GKS!S 99/E/54
Climate simulations for the last interglacial period by means of climate models of different complexity
Worn Fachbereich Geowissenschaften der Universitat Hamburg als Dissertation angenommene Arbeit)
Authoress: M. L. Montoya (Meteorological Institute, University Hamburg, and as a guest in the GKSS Research Centre, Inst. of Hydrophysics)
GKSS-Forschungszentrum Geesthacht GmbH ● Geesthacht ● 1999 Address: MariaLuisaMontoya PotsdamInstitutefor ClimateImpactResearch(PIK) PO BOX601203 14412Potsdam,Germany Phone:+49-331-288-2566
Die extemenBerichtederGKSSwerdenkostenlosabgegeben. The delivery of the externalGKSS reportsis free of charge,
Anforderungen/ltequests: GKSS-ForschungszentrumGeesthachtGmbH Bibliothek/Library Postfach 1160 D-2 1494 Geesthacht Germany Fax.:(49)04152/871717
Als Manuscriptvervielfaltigt. FiirdiesenBerichtbehaltenwir unsalleRechtevor.
GKSS-ForschungszentrumGeesthachtGmbH - Telefon (04152 )87-O Max-Planck-Stra13e”D-21502Geesthacht/Postfach1160”D-21494Geesthacht DISCLAIMER
Portions of this document may be illegible in electronic image products. Images are produced from the best available original document. mm!!!!!
GKSS 99/E/54
Climate simulations for the last interglacial period by means of climate models of different complexity (Vom Fachbereich Geowissenschaften der UniversitatHamburg als Dissertation angenomnzeneA,rbeit)
M. L. Montoya
128 pages with 57~gures and 9 tables
Abstract Climatic conditions during the last interglacial(125,000 years before present) are investigated with two climatemodels of differentcomplexity: the atmosphere-oceangeneralcirculationmodel ECHAM-l/LSG and the climate system model of intermediatecomplexity CLIMBER-2. In par- ticularthe role of vegetationat the lastinterglacialmaximum, and its importancefor a consistent simulation of the Mid-Holocene climate, has been investigated (EU project ASPEN: Air-Sea Wave Processes in Climate Change Models). Comparisonof theresultsof thetwo modelsreveals a broad agreementin most large-scalefeatures.Nevertheless,discrepanciesare also detected.Es- sentiaIIy,the models differ in their ocean circulationresponses. Profiting of the fast turnaround timeof CLIMBER-2, a number of sensitivityexperimentshave been performed to try to explain the possible reasons for these differences, and to analyze additionaleffects not included in the previoussimulations.Inparticular,therole of vegetationatthe lastinterglacialmaximumhasbeen investigated.Comparison of the simulated responses against CLIMAP reconstructed SS’TSfor Marine Isotope Stage 5e shows a satisfactory agreementwithin the data uncertainties.
Klimasimulationen fiir die letzte interglaziale Periode anhand von Klimamodellen unterschiedlicher Komplexitat
Zusammenfassung
Die klimatischenBedingungenwahrendder letzten interglazialenPeriode (vor 125000 Jahren) werden anhand zweier Klimamodelle unterschiedlicher Komplexitat untersucht:dem C)zean- Atmosphare gekoppelten allgemeinen Zirkulatonsmodell ECHAM- l/LSG und dem K.lima- systemmodell mittlererKomplexitat CLIMBER-2. Insbesonderewurde die Rolle derVegetation in der letzten interglazialen Periode und ihre Bedeutung fir eine konsistente Simulation des mittelholozaenischenKlimas untersucht(EU-ProjektASPEN: Air-SeaWave Processes in Climate ChangeModels – ,,Klimavariationenin historischenZeiten”). Der Vergleich der Ergebnissebei- der Modelle zeigt eine gute Ubereinstimmung der meisten der grofiskaligen Eigensch.aften, allerdings zeigen sich such Unterschiede. Die Modellergebnisse unterscheidensich im wesent- Iichen in der Antvvortder Ozeanischen Zirkulation. Die kurze Rechenzeit von CLIMBER-2 ermoglicht eine nahereUntersuchungdieserUnterschiedeund zusatzlicherAntriebsfaktoren,die in den bisherigenSimulationennichtbetrachtetwurden.Insbesonderewurde die Rolle der Vege- tationin der letzteninterglazialenPeriode untersucht.Der Vergleich der simuliertenAntworten mit den von CLIMAP rekonstruiertenOzeanoberflachentemperaturenzeigen eine befriedigende Ubereinstimmunginnerhalbder Ungewifiheit der Daten.
Manuscriptreceived/Manuskripteinganginder Redaktion:4. Oktober 1999 Contents
Contents i
Glossary v
List of Figures vii ... List of Tables X111
1 Introduction 1 1.1 Climate models for the study of paleoclimates ...... 1
2 The geological evidence for the LIG 7 2.1 The definition of the LIG in the terrestrial and marine records. Chronology. 7 2.2 Reconstructions on regional to global scales ...... 9 2.2.1 The CLIMAP SST reconstruction for MIS 5e ...... 9 2.2.2 LIGA 1991 ...... 10 2.2.3 The Last Interglacial in northwestern Europe ...... 10 2.3 Evidence on the deep ocean circulation ...... 12 2.4 Theice core record ...... 12 2.5 The Monsoon atthe LAG ...... 14
3 The experimental set-up 15 3.1 The ECHAM-1/LSG climate model ...... 15 3.1.1 The atmospheric component ...... 15 3.1.2 The oceanic component ...... 16 3.1.3 The coupling ...... 17 3.2 The control run. .o. o...... O...... 17 3.3 The external forcing .. o.... .O. o . . ..e . . . .. a.. .OO. 19
i ii CONTENTS
3.3.1 The definition of the seasons ...... 19 3.3.2 Changes in incoming solar radiation ...... 21
4 Results with the ECHAM-1/LSG coupled GCM 23 4.1 Temporal evolution ...... 23 4.2 Large-scale atmospheric patterns ...... 24 4.2.1 Northern summer ...... 27 4.2.2 Northern winter ...... 33 4.2.3 Armual mean ...... 34 4.3 Ocean circulation ...... 37 4.4 Climate variability ...... 42 4.4.1 Atmospheric variability: synoptic scale disturbances . . . . . 44 4.4.20 cean variability ...... 45 4.5 Summary...... 48
5 Results with the CLHV!!BER-2 climate system model 51 5.1 The CLIMBER-2 climate system model ...... 52 5.2 Experimental set-up ...... 53 5.3 Temporal evolution ...... 53 5.4 Comparison of mean simulated fields ...... 54 5.4.1 Near surface temperature ...... 55 5.4.2 Precipitation ...... 55 5.4.3 Atmospheric Circulation ...... 59 5.4.4 Mean ocean circulation ...... 60 5.5 Sensitivity experiments ...... 61 5.5.1 Prescribed freshwater flux run ...... 62 5.5.2 Radiative forcing ...... 63 5.5.3 The role of interactive vegetation ...... 66 5.6 Feedback Analysis ...... 71 5.7 Summary ...... 77
6 Comparison against reconstructed SSTS 81 6.1 CLIMAP SSTS ...... 81 6.2 Model-data intercornparison ...... 82 6.3 Global temperature differences Erompresent ...... 84 6.4 Summary ...... 86
7 Conclusions and Discussion 89 7.1 ECHAM-1/LSG . . < ...... 90 7.2 CLIMBER-2 . . . . < ...... 92 7.3 Model-data intercornparison ...... 94 7.40utlook ...... 95
8 Acknowledgments 97
9 Appendix 99 ... CONTENTS m
References 115
Glossary
AABW: Antarctic Bottom Water. AET: Actual Evapotranspiration. AGCM: Atmospheric General Circulation Model. A: Atmosphere. AO: Atmosphere-Ocean. AOV: Atmosphere-Ocean-Vegetation. BATS: Biosphere-Atmosphere Transfer Scheme. CLIMAP: Climate Long Range Investigation and Mapping Project. CLIMBER: Climate and Biosphere. DJF: December-January-February. EBM: Energy Balance Model. ECHAM: European Centre and Hamburg. ECMWF: European Centre for Medium Range Weather Forecasts. EOF: Empirical Orthogonal Function. GCM: General Circulation Model. GDD: Growing Degree Days. GIN: Greenland-Iceland-Norwegian. GISP2: Greenland Ice Sheet Project. GRIP: Greenland Ice Core Project. IRD: Ice Rafted Debris.
