., Mosbrugger, V., Utescher, T., 1997. Palynological studies in the Neogene ofthe open-pit mines Inden and Bergheim. Courier Forschungsinstitut Senckenberg201, 29–46.

Belz,G.,Mosbrugger,V.,1994. Systematisch-palaäoökologischeundpaläoklimatischeAnalysevon Blattfloren im Mio-/Pliozän der Niederrheinischen Bucht (NW-Deutschland).Palaeontogr. Abt. B 233, 19–156.

Boehme,M., 2003. TheMiocene climatic optimum: evidence fromectothermic vertebratesof Central Europe. Palaeogeogr. Palaeoclimatol. Palaeoecol. 195, 389–401.

Boehme, M., Ilg, A., Winklhofer, M., 2008. Late Miocene “washhouse” climate in Europe.Earth Planet. Sci. Lett. 275, 393–401.

Bozukov, V., Utescher, T., Ivanov, M., 2009. Late Eocene to early Miocene climate andvegetation of Bulgaria. Rev. Palaeobot. Palynol. 153, 360–374.

Bruch, A.A., Utescher, T., Alcalde Olivares, C., Dolakova, N., Mosbrugger, V., 2004. Middleand Late Miocene spatial temperature patterns and gradients in Central Europe —preliminary results based on palaeobotanical climate reconstructions. Courier For-schungsinstitut Senckenberg 249, 15–27.

Bruch, A.A., Utescher, T., Mosbrugger, V., Gabrielyan, I., Ivanov, D.A., 2006. Late Mioceneclimate in the circum-Alpine realm— a quantitative analysis of terrestrial palaeofloras.Palaeogeogr. Palaeoclimatol. Palaeoecol. 238, 270–280.

Bruch, A.A., Uhl, D., Mosbrugger, V., 2007. Miocene Climate in Europe — patterns andevolution. Palaeogeogr. Palaeoclimatol. Palaeoecol. 253, 1–7.

Kottek, M., Grieser, J., Beck, C., Rudolf, B., Rubel, F., 2006. World Map of the Köppen-Geiger climate classification updated. Meteorol. Z. 15, 259–263.

Mai, D.H., 1995. Tertiäre Vegetationsgeschichte Europas. Gustav Fischer, Jena.Mai, D.H., Walther, H., 2000. Die Fundstellen Eozäner Floren des Weißelster-Beckens

und seiner Randgebiete. Altenbg. nat.wiss. Forsch. 13, 1–59.Matthews, E., 1983. Global vegetation and land use: new high-resolution data bases for

climate studies. J. Clim. Appl. Meteorol. 22, 474–487.Micheels, A., Bruch, A.A., Uhl, D., Utescher, T., Mosbrugger, V., 2007. A Late Miocene

climate model simulation with ECHAM4/ML and its quantitative validation withterrestrial proxy data. Palaeogeogr. Palaeoclimatol. Palaeoecol. 253, 251–270.

Mosbrugger, V., 1989. Die Pflanzenwelt des Ober-Oligozäns von Rott. In: v. Koenigswald,W. (Ed.), Fossillagerstätte Rott. Rheinlandia, Siegburg. 82 p.

Mosbrugger, V., Utescher, T., 1997. The coexistence approach — a method for quan-titative reconstructions of Tertiary terrestrial palaeoclimate data using plant fossils.Palaeogeogr. Palaeoclimatol. Palaeoecol. 134, 61–86.

Mosbrugger, V., Utescher, T., Dilcher, D.L., 2005. Cenozoic continental climatic evolutionof Central Europe. Proc. Natl. Acad. Sci. 102 (42), 14964–14969.

New, M., Lister, D., Hulme, M., Makin, I., 2002. A high-resolution data set of surfaceclimate over global land areas. Climate Res. 21, 1–25.

Roth-Nebelsick, A., Utescher, T., Mosbrugger, V., Diester-Haass, L., Walther, H., 2004.Changes in atmospheric CO2 concentrations and climate from the Late Eocene toEarly Miocene: palaeobotanical reconstruction based on fossil floras from Saxony,Germany. Palaeogeogr. Palaeoclimatol. Palaeoecol. 205, 43–67.

Steppuhn, A., Micheels, A., Bruch, A.A., Uhl, D., Utescher, T., Mosbrugger, V., 2007. Thesensitivity of ECHAM4/ML to a double CO2 scenario for the Late Miocene and thecomparison to terrestrial proxy data. Glob. Planet. Change 57, 189–212.

Traiser, C., Klotz, S., Uhl, D., Mosbrugger, V., 2004. Environmental signals from leaves— aphysiognomic analysis of European vegetation. New Phytol. 166, 465–484.

Uhl, D., Mosbrugger, V., Bruch, A.A., Utescher, T., 2003. Reconstructing palaeotempera-tures using leaf floras — case studies for a comparison of leaf margin analysis andthe coexistence approach. Rev. Palaeobot. Palynol. 126, 49–64.

Utescher, T., Mosbrugger, V., Ashraf, A.R., 2000. Terrestrial climate evolution in NorthwestGermany over the last 25 million years. Palaios 15, 430–449.

Williams, J.W., Jackson, S.T., 2007. Novel climates, no-analog communities, and eco-logical surprises. Front. Ecol. Environ. 5, 475–482.

Zachos, J.C., Pegani, M., Stone, L., Thomas, E., Billups, K., 2001. Trends, rhythms, andaberrations in global climates 65 Ma to present. Science 292, 686-293.

Zachos, J.C., Röhl, U., Schellenberg, S.A., Sluijs, A., Hodell, D.A., Kelly, D.C., Thomas, E.,Nicolo, M., Raffi, I., Lourens, L.J., McCarren, H., Kroon, D., 2005. Rapid acidification ofthe ocean during the Paleocene–Eocene thermalmaximum. Science 308,1611–1615.