v vi GLOSARIO
JJA: June-JuIy-August. KYR Thousand years ago. K131?: Thousand years Before Present. LGM: Last Glacial Maximum. LIG: Last Interglacial. LIGA GROUP: Working group for the study of the Last Interglacial in the Arctic and sub-Arctic. LS~: Large Scale Geostrophic. MIS: Marine Isotope Stage. MTC O: Mean Temperature of the Coldest Month. NADW: North Atlantic Deep Water. OGCM: Ocean General Circulation Model. PIK: Potsdam Institute for Climate Impact Research. P-E: precipitation minus evaporation. PPMV: parts per million (in volume). SLP: sea Level Pressure. SS S: Sea Surface Salinity. SST: Sea Surface Temperature. THC: Thermohaline circulation. VEC ODE: Vegetation Continuous Description Model. List of Figures
3.1 The ECHAM-1 model orography (m) in T21 resolution and the model continents ...... 16
3.2 Time vs. latitude (top) and celestial longitude vs. latitude (bottom) differences in zonally averaged incoming solar radiation at the top of the atmosphere (125,000 years ago minus present, Win-2)...... 20
3.3 Position of the solstices and equinoxes a) for the control run (the present), and b) the Eemian run (125,000 years ago). Dates of solstices and equinoxes are given in parenthesis. VE: Vernal equinox, AE: Autum-- nal Equinox, SS: Summer solstice, WS: Winter Solstice...... 21 3.4 Difference in zonally averaged mean insolation in JJA (dotted line), DJF (dashed line) and in the annual mean (solid line) (Win-2)...... 22
4.1 Time series of the mean annual a) total sea ice extent (m2), b) total sea ice volume (m3), in the Northern (solid line) and Southern Hemisphere (dashed line), and c) globally averaged near-surface (2 m) temperature (°C), for the control (thin line) and Eemian run (thick line)...... 24 4.2 Time series of the mean annual globally averaged temperature of the ocean at a) 75 m, b) 250 m, c) 1000 m, d) 2000 m, and e) 4000 m depth for the control (dashed line) and Eemian run (solid line) (“C). Note that the vertical scale is the same for all levels...... 25 4.3 Time series of the mean annual globally averaged salinity of the ocean at a) 75 m, b) 250 m, c) 1000 m, d) 2000 m and e) 4000 m depth for the control (dashed line) and Eemian run (solid line) (psu). Note that the vertical scale is the same for all levels...... 25
vii ... Vlll LIST OF FIGURES
4.4 Time series of a) mass transport of the Antarctic Circumpolar Current (ACC) across the Drake passage, b) outflow of North Atlantic Deep Water (NADW) at 30°S, and c) amplitude of the Atlantic overturning circulation, defied as the maximum value of the zonally averaged mass transport strearnfunction of the meridional overturning circulation in the Atlantic Ocean, for the control (dashed line) and the Eemian run (solid line) (Sv)...... 26 4.5 Mean difference (Eemian minus control run) in JJA in a) net surface solar radiation (Isoline spacing (IS) = 10 Win-2); b) total cloud cover (IS = 5%); c) near-surface temperature (IS = I“C), d) SLP (IS = 1 mb), e) 10-m winds (ins-l), f) total precipitation and g) precipitation minus evaporation (isolines at +15, +10, +5, +2, +1, +0.5 mm day–l); h) 200-hPa u-velocity (IS = 2 ins-l); h) 200-hPa streamfunction (IS = 3x 106m2s-l); i) 200-hPa velocity potential (IS = 106m2s-1) Shading indicates the level of local recurrence p. Light shading: p>O.8 or p 4.13 Mean integrated freshwater flux from 90”N southward for the three ocean basins and the global ocean for a) the control run b) the difference Eemianminus control run(Sv)...... 44 4.14 Mean northward heat transport for the three ocean basins and the global ocean for a) the control run b) the difference Eemian minus control run (Pw) ...... 45 4.15 Difference (Eernian minus control run) in the intra-seasonal standard deviation of the band-pass geopotential height at 500-mb in a) northern summer (JJA) a) northern winter (DJF) (IS = 1 m)...... 46 4.16 Standard deviation of the yearly horizontal barotropic streamfunction (top left), the zonally averaged mass transport streamfunction of the meridional circulation in the Atlantic (top right), in the Pacific (bottom left) and in the Indian Ocean (bottom right) (all in 0.05 Sv)...... 47 4.17 First EOF of the yearly anomalies of the horizontal barotropic stream- function (top left), the zonally averaged mass transport streamfunction of the meridional circulation in the Atlantic (top right), Pacific (bottom left) and in the Indian Ocean (bottom right) (0.05 Sv)...... 48 4.18 First principal component of yearly anomalies of a) the horizontal barotropic streamfunction, and b) the zonally averaged mass transport streamfunc- tion of the meridional circulation in the Atlantic Ocean and c) in the Pacific Ocean, as simulated in the Eemian run...... 49 5.1 Time series of the mean annual globally averaged surface air temperature a) in the control run, b) in the Eemian run (0C) ; Arctic sea ice cover c) in the control run, d) in the Eemian run (106 km2); Antarctic sea ice cover e) in the control run, f) in the Eemian run (106 krn2); amplitude of NADW g) in the control run, h) in the Eemian run (Sv)...... 54 5.2 Difference (Eem minus present control run) in mean near-surface tem- perature as simulated by ECHAM-1/LSG smoothed to CLIMBER-2’S spatial resolution (a, c, e), and by CLIMBER-2 (b, d, f), respectively, in JJA (a, b), in DJF (c, d), and in the annual mean (e, f). Isoline spacing: l°C ...... 56 5.3 Mean annual cycle of total a) Arctic and b) Antarctic sea ice cover (poleward of 65”N and 65°S, respectively) for the control (dashed line) and Eemian run (solid line) (106 km2)...... 57 5.4 Difference (Eem minus present control run) in mean precipitation as simulated by ECHAM-1/LSG smoothed to CLIMBER-2’s spatial reso-- lution (a, c, e), and by CLIMBER-2 (b, d, f), respectively, in JJA (a, b), in DJF (c, d), and in the annual mean (e, f). Isolines at +15, +10, +5, +2, +1, +0.5, +O.lmm day-l...... 58 5.5 Mean difference (Eemian minus control) in zonally averaged precipi- tation over land in JJA, in DJF, and in the annual mean (bottom) as simulated by ECHAM-1/LSG (solid line) and CLIMBER-1 (dashed line) (mmday-l) ...... 59 x LIST OF FIGURES 5.6 Mean difference (Eemian minus control) in zonally averaged zonal ve- locity for a) JJA and b) DJF, as simulated by CLIMBER-2 (m s-l). . 60 5.7 Zonally averaged mass transport streamfunction of the meridional over- turning circulation as simulated by CLIMBER-2 in the Atlantic Ocean and Pacific Ocean in the control run (a, b), and for Eem minus the control run with present C02 concentration (c, d) (Sv)...... 61 5.8 Mean difference (Eemian minus control) in northward heat transport a) in the Atlantic Ocean, b) in the Pacific Ocean, c) in the Indian Ocean (bottom) d) by the atmosphere, as simulated by CLIMBER-2 (PW). . 62 5.9 As in Figure 5.7 but for: Eem with prescribed freshwater flux minus the control run with preindustrial C02 concentration (a, b); Eem sim- ulated with the atmosphere-ocean-vegetation version minus the control run with preindustrial C02 concentration (c, d); Eem simulated with the atmosphere-vegetation version (e, f) (Sv)...... 64 5.10 Northward heat transport response simulated by CLIMBER-2 in the Atlantic, Indian and Pacific Oceans for the experiment with prescribed freshwater flux with respect to the control run (left) and with respect to the Eemianrun (right) (PA) ...... 65 5.11 Difference in mean near-surface temperature a) in JJA, b) DJF, and c) the annual mean as simulated by CLIMBER-2 for Eem with prescribed freshwater flux minus present control run. Isoline spacing: l“C. . . . . 66 5.12 Difference in mean near-surface temperature in a) JJA, b) DJF, and c) the annual mean as simulated by CLIMBER-2, for Eem minus prein- dustrial control run. Isoline spacing: l“C...... 67 5.13 Difference in mean precipitation in a) JJA, b) DJF, and c) the annual mean as simulated by CLIMBER-2 for Eem minus present control run. Isolines at +15, +10, +5, +2, +1, +0.5, +0.1 mm day-l...... 68 5.14 Difference in mean near-surface temperature in a) JJA, b) DJF, and c) the annual mean as simulated by CLIMBER-2 for Eem with the atmosphere-ocean-veget ation version minus preindustrial centrol run. Isoline spacing: l°C ...... 69 5.15 Differences between the Eemian climate simulation with the atmosphere- ocean-vegetation model AOV and the preindustrial control run in the fraction of area covered by forest and by desert...... 70 5.16 Zonally averaged trees (a) and desert (b) fraction for the CLIMBER-2 control run (grey), and the Eemian climate simulations with the atmosphere- vegetation model AV (white) and with the atmosphere-ocean-vegetation model AOV (black) (fraction). All simulations run with preindustrial COz concentration ...... 71 5.17 Difference in mean precipitation in a) JJA, b) DJF and c) the annual mean as simulated by CLIMBER-2 for Eem with the atmosphere-ocean- vegetation version minus preindustrial control run. Isolines at +15, +10, +-5, +2, +1, +0.5, +l.lmmday-l...... 72 LIST OF FIGURES xi 5.18 Difference in mean near-surface temperature a) JJA, b) DJF, and c) the annual mean as simulated by CLIMBER-2 for Eem with the atmosphere- vegetation version minus preindustrial control run. Isoline spacing: 1°C. 73 5.19 Difference in mean near-surface temperature in a) JJA, b) DJF, and c) the annual mean as simulated by CLIMBER-2 for Eem with the atmosphere version minus preindustrial control run. Isoline spacing: lot ...... 74 5.20 Difference in mean precipitation in a) JJA, b) DJF, and c) the an- nual mean as simulated by CLIMBER-2 for Eem with the atmosphere- vegetation version minus preindustrial control run. Isolines at +15, +10, +5, +2, +1, +0.5, +O.lmm day-l...... 75 5.21 Difference in mean precipitation in a) JJA, b), DJF, and c) the annual mean as simulated by CLIMBER-2 for Eem with the atmosphere version minus preindustrial control run. Isolines at +15, +10, +5, +2, +1, +0.5, +O.lmm day-l ...... 76 5.22 Zonally averaged temperature differences with respect to the preindus- trial control run) for JJA (upper panel) and DJF (lower panel) for the four runs with CLIMBER-2: Eem with the atmosphere-ocean-vegetation. version minus preindustrial control run (solid line); Eem minus prein- dustrial control run (dotted line); Eem simulated with the atmosphere- vegetation version minus preindustrial control run (dashed line); Eem simulated with the atmosphere only version minus preindustrial control run(dashed-dotted), (°C) ...... 78 6.1 CLIMAP minus model-simulated SST anomalies (difference between Eemian and present summer and winter average SSTS) at CLIMAP core locations. Thin (bold) digits indicate there is (no) significant difference at l-a level between the CLIMAP and simulated SST anomalies (“C). . 84 6.2 a) Zonally averaged mean annual SST anomalies estimated from CLIMAP SST reconstructions, with their errors calculated from the standard er- rors of seasonal estimates (solid line) and simulated SST anomalies by ECHAM-1/LSG only at grid points which contain CLIMAP cores. cor- rected in order to refer them to the average climatic conditions of the last 1500 years (GCM, dotted line), and by CLIMBER-2, in its ocean- atmosphere-vegetation, relative to preindustrial C02 (AOV-CL, dashed- pointed line) (“C). Note that errors in the zonally averaged simulated SSTS have been omitted; b) Latitudinal distribution of the number of cores for which core-top estimates exist...... 85 6.3 Zonally averaged near surface temperature change (Eemian minus mod-- em) as simulated by ECHAM-1/LSG (solid line), by CLIMBER-2 in its atmosphere-ocean-vegetation version, and by (dashed line) a linear EBM(dotted line), in °C ...... 86 xii LIST OF FIGURES 9.1 Mean difference in the control run in JJA in a) net surface solar radiation (Isoline spacing (IS) = 10 Win-’); b) total cloud cover (IS = 5%); c) near-surface temperature (IS = 10C), d) SLP (IS = 1 mb), e) 10-m winds (ins-l), f) total precipitation and g) precipitation minus evaporation (isolines at +15, +10, +5, +2, +1, +0.5 mm day-’); h) 200-hPa u- velocity (IS = 2 ins-l); h) 200-hPa streamfunction (IS = 3x 10Gm2s–1); i) 200-hPa velocity potential (IS = 106m2s–1) Shading indicates the level of local recurrence p, Light shading: p>O.8 or p 3.1 Boundary conditions for the Eemian and control runs...... 19 4.1 Differences (Eemian minus control) in mean simulated sea ice cover (m2) and volume (m3) in JJA, in DJF, and in the annual mean for both[ hemispheres and global. Results indicate averages over the last 300 years of the Eemian and control runs)...... 28 4.2 Difference (Eemian minus control) in mean simulated precipitation andl precipitation minus evaporation for selected areas (land, ocean, global, Northern Hemisphere land, Northern Hemisphere ocean, Northern Hemi- sphere global, Southern Hemisphere land, Southern Hemisphere ocean, Southern Hemisphere global ) in JJA, in DJF, and in the annual mean (mm day-l); * indicates those values which are above the 0.8 (or below the0,2) level of recurrence...... 40 4.3 Total variance explained by each of the first EOFS plotted in Figure 4.17’ (%) ...... 47 4.4 Correlation between the first principal components plotted on Figure 4.18 48 5.1 Boundary conditions for runs with CLIMBER-2 ...... 53 5.2 Annually (JJA) averaged temperature (in ‘C) and precipitation (in mm day-l) over land, and land plus ocean for selected areas (NH: Northern Hemisphere; SH: Southern Hemisphere; N-AFR: North Africa between 10 and 30°N; subscript L indicates averages over land only). lEemian - present ; 2Eemian - preindustrial CLIMBER-2 (C02 = 280 ppmv) . . 57 5.3 Feedbacks and synergisms due to ocean and vegetation involved in an nually averaged temperature (in ‘C) and precipitation (in mm day–l) changes at the Eemian over land plus ocean for selected areas (NH: Northern Hemisphere; SH: Southern Hemisphere; N-AFR: North Africa between 10 and 30°N; L indicates averages over land only)...... 77 ... Xlll xiv LIST OF TABLES 6.1 Difference (Eemian minus control) in mean near-surface temperature for selected areas (global., Northern Hemisphere, O-30”N, 30-90”N, 30-90”N over land only, and for northern summer, as simulated by ECHAM- l/LSG, corrected values to compensate for the different C02 concentra- tions of our Eemian and control runs (2nd row)...... 86 CHAPTE:R 1 Introduction 1.1 Climate models for the study of paleoclimates The potential of human activities to affect climate and other environmental systems, and the fact that such changes could have a substantial impact on human and natural systems has motivated the need for a quantitative evaluation of this impact through the mathematical simulation of the climate system. Efforts in this line initially mainly focused in modeling the different components of the climate system separately. IDue to the complexity of the climate system, a hierarchy of models is necessary to study its full response. The simplest models are conceptual and aim at understanding individual processes, providing insight into their relative import ante. Many processes which are not directly related to the one studied, but which nevertheless might be important for the global climate system response are neglected or parameterized. Among at- mospheric models, the simplest physically oriented models are energy balance models (EBMs), ranging from zero to two-dimensional. Zero-dimensional models predict the global temperature of the Earth from the net in and out-coming global radiative fluxes. Starting from such a model, modifications can be gradually added to approach more realistic conditions. For example, the atmosphere can be vertically divided into differ- ent levels to obtain a satisfactory parameterization of the vertical long-wave radiation; radiative-convective models accounting for convection, which is responsible for most of the upward heat transport from the lower to the upper levels of the atmosphere, can be used. In addition, the number of dimensions can be increased up to two. In such models, for instante, the horizontal heat transport by eddies and by the large-scale circulation is parameterized by diffusive heat transport. At the other end, the most complex models are general circulation models (GCMS). Their development was very much impulsed by the advent of computers. These attempt 2 CHAPTER 1. INTRODUCTION to give a realistic simulation of the observed phenomena by providing the most compre- hensive mathematical description of the system. Thus, atmospheric GCMS (AGCMS) are capable of resolving the daily weather fluctuations which are believed to play a cru- cial role in the climate response. Modeling of the ocean general circulation has much in common with its atmospheric counterpart, although there are physical and techni- cal features unique to the ocean. Among ocean models, in increasing complexity, box models, zonally averaged models of the meridional circulation on a vertical-meridional plane, and finally, ocean GCMS (OGCMS) have been employed. Only in the last decade have experiments been undertaken towards a comprehen- sive coupling of the different components of the climate system. AGCMS were initially coupled to swamp and to mixed layer ocean models, and finally to OGCMS. ICou- pled GCMS are the state-of-the-art tools to simulate climate under given boundary conditions and to study the climatic response to changes in these and in the external forcing, and constitute a valuable tool for the study of climate variability, in particular the large-scale, low-frequency ocean variability, otherwise not feasible due to the lack of observational data. All climate models include a number of adjustments to reproduce the present ob- servations. Hence, an accurate simulation of the present climate does not guarantee a correct performance of the model under different boundary conditions. The sensitivity of climate models can be calibrated against the observed climatic record. This requires the knowledge both of the climate forcing and of the climatological data corresponding to that period. One means ;s provided by the use of the instrumental record (the last N 100 years). This has the advantage of using direct observations of climatic parameters with increasingly dense spatial coverage. However, the determination of the climate forcing is less satisfactory: in addition to increasing greenhouse gas con- centrations, stratospheric aercmolsof volcanic origin, variations in solar radiation and anthropogenic sulfate aerosols also have an imprint on climate and their occurrence in the past and their contribution are difficult to quantify. Natural variabilityy due to internal processes may as well produce climatic variations of magnitude comparable to the variations forced by the (external) former processes. An alternative approach ccmsists in resorting to the simulation of past climates and the comparison of the results to the evidence from the geological record. If sufficiently realistic, this approach additionally provides explanations of the reconstructed climatic features in term of specific physical processes (changes in atmospheric or oceanic circu- lation, in precipitation or evaporation, etc.). During the last 20 years, most efforts con- cerning paleoclimat e simulaticms have been devoted to AGCMS. In this case, the state of the other components of the climate system must be specified: i.e., the land-ocean distribution, ice sheets, vegetation (mainly through albedo and surface roughness), sea surface temperatures (SSTS), ;seaice extent, and the atmospheric composition (mainly in terms of an equivalent CC~Zconcentration). Starting from the initial conditions, the numerical simulation provides the atmospheric circulation in equilibrium with the prescribed boundary conditions for a specific time slice of the past. In a different ap- proach, numerical simulations with simpler models have also been used to simulate the transient changes that take place over decades, centuries or longer time scales. 1.1.CLIMATE MODELS FOR THE STUDY OF PALEOCLIMATES 3 Much effort has been devoted to the climate reconstruction of a few past time periods, mainly during the Late Quaternary (0-20 kBPl), a period for which the re- constructed paleodata are abundant and accurately dated (Crowley and North, 1991): the early and mid Holocene (9 and 6 kBP, respectively), which were both periods with enhanced northern summer insolation, (LGM, 21 kBP), when continental ice-sheets covered most of mid-high northern latitudes. Paleoclimate simulations have mainly focused on such periods, for which a complete set of boundary conditions is known. In addition, sensitivity studies in which the response of the climate system to the change in individual boundary conditions is studied, while fixing the others to prescribed val- ues, help to estimate the sensitivity of climate to individual changes in the forcing. Paleoclimate simulations and sensitivity studies have shed light on the orbital control and the influence of surface boundary conditions and C02 on climate (Kutzbach and Guetter, 1986; Prell and Kutzbach, 1987; Kutzbach and Liu, 1997), in particular at the mid-early Holocene (Kutzbach and Gallimore, 1988; Mitchell et al., 1988; Hewitt and Mitchell, 1996; Hall and Valdes, 1997; Hewitt and Mitchell, 1998; Kutzbach and Liu, 1997) and at the LGM (Ganopolski et al., 1997; Dong and Valdes, 1998; Bush and Philander, 1998), and on the role of vegetation feedbacks in past climate changes (Crowley and Baurn, 1997; Kutzbach et al., 1996; Texier et al., 1997; Ganopolski et al., 1998; Kubatzki and Claussen, 1998). In this line, we have used two models of different complexity to simulate climatic conditions at 125 kBP, the maximum of the last interglacial (LIG) period (the Eemian, N 120 – 130 kBP): a coupled ocean-atmosphere GCM and a global climate system model of intermediate complexity. The aim of this study is to present a best guess of climate at this time period as simulated by a state-of-the-art coupled models. Evidence for warmth from land, ice and, in a smaller number, ocean records from mid-high northern latitudes have led to the widespread assumption that the LIG was possibly the warmest time period of the last 200,000 years in terms of globally averaged temperature, and to regard this and other interglacial periods from the Late Quaternary (6 kBP, 9 kBP) as analogs for a future potentially warmer world due to rising greenhouse gas levels in an attempt to infer future regional climate patterns (Kellogg, 1977; Fung et al., 1988; Budyko and Izrael, 1991; Zubakov and Borzenkova, 1990; Shabalova and Konnen, 1995). However, based on the different nature of past and future climate forcings, such approaches have sometimes been criticized (Crowley, 1990). Differences with respect to present in the seasonal cycle of insolation at the LIG were greater than at other warm time periods of the Late Quaternary, thereby improving the signal-to-noise ratio both in the paleoclimat e reconstructions and in the simuli~tions. In addition, paleo-environmental records are not affected by human activity, as would be the case during the Holocene. At the same time, climatic conditions were still close to present ones. Hence, it is in principle reasonable to ask whether the GCM employed, with its current adjustments, parameterizations and correction terms is appropriate or not to simulate the corresponding climate response. In particular, as will be discussed in Chapter 4, the coupling of the atmosphere and ocean components involves the use of a flux correction term. Such an approach is appropriatee as long as the working point 1Thousandyearsbeforepresent 4 CHAPTER 1. INTRODUCTION does not radically differ from the climate for which the flux correction was built. For instance, the existence of a flux correction terms prevents us using such a model to simulate, for example, the climate of the LGM. Few paleoclimate simulations have so far been carried out with AGCMS coupled to oceanic GCMS (O GCMS), and the first of these experiments constitutes the first attempt to use a coupled ocea,n-atmosphere GCM to simulate the climate of the LIG. Previous climate simulations of this period have been carried out with energy balance models (EBMs) (Crowley and Kim, 1994) and AGCMS with ocean boundary conditions prescribed to modern values (Royer et al., 1984; Prell and Kutzbach, 1987; Kutzbach et al., 1991; de Noblet et al., 1996) or coupled to static mixed-layer ocean models (Harrison et al., 1995). However, such models are not able to simulate, e.g., the heat transport by the ocean and the heat storage by the deep ocean, processes all which have a strong influence on the equilibrium climate and on the climate response to external forcing, and it is generally recognized that reliable estimates of the climate response to greenhouse warming can be achieved only by means of fully coupled ocean-atmosphere GCMS (Houghton et al., 1990). The comparison of our results to the previous ones and the inspection of the differences between them helps to shed light on the role of the ocean in the climate response to the perturbed radiative forcing at this time period, ignored in the previous studies. To avoid including systematic model errors, the focus will be on the anomalies of the simulated LIG climate with respect to a modern climate, as simulated in a control run with constant external forcing. Both the mean anomalous state of the main atmospheric and oceanic large-scale fields and their climate variability at different time-scales will be analyzed. Concerning the set of experiments performed with the second model, the term in- termediate complexity refers to the fact that this model fills the gap between GCMS and simplest models such as EBMs. As shown below, the large-scale simulated responses of the two models agree reasonably well. However, a certain number of disagreements is also found. The fast turnaround time of this simpler model permits carrying out further sensitivity studies which help to shed light on the possible reasons for such discrepan- cies. Moreover, in recent times the import ante of changes in vegetation in the climate response has increasingly been acknowledged. Due to computational limitations, this component was ignored in the first simulation with the coupled ocean-atmosphere GCM, but has been considered in our simulations with this intermediate-complexity model. The validation of climate models in paleo-simulations is hindered by several prob- lems: First, few quantitative reconstruction exist yet. Second, a large gap exists be- tween the typical spatial scales of models and reconstructions. With their typical spatial resolution, GCMS simulate reliably the main large-scale climatic features but their skill at finer (regional to local) scales is much smaller, while paleo-reconstructions generally give information on the local or, at most, the regional scale. One of the most complete quantitative reconstructions for the LIG is that of SSTS at Marine Isotope Stage (MIS) 5e carried out by the CLIMAP (Climate: Long-Range Investigation and Mapping) Project Members (1984). Since the SST field is generally (except at frontal zones) quite homogeneous, thle regional climate will not differ significantly from the 1.1.CLIMATE MODELS FOR THE STUDY OF PALEOCLIMATES 5 large-scale prediction. Hence, this reconstruction has the advantage that it can be directly compared to the GCM output. The contents of this thesis are organized as follows: Chapter 2 briefly reviews the current view of climate at the LIG based on paleoclimatic evidence. Chapter 3 describes the experimental set-up (the coupled model used and the prescribed external forcing). Chapter 4 describes the main results of the climate simulation for 125 kBP with the coupled GCM: the temporal evolution of the sea ice, atmosphere and ocean systems, the mean differences with respect to the control run in the thermal response, the large-scale atmospheric circulation, the hydrological cycle and the atmospheric variabilityy, the mean annual differences in ocean circulation and the main modes of ocean circulation variations. In Chapter 5 the results of an additional climate simulation for 125 KBP with an intermediate complexity model. are compared to the former results with the coupled GCM. Results of several sensitivityy experiments analyzing the relative influence of changes in orbital parameters and in C02 concentration, which were performed in an attempt to explain the disagreements between these two simulations, are presented, and the response of interactive vegetation to the altered boundary conditions is analyzed. In Chapter 6, the thermal responses of the two former models are compared to each other an and to reconstructed SSTS (1984) for MIS 5e. Finally, chapter 7 summarizes and discusses the main results. CHAPTE:R 2 The geological evidence for the LIG 2.1 The definition of the LIG in the terrestrial and marine records. Chronology. The terrestrial record of the LIG (the Eemian on land, see below) is very well doc- umented in Europe thanks to palynological analysis of hundreds of records. Many of these are continuous throughout the Late Pleistocene, permitting the development of regional palynostratigraphical schemes which can easily be correlated with each other. The stratigraphical framework of the LIG is characterized by the development of thermophilous associations reflecting the estabolishmentof warm, humid conditions (Behre, 1989). The upper Quaternary sequences in northwestern Europe and in the Mediterranean region were characterized already in the sixties thanks to the discovery of pollen sequences from the Netherlands and Macedonia, respectively. The former comprise, among others, the type locality for the Eemian interglacial (see van der Harnmen (1971) and references therein). The latter record, obtained from the Tenaghi Philippon (Macedonia) (Wijmstra, 1969), is continuous through the last 500 kyr and shows a warm period, called Pangaion interglacial, which is characterized by the dom- inance of forest and corresponds to the northwestern European Eemian. In the original 6180 stratigraphy obtained from the planktonic foraminifera, deep- sea sedimentary record (Emiliani, 1955), MIS 5 was initially broadly defined as an interglacial and subsequently subdivided into five substages, 5a to 5e (Shackleton, 1969). Substages 5a-d are associated with an early glacial phase. Substages 5a, 5c, and 5e correspond to local 6180 minima reflecting episodes of reduced continental ice volume following high northern summer insolation; MIS 5e corresponds to the substage of minimum ice-sheet volume. On the basis of the striking similarities found between climatic conditions regis- 7 8 CHAPTER 2. THE GEOLOGICAL EVIDENCE FOR THE LIG tered in the deep-sea and in the continental records, Shackleton (1969) was the first to suggest a correspondence between MIS 5e and the Eemian stage on northwestern Eu- rope. This correspondence wa~$established several years later by Turon (1984) through correlation of the La Grande Pile peat bog pollen record (Woillard, 1978), continuous through the last 140 kyr, with that obtained through marine palynology of a deep-sea sediment record off the coast of the Iberian Peninsula. The correspondence of these with the Ipswichian (Britsin), Sangamon (North America) amd Mikulino (Russia) was subsequently demonstrated (Eehre, 1989). The development of an absolute chronology for the Pleistocene from the deep-sea sediment record has been precluded above the range of 14C (w45 kyr), since other radiometric techniques, such as uranium series dating, are beset with questionable as- sumptions. Instead, efforts in this direction have aimed at tuning the 6180 record to the Earth’s orbitally driven insolation variations, whose dependence in time is well known, leading to the standard SPECMAP J180 chronology (Imbrie et al., 1984; Mar- tinson et al., 1987), according to which the last interglacial spanned ~127-116 kyr. Support for this chronology hi~sbeen provided by radiometrically derived ages for the last high-level stands (events 51a,5Cand 5e) centered around 82, 105 and 125 kBP. The technique employed consisted in a particle-counting in z30Th/zsAUfrom coral terraces. With a typical error of 10 kyr, it has been replaced in the last years by more precise mass spectrometric techniques to measure uranium and thorium isotopes, which has reduced the typical error to about 1.5 kyr. Application to fossil corals (Edwards et al., 1987; Chen et al., 1991) has yielded for the LIG (corresponding to an elevation of the sea level 2-8 m higher than present) an age of 120-130 kBP. A chronology for the last 160 kyr based on an ice-flow model was derived from the Vostok ice core record in ,Antarctica (Jouzel et al., 1987). This chronology agreed satisfactorily with that of the SPECMAP stacked 6180 record, but the duration of the LIG disagreed considerably: ~~22kyr, spanning 139-117 kyr, versus Nll kyr (127-116 kyr) from the SPECMAP record. The extension of the Vostok ice core record from 160 kyr to about 220 kyr (Jou:zel et al., 1993) led to the development of a new chronol- ogy which much closely agreed.with the SPECMAP 6180 chronology, but differences of about 6 kyr still existed at the peaks of the LIG. However, given the typical error in the ice-flow model used to establish the Vostok chronology is ca. 1570, the aforementioned disagreement should possibly not be surprising. In addition, a new chronology obtained by uranium series dating on carbonate 6180 of a calcite vein in Devils Hole (Ludwig et al., 1992; Winograd et al., 1992) showed a remarkable agreement with SPE!CMAP, except at MIS 5e, where the Devils Hole chronology shows a slightly earlier onset (140 kBP) and a longer duration (20 kyr). Such a discrepancy is critical due to the small uncertainties in the dating techniques. Possible explanations are that changes in the accumulation rate could have affected the deposition rate of 230Th generated by the decay of uranium in the Devils Hole Fissure (Edwards and Gallup, 1993; Shackle- ton, 1993), or that the Devils Hole chronology is correct but rather reflects regional variations in climate (Imbrie, 1993; Crowley, 1994). More recent application of mass spectrometric techniques to measure uranium and thorium isotopes in coral fossils (Gallup et al., 1994) and in marine sediments from 2.2.RECONSTRUCTIONS ON REGIONAL TO GLOBAL SCALES 9 the Bahamas, where algae and inorganic processes produce aragonite with uranium concentrations far in excess of those found in foraminifera and which equal or exceed those found in fossil corals (Slowey et al., 1996) have yielded for the LIG an age of 120-130 kBP. 2.2 Reconstructions on regional to global scales Inferences about ice-volume at the LIG versus today are provided mainly by indirect evidence. Most direct evidence has either been eroded or lies buried under younger sediments or ice sheets. Deep-sea sediment cores generally do not have a high enough resolution and those which do anyway provide ambiguous evidence. A wide consensus exists that during the LIG less ice existed than today based mainly on the evidence of higher sea level (up to 6 m) than present from fossil coral reef terraces dated by uranium series (Mesolella et al., 1969; Bloom et al., 1974; Ku et al., 1974). 6 m of global sea-level equivalent is roughly the volume of ice contained today in each the Greenland and the West Antarctic ice sheet; hence either complete disappearance of either or partial disappearance of both ice bodies has been suggested. The main attempts to compile and synthesize evidence to obtain a view of the climate at the LIG at spatial scales larger than local have been carried out by the following groups: i) the CLIMAP (Climate: Long-Range Investigation and Mapping Project) Group (CLIMAP Project Members, 1984); ii) the LIGA Group (Working group for the study of the last interglacial in the Arctic and sub-Arctic) (LIGA Mem- bers, 1991); iii) recent works by Zagwijn (1996) and Aalbersberg and Litt (1998). 2.2.1 The CLIMAP SST reconstruction for MIS 5e The CLIMAP Project Members (CLIMAP Project Members, 1984) reconstructed sum- mer and winter SSTS at MIS 5e for all basins by applying previously derived trans- fer functions to census counts of one of the three biotic components of the sediment (foraminifera, coccoliths, and radiolaria). Results showed that SSTS at the LIG were in generaI very similar to today’s: about 60% of the SST estimates for the LIG agreed with observed (present) values within the typical standard error of 1.5”C, and many of the disagreements disappeared if core-top estimates (that is, the result of applying the transfer functions to the population of the core-top, which are calibrated against observed SSTS) were used instead of present observations. Areas showing divergence with modern temperatures included the Southern Indian Ocean, where results sug- gested slightly lower SSTS at MIS 5e compared to present, the mid-high latitudes of the North Atlantic, where temperatures were higher at MIS 5e than present, especially in winter, and the western equatorial Atlantic, the Caribbean Sea and the Gulf of Mexico where temperatures were 2-3°C cooler than present. 10 CHAPTER 2. THE GEOLOGICAL EVIDENCE FOR THE LIG 2.2.2 LIGA 1991 Inspired by the paleo-analog idea, the LIGA Members compiled preexisting data for MIS 5 from 95 high northern latitude sites originating in the the deep-sea record, ice cores, and terrestrial records to study the environmental condkions and climatic changes associated to an increase of temperature at high northern latitudes. At the LIG, climatic conditions (qual.itatively) considerably warmer than present were found throughout the circumpolar and circum-boreal terrestrial regions. In central and west- ern North America, pollen records suggest conditions similar or slightly warmer, while macrofossil remains indicate much warmer conditions. Pollen records from Scandinavia and western Europe suggested mean annual temperatures 2-3°C higher than present, while eastern Europe winter temperatures might have been up to 5°C higher than present. Conditions were moist er over northwest North America and northern Eura- sia. In southeastern Camada, both pollen data and macrofossil assemblages indicate temperatures around 4°C warmer than present. Paleoecological indicators show a shift of subpolar and borezd bioclimatic zones by up to several thousand km north of their present limits. The GIN (Greenland, Iceland, Norwegian) Seas were occupied by sub- polar water masses (Kellog, 1980; CLIMAP Project Members, 1984). Summer SSTS might have been 8°C higher than present in the Labrador sea (de Vernal et al., 1991), suggesting less pronounced North Atlantic west-east SST gradients and very efficient latitudinal and eastward transport of warm, moist air masses through the circumpolar regions. The conclusion was that high northern latitudes experienced warmer condi- tions than at present during the LIG maximum, and that such conditions were not re-established, even during the early-mid Holocene. 2.2.3 The Last Interglacial in northwestern Europe Several long pollen sequences in Europe have been reconstructed covering the last in- terglacial/glacial cycle: La Grande Pile (Woillard, 1978; de Beaulieu and Reille, 1992) (les Vosges, France), Padul in southern Spain (Pens and Reille, 1988), Lac du Bouchet and the Ribains Maars in Vellay Plateau from the French Massif Central (Reille and de BeauIieu, 1988; de Beaulieu and Reille, 1992b), Les Echets near Lyon (de Beaulieu and Reille, 1984; de Beaulieu and Reille, 1984b), Monttichio (Watts, 1985) and Vane di Castiglione (Follieri and %dori, 1988) in Italy, and Tenaghi Philippon (Wijmstra, 1969) and Ioaninna (Tzedakis, 1993) in Macedonia and Greece, respectively. All se- quences from northwestern Europe present the same main climatic variations, gen- erally expressed in terms of the arboreal pollen sum, and similar pollen zones have been recorded for nearby regions (e.g. eastern Europe). Essentially, they reveal the presence of a succession of typical overlapping acmae of the following trees: 13etzda, Pinus, Ulmus, Quercus, Corylus, TaxUs, Carpinus, Picea, and Pinus. AU show three warm periods correlated with the warm isotopic stages 5a, 5Cand 5e (Mangerud, 1989). Therefore, pollen diagrams for the Eemian are readily correlated, even more than those for the Holocene. Only for mcmedistant are= such as Norway (Mangerud et al., 1981) or the Mediterranean regions (Follieri and Sadori, 1988) do these similarities not hold. Quantitative analysis of pollen records has been limited. Guiot et al. (1989) re- 2.2.RECONSTRUCTIONS ON REGIONAL TO GLOBAL SCALES 11 constructed mean annual temperature and the annual sum of precipit ation for the last 140 kyr at La Grande Pile (Woillard, 1978) and Les Echets (de Beaulieu and Reille, 1984; de Beaulieu and Reille, 1984b). The Eemian appeared as a warm and humid episode with higher than present mean annual temperature (1-3”C) and lower than present mean annual precipitation (by *200 mm day-l) at both sites. The first ten- tative monthly reconstructions for these data were also provided by Guiot (1990). His January and July reconstructions showed the second part of the Eemian (125-115 kBP) was characterized by an increase of January temperatures by 2-3°C and an decrease of July temperatures about 2°C coupled with a rise of precipitation, features of abmore oceanic climate, consistent also with Guiot et al. (1993) reconstructed total evapotran- spiration (AET) and summer and mean annual temperatures - these constrained by coleoptera fossils - from the pollen data of La Grande Pile. A cooling event registered around 125 kBP was registered which still needs correlation with other marine a,ndice core data. The climatic history of central and northwestern Europe during the LIG has been analyzed on a regional scale in recent works by Zagwijn (1996) and by Aalbersberg and Litt (1998). In the former study, data from 31 preexisting pollen diagrams plus botani- cal macrofossil data were used to estimate mean summer and winter temperatures from central and northwestern Europe. In the latter, temperatures were reconstructed at 106 sites cent aining abundant pre-published and non-published paleobotanical, periglacial and coleopteran data in northwestern Europe. Biostratigraphical zones were used in both studies as time slices. Results of both studies broadly agree. Summer tennpera- tures peaked at the first phase (the thermal maximum of the LIG, in which the north- western European climate reached its full temperate character) and dropped during the subsequent phase. Minimum mean July temperature in this time slice shows high values up to 20°C in Poland and southeastern Germany, up to 18°C in northwestern Germany, the Netherlands and southeastern England and up to 16°C in Denmark and Finland, suggesting a southeast to northwest temperature gradient. Many coleoptera species which today are found in central and southern Europe were found as far north as the British Isles, suggesting a warmer than present climate in the British Isles at the LIG, with winter temperatures in this region similar to present values in central Europe and summer temperatures in southern England above 20”C. Climate had a sub-continental character in central Germany and Poland while in the west it was un- der the moderating influence of the nearby ocean; yearly precipitation was above 600 mm for the whole transect. Winter temperatures reached their maximum value during the second phase. Hence, climate was milder, with a more oceanic character. West- east and meridional trends of isotherms were similar to the present ones. A strong oceanic influence was present, even in Southern Poland and Germany, related to the higher sea level, and possible connections between the North and the Baltic seas which would allow warm ocean water to reach further east than in the present. Precipi.tation was over 600 mm in all northwestern Europe. Temperatures dropped further in the final phase, first winter and then summer temperatures, marking the deterioration of climate towards the subsequent glacial period. 12 CHAPTER 2. THE GEOLOGICAL EVIDENCE FOR THE LIG 2.3 Evidence on the deep ocean circulation &3C reconstructions of TC02 from benthic foraminifera in the deep ocean during MIS 5e (Duplessy et al., 1984; Duplessy and Shacldeton, 1985) showing a global decrease of J13C of wO.15 per mil with respect to present values were initially attributed to a 7% reduction of the continental biomass, explained in terms of tropical rather than temperate regions’ vegetation. Since highest J13C values were found in the Norwegian Sea and because 613C values dlecreasein the Indian and Pacific Oceans according to a simiIar pattern as in the present, a similar deep-ocean circulation pattern as today was suggested. Nevertheless, one striking difference with respect to present values were the low J13C values found in the eastern Atlantic Ocean at MIS 5e, and hence the stronger east-west #3C gradients. Since no evidence was found for changes in productivity throughout the Atlantic Ocean, this implied that the influence of AABW (Antarctic Bottom Water) was much greater than that of NADW (North Atlantic Deep Water) at MIS 5e. This could be explained in terms either of stronger AAB W or of weaker NADW. Evidence for NADW production close to present values at MIS 5e has also been provided by records from the Bahamas showing &3C levels at MIS were close to 1 per mil (Keigwin et al., 1994), Nevertheless, higher resolution records from the North Atlantic have indicated some (subtle) NADW variability at MIS 5e (see below). 2.4 The ice core record Climatic evidence for the LIG from the ice core record is limited to the three cores which penetrate thk time period: the Vostok ice core in Antarctica (Jouzel et al., 1987; Jouzel et al., 1993), and the GRIP (Greenland Ice Core Project Members, 1993) and GISP2 (Greenland Ice sheet Project) (Grootes et al., 1993; Taylor et al., 1993) ice cores in Greenland. The Vest ok ice core provides a continuous record throughout the last climatic cycle (160 kyr) (Jouzel et al., 1987) and has more recently been extended to span the last 220 kyr (Jouzel et al., 1993). The local surface air temperature was estimated from the JD content of the proper ice, and the atmospheric concentration of C02 was determined by analyzing the air bubbles trapped in the ice. The LIG appears in this record to contain a pulse of W5 kyr with temperatures up to 2°C higher than present. In 1992 a second ice core reaching ca. 250 kBP was drilled at Summit (Greenland), the GFUP ice core (Dansgaard et al., 1993; GRIP Members, 1993). A close correlation was found between low-frequency features in the GRIP and the Vostok, SPECMAP and Devil’s Hole 6180 records. However, striking differences appeared at higher frequencies (<5kyr), especially at the LIG; this is related to the question of evidence for climate variability at the LIG. During several years, one of the most intriguing questions concerning the paleoclimatic evidence for the LIG has been that regarding the the question of climatic variability during this time period. The GRIP core (Dansgaard et al., 1993; GRIP Members, 1993) revealed a highly structured pattern of variation on time-scales from several thousand years down to less than 100 years. At MIS 5e, three warm substages (5el, 5e3 and 5e5) appear interrupted by several abrupt cooling events with temperature changes up to 14°C leading in only 1-2 2.4.THE ICE CORE RECORD 13 decades to conditions similar to those of mid-glacial times. Although it was lknown that high-amplitude, high-frequency variability occurred during glacial stages, such oscillations in a warm period appeared to be new. In the light of this new evidence, the past 8 kyr appeared as an exceptionally stable time period of the last 250 kyr, posing the question of why similar oscillations are not found in the present climate. Comparison against the record of a second ice core located 28 km away, the G1SP2 ice core (Grootes et al., 1993; Taylor et al., 1993), showed an excellent agreement throughout the first 2700 m. However, the deepest 10% of the cores, spanning the Eemian interglacial and the previous deglaciation, differed significantly, and the high- frequency, high-amplitude variability suggested by the GRIP ice core was not recorded in the GISP2 record. It was suggested then that either one or both records might have been disturbed by ice-flow deformation (Grootes et al., 1993; Taylor et al., 1993). In the first case, the abrupt events seen in the GRIP record would not represent a real climate signal. This discrepancy motivated the study of Eemian sections of northwestern European pollen records and of the MIS 5e sections of the deep-sea sediment record: the excellent correlation found between surface North Atlantic records and the JlsO record of ice cores during the last deglaciation and MIS 3 (Lehman and Keigwin, 1992; Bond et al., 1993) suggested that if the oscillations registered during MIS 5e in the GRIP core were real, these should be reflected by the marine sedimentary record. Support for the idea that rapid climatic oscillations did occur during the Eemian has been provided by several land and ocean studies: Records of the Lac du Bouchet (Massif Central, France) over the past 140 kyr (Thouveny, 1994) show five oscillations supposedly related to substages 5e5 to 5el described in the GRIP record. A reconstruction for Bispingen (Field et al., 1994) showed significant variability at the Eemian including one major oscillation suggesting cooling to mid-glacial levels, consistent with evidence from the GRIP ice core. Reconstructed summer SSTS in the Norwegian Sea show the increasing trend of SSTS since the early MIS 5e was interrupted by an abrupt cooling event of ca. 3.5°C by mid-MIS 5e (C!ortijo, 1994). Proxy temperature records of several cores in the GIN seas during MIS 5e (Fronval and Jansen, 1996) display a number of in phase high-frequency, high-amplitude oscillations with three main cooling events versws only one in the Holocene. A very high-resolution deep-sea sediment record from the North Atlantic (McManus et al., 1994) showed considerable sub-orbital (millenial) variability back to ca. 1110kyr, closely matching that in the GRIP ice record, and probably reflecting a repositioning of the polar front (McManus et al., 1994). A record from the tropical Atlantic (Bahamas) (Keigwin et al., 1994) also showed a good correlation with the GRIP ice core record back to 110 kyr, indicating significant millenial scale variability probably due to rapid NADW production changes, thus suggesting a link between NADW production and air- temperature over Greenland caused by changes in northward heat transport. But no evidence for comparable variability was found in either record at MIS 5e. Benth.ic and surface deep-sea sediment records of the subpolar North Atlantic (Oppo et al., 1997) and the Bermudas (Adkins et al., 1997) indicate that climatic shifts linked to deep- water reorganizations can occur in less than a few hundred years, showing that MIS 5e 14 CHAPTER 2. THE GEOLOGICAL EVIDENCE FOR THE LIG was preceded and ended through abrupt reduction of NADW formation. Within MIS 5e, the records do show evidence for some variability of the THC (suggesting NADW was highest at 125 kBP and lc~werat the beginning and the end of MIS 5e), but much smaller than during the preceding and following glacial periods and too small to have caused large imprints in surface air temperature over Greenland (Oppo et al., 1997). In 1995, the GISP2-GRIP .JointWorkshop (Wolfeboro, New Hampshire, USA) con- cluded that both the GRIP and GISP2 ice cores had suffered stratigraphic disturbances in ice older than 110 kyr (Peel, 1995). Nevertheless, the GRIP record has highlighted how little is known about climate stability at warm periods and has inspired detailed climate studies on land and in marine records as the aforementioned ones. 2.5 The Monsoon at the LIG A variety of data show evidence for a more humid climate in the African and Asian monsoon sectors between 141 and 70 kBP, with a wet peak at 130-125 kBP. In the Sahara, several wet episodes during the LIG are documented by U/Th series dated records of paleo-lakes in central Libya, northern Mali and the Egyptian desert (Petit- Maire et al., 1980; Petit-Maire, 1982). In northern Mali, thick lacustrine lime-stones with fresh-water mollusks have been observed and dated around 125 kBP (Petit-Maire, 1993). In the Atlantic Ocean, the variable occurrence of wind blown freshwater diatoms in deep sea sediments has been attributed to changes in the African monsoon (Pokras and Mix, 1985). Low numbers of freshwater diatoms indicating humid conditions, high level lakes and low deflation of diatoms are obtained from 150-125 kBP. In the Arabian Peninsula, humid conditions are dated 142 to 108 kBP: in Oman, alteration and pedogenesis were active during MIS 5e (Salanville, 1992). KIein et al. (Klein et al., 1990) describe evidence for monsoonal rains up to the Sinai. Along the coast of Sudan and the Red Sea, stage 5e coral reefs were interbedded with sediments brought by river runoff. The three warmer peaks on MIS 5e are documented by organic- rich layers (sapropels) off the Nile delta, showing evidence of enhanced tropical rains over Ethiopia and East Africa (Rossignol-Strick et al., 1982; Petit-Maire, 1993). In the Arabian Sea, strong southwesterly winds cause coastal upwelling, characterized by low SSTS and high nutrient flux and productivity during summer, favoring both high assemblages of Globigerina bulloides, a planktonic foraminifera typical of subpolar wa- ters, and high foraminifera shell accumulation rates in the underlying sediments. High values of these indicators suggest evidence for strong south westerly winds broadly cor- responding with 125 kBP (Anderson and Prell, 1993; Clemens et al., 1991; Emeis et al., 1991; Prell, 1984% Prell, 1984,b; Prell and Kutzbach, 1987). In addition, estimates of paleo-salinity near the Maldivas leading to lower than present values at interglacial times at 115-125 kBP, indicate that higher than present freshwater flux into this region suggesting the southwest monsoon was strengthened during this time interval (Rostek et al., 1993). In China, loess-paleosol sequences provide evidence for wetter conditions (An et al., 1991), in particular at MIS 5e. CHAPTEIR 3 The experimental set-up 3.1 The ECHAM-1/LSG climate model The climate model used in this experiment (hereafter, the Eemian run) is the ECHAM-1 T21/LSG coupled ocean-atmosphere GCM (Cubasch et al., 1992). This model has been used in one of the few millennia coupled runs performed at the moment (von Storch et al., 1997), which constitutes our reference or control run, and in several climate change experiments (Cubasch et al., 1992; Cubasch et al., 1994; Cubasch et al., 1995; Bakan et al., 1991). 3.1.1 The atmospheric component The atmospheric component (ECHAM-1) is a second generation version of the ECHAM model (Roeckner et al., 1992), which evolved from the numerical weather predic- tion model developed at the European Centre for Medium Range Weather Forecasts (ECMWF), and was adapted for climate study purposes at the Max-Planck-Institut for Meteorology in Hamburg. It is a spectral model (i.e., horizontal equations are solved using the spectral method) with a resolution given by a triangular cut-off at zonal wave number 21. At each time step, variables are transformed into a Gaussian grid of about 5.6° x 5.6° (i.e., 64x 32 grid points) to calculate the non-linear advection terms and physical parameterized processes (Figure 3.1). Vertically, the model is discretized in a hybrid a-p-coordinate system with 19 levels. The time integration is semi-implicit with a time step of 40 minutes. ECHAM-1 is based on the primitive equations with vorticity and divergence, temperature, spe- cific humidity, the logarithm or surface pressure, and the cloud water mixing ratio (liquid and ice) as prognostic variables. Physical processes such as radiation, con- vection, cloud formation, precipitation, vertical turbulent mixing, gravity wave drag, 16 CHAPTER 3. THE EXPERIMENTAL SET-UP Figure 3.1: The ECHAM-I model orography (m) in T21 resolution and the model continents. phase transition and soil processes are parametrized, either because they cannot be resolved explicitly or because a more accurate treatment would exceed the computer resources. The diurnal cycle is included. The runoff into the ocean is calculated using a simple surface hydrology model. Since in ECHAM-1 the calving of glaciers is not considered, an attempt is made to correct the error caused by the accumulation of snow over glaciers using the flux correction for freshwater flux (see below). An enve- lope orography is applied, which is defined as the sum of the mean orography plus the sub-grid scale standard deviation. The complete description of the parameterization schemes employed can be found in Roeckner et al (1992). 3.1.2 The oceanic component The oceanic component is the Large Scale Geostrophlc (LSG) model (Maier-Reimer et al., 1993) which is based cm a numerical formulation of the primitive equations appropriatee for large scale geostrophlc motion. The model has 11 variably spaced levels in the vertical (centered at depths of 25, 75, 150,250,450, 700, 1000,2000,3000, 4000 and 5000 m) and a horizontal E-grid formed by two overlapping 5.6° x 5.6° grids (one for scalar and the other for vectorial fields), corresponding to an effective grid size of 4°, which are interpolated onto the Gaussian grid used in ECHAM- 1 T21. The nonlinear advection of momentum is neglected, and fast gravity waves are damped by an implicit time integration scheme with a time step of 30 days (which is reduced to 1 day in the coupled model for the computation of sea ice, temperature and salinity at the two uppermost ocean levels in order to resolve the response of the upper ocean to the synoptic variabilityy oft he atmosphere). Salinity and temperature transport are computed with an upstream advection scheme. A small explicit horizontal diffusion 3.2.THE CONTROL RUN 17 and a moderate viscosity are introduced to counteract the tendency for mode-splitting in the E-grid used in the horizontal discretization scheme. Vertical convective mixing occurs whenever the stratification becomes unstable. Sea ice is computed from the ice heat balance and the advection by oceanic currents, using a simplified viscous rheology. A realistic bottom topography is included. 3.1.3 The coupling Before coupling, the ocean model was integrated for 5000 years in a spin-up run driven by monthly climatological wind stress fields (Hellerman and Rosenstein, 1983) and a feedback to an effective monthly mean air temperature constructed from the COADS (Woodruf et al., 1987) and the annual-mean surface salinity from Levitus (1982). The fresh water fluxes diagnosed from this run, together with the same wind stress and air temperature forcing as used before, were used as external forcing for a subsequent integration of 2000 years. The final state of this integration was then coupled to the ECHAM-1 model. The atmospheric and ocean models are integrated synchronously, but each with its own time step, and they are coupled by the air-sea fluxes of momentum, heat (sensible, latent and radiative) and fresh water (precipitation minus evaporation, plus runoff along the coastal boundaries). The fluxes are calculated by the atmospheric model using the SST and the sea ice thickness as surface boundary conditions. Thus, the fluxes computed at each 40 minute time step by the atmospheric model are accumulated over the ocean time step and then transferred to the ocean model. A flux correction is applied in order to minimize a climate drift of the coupled system away from the climatologies simulated by the uncoupled models (Sausen et al., 1988). The flux correction is based in a linearization of the fluxes which couple the ocean and the atmosphere and, for small perturbations, it is assumed that it has no effect on the response of the climate. Such methods have been applied in greenhouse warming simulations to this and other models by different modeling groups (M.anabe et al., 1991; Manabe et al., 1992; Cubasch et al., 1992; Cubasch et al., 1995). Its use in climate simulation or sensitivity experiments has cast some doubt with respect to the internal consistency of such experiments. An assessment of the potential errors due to this technique would require a comparison to the true response of a coupled model without drift. However, sensitivity experiments with simple box models (Egger, 1997) suggest that the performance of the flux correction is especially suspicious in climate sensitivity experiments in which a transition to a new, essentially different equilibrium state takes place, which is clearly not our case. 3.2 The control run The starting point of the present study was the control run performed with the ECHAM-1 LSG coupled GCM in an attempt to reproduce the present climate, in which the atmospheric C02 concentration was fixed to 330 ppmv. The temporal evolution of the atmosphere and ocean, the time-mean states, and the variations of the qua.si- 18 CHAPTER 3. THE EXPERIMENTAL SET-UP stationary atmosphere-oceans ystem in the control run have been analyzed in previous works (Cubasch et al., 1992; von Storch, 1994; von Storch et al., 1997). The reader is referred to these works for a thorough description of the performance of the model in the control run; herein the main results are briefly summarized. Nevertheless, as a reference for the next chapters, a considerable number of the main results in the control run are shown in the Appendix. The main features of the atmospheric circulation are well reproduced. Main er- rors are thought to be due to) the low horizontal resolution and to the inappropriate gravity wave drag parametrization in ECHAM- 1 and have also been found in uncou- pled runs of this model (Roeckner et al., 1992). In December-February (DJF), the intensity of the Aleutian Low is overestimated, while excessive high pressures over the Arctic ocean lead to an underestimation of the Icelandic Low, which is located too far southwards. In June-August (JJA), the Northern Hemisphere subtropical highs are simulated realistically, while those in the Southern Hemisphere are too low by ca. 5 hPa and the polar trough is overestimated, so that the meridional pressure gradient in mid-latitudes is underestimated, a feature typical of low resolution models (Xu et al., 1990). The zonally averaged zonal wind is simulated realistically. The subtropical jets are well simulated both in location and magnitude, in summer and winter, but the winter double jet is not reproduced. Midlatitude baroclinic disturbances are not well resolved and cyclonic activity is underestimated. As a consequence, precipitation is also too low in the North Pacific and North Atlantic storm tracks in DJF, as well as in the southern midlatitude belt in DJF. Areas of total precipitation under 0.5 mm day-l are well captured by the model. The general structure of the mean oceanic circulation is reasonably well simulated. Main faiIures are the underestimation of the Kuroshio and the Gulf Stream trans- ports, an overestimation and too farther south location of the maximum value of the mean meridional circulation in the Atlantic, with too weak Antarctic bottom water in- flow, and overestimated equatorial upwelling in the Pacific. The Antarctic circumpolar current (ACC) is not well represented either and appears too broad in the model in comparison with observations. The sea ice component performs unsatisfactorily in the coupled model. After the coupling of the atmospheric and oceanic components in the control run, sea ice suffers a strong drift (von Storch et al, 1997). This is a supposedly consequence both of the simplified sea ice model and the flux correction applied (Cubasch et al., 1994; von Storch et al., 1997): the flwc correction, which is constant in time (except for the seasonal cycle) becomes inconsistent once the sea-ice distribution varies. As already mentioned, the flux correction. is applicable only in the case of small changes, for which a linearization is valid. The flux correction is calculated from the uncoupled runs and the observations; if at a certain point a mismatch between these exists (e.g. the ocean GCM does not yield sea-ice which is present in the observations) the flux correction at this point will be large; if then it is applied on a certain grid point where sea-ice conditions have changed from the uncoupled to the coupled run, unrealistic conditions (no sea ice or accumulation of sea ice) will result. For these reasons, main discrepancies occur at the sea ice edge. 3.3.THE EXTERNAL FORCING 19 Coz IEccentricity I Obliquity I Angle of perihelion I (PPmv) (degrees) (degrees) Control run 330 0.017 23.45 282.16 125 kBP 267 0.040 23.79 127.27 Table 3.1: Boundary conditions for the Eemian and control runs. 3.3 The external forcing In the Eemian run the ECHAM-1 T21/LSG climate model was integrated for 510 years starting at year 600 oft he control run. The external forcing of these runs differ in two ways (Table 3.1). First, instead of 330 ppmv as in the control run, the C02 concentra- tion in the Eemian run was prescribed to 267 ppmv, consistent with mean C02 levels of about 270 ppmv estimated for the Eemian from measurements in ice cores (Barnola et al., 1987). Second, the incoming solar radiation at the top of the atmosphere was changed by setting the Earth’s orbital parameters to their values at 125 kBP (E]erger, 1978): the eccentricity of the Earth’s orbit was greater, obliquity was slightly greater, and, with the vernal equinox fixed at March 20~ in both experiments, perihelion took place in northern sumer (July) instead of in northern winter (January), as in the present. The rationale behind this way of proceeding is the assumption that, at the time scales considered, the equilibrium climate essentially depends on the distribution of the incoming solar radiation at the top of the Earth’s atmosphere and the atmo- spheric C02 concentration. However, it should be taken into account that climate can be described as a random process whose parameters are determined by the external forcing, each realization of climate as simulated by a climate model being a random re- alization of this process. Hence, it should not be expected the simulation to reproduce in detail the paleoclimatic states reconstructed from paleo evidence (which in. turn, constitute another realization). Therefore, simulations should be done in ememble mode reflecting the inherent uncerttinty of the climatic process. e.g. by varying the initial state or the forcing within its range of uncertainty. The observed paleo climatic state should be a credible member of the ensemble. 3.3.1 The definition of the seasons The difference between the Eemian and present orbital parameters gives rise to a problem when comparing monthly and seasonal mean simulated fields (Kutzbach and Gallimore, 1988; Joussaume and Braconnot, 1997). Basically, for given values of eccentricity and obliquity, insolation over a given point on the outer limit of the Earth’s atmosphere at a fixed time is a function of the true longitude @, that is, the angle measured anti-clockwise from the vernal equinox (Figure 3.3). The precession of the equinoxes and the change in eccentricity translates into a variation of the dates at which a given true longitude @ is attained. As an example, maximum insolation in the Northern Hemisphere is received at the summer solstice, which, due to precession, takes place at different calendar dates for the present and for 20 CHAPTER3.THE EXPERIMENTAL SET-UP 125 kBP (Figure 3.3). The mismatch between the dates which correspond to a given value of @ increases as the Ea,rth moves farther away from the vernal equinox, which has been fixed in the two runs, up to about 12 days for @ = 210°. This effect is appar- ent in Figure 3.2, which shows the time versus latitude (top) and celestial longitude vs. latitude (bottom) diagrams of the zonally averaged differences in incoming solar radiation at the top of the atmosphere (125 kBP minus present). In the vicinity of the vernal equinox, both representations are essentially the same, while the disagree- ments between them increase as we move farther from the vernal equinox towards the autumnal equinox. merMkm/urne diatrib.ti.. 90N 60N 20N o 30S L?oe SOB in Feb. Mm-. Apr. Mev. Jun. Jul. AUS. fJep. OCL Nov. Dec. Jan, mridian/rAit diitributio” mm 60N 24N o 30s m @o@ 270 2C0 220 (W) 30 60 90 120 160 (A@ 210 340 270 Figure 3.2: Time vs. latitude (top) and celestial longitude vs. latitude (bottom) differences in zonalfy averaged incoming solar radiation at the top of the atmosphere (125,000years ago minuspresent, Win-2]. Since our interests focus on summer and winter mean fields, we have considered the orbit of the Earth (Monin, 1986) to redefine Northern Hemisphere summer and winter from an astronomical point of view, that is, as those which correspond to the same interval in @) as present Jurle-July-August (JJA) and December-January-February (DJF). Eemian Northern Hemisphere summer and winter are redefined as the periods between days 150 to 231, and days 320 to 59, respectively, in our 360-day model. Of course, for the centrol run the definitions of summer and winter are the usual ones. The fact that at 125 kBP perihelion occurred in Northern Hemisphere summer while in the present it occurs in Northern Hemisphere winter translates into a shorter summer and a longer winter for this epoch with respect to the present. The results show 3.3.THE EXTERNAL FORCING 21 that in our case the difference between the astronomical and meteorological (JJA, DJF) seasonal means is small compared to the actual difference fields between the Eemianand the present. Hence, wehaveadopted theusual JJAand DJF mean fields as northern summer and winter, respectively. However, this problem should :not be ignored when analyzing paleoclimate simulations for other time periods. In particular, due to the reasons specified above anomalies spurious insolation anomalies appear in autumn which would lead to misleading interpretations from the mean autumnal field. a)Controlrun WS (23.12) b)Eemianrnn WS (14.12) SS(20.6) VE (21,3) Figure 3.3: Position of the solstices and equinoxes a) for the control run (the present), and b) the Eemian run (125,000 years ago). Dates of solstices and equinoxes are given in parenthesis. VE: Vernalequinox, AE: Autumnal Equinox, SS: Summersolstice, WS: Winter Solstice. 3.3.2 Changes in incoming solar radiation The variation of the orbital parameters of the Earth modulates the incoming solar radiation at the top of the atmosphere. Differences in seasonal solar radiative forcing from present during the last interglacial were larger than during the Late Quaternary, and 125 kBP was the time of maximum seasonal insolation difference. Figure 3.2 shows Northern Hemisphere summer insolation increased at all latitudes, with an a,verage increase of 50 Win-2 (1270). The maximum increase of insolation (60 Win-2) took place at high northern latitudes in northern summer as a consequence of the timing of perihelion, which enhanced insolation mainly at mid-high northern latitudes, and the greater obliquity, which affected high northern latitudes. Winter insolation, in turn, 22 CHAPTER 3. THE EXPERIMENTAL SET-UP decreased at all latitudes with an average value of -26 Win-2 (11%); insolation decreases were larger at subtropical southern latitudes since at high latitudes its decrease due to the timing of perihelion was attenuated by the greater obliquity. To summarize, insolation was increased (reduced) at all latitudes during northern summer (winter), that is, the seasonal cycle of insolation was amplified at all latitudes in the Northern Hemisphere, while attenuated in the Southern Hemisphere. The insolation anomaly essentially affected the seasonal insolation cycle, since the change in mean annual incoming insolation was below 2 Win-2 at all latitudes (Figure 3.4). Annual mean insolation increased slightly at high latitudes and decreased at low latitudes. This was causeci essentially by the greater obliquity (summer insolation increases at high latitudes are not offset by tinter insolation decreases, since polar winters already receive no insolation). The change in the globally averaged incoming insolation is negligible (0.23 Win-2). ---- .,, ...... *~.. ..m. ..= ...... i.... m Gw Figure 3.4: Differencein zonall,yaveragedmeaninsolationin JJA (dotted line), DJF (dashed line] and in the annual mean (solid line) (Win-2). CHAPTE:R 4 Results with the ECHAM-1/LSG coupled GCM In this chapter the results of the climate simulation for 125 kBP with the coupled ocean-atmosphere GCM EC HAM- l/LSG are presented. The focus is on the the mean seasonal and mean annual large-scale atmospheric fields, the mean large-scale ocean circulation, the synoptic scale atmospheric variability, and the dominant modes of the large scale oceanic variations. In order to eliminate systematic errors, anomalies of the fields simulated for 125 kBP with respect to the control run are considered. 4.1 Temporal evolution The temporal evolution of the sea-ice, atmosphere and ocean systems in the control run are fully described in von Storch et al. (1997). After the change of the external forcing in year 600 of the control run to Eemian conditions, the sea ice and the near- surface temperature require slightly more than 100 years to attain quasi-stationary equilibrium (Figure 4.1). The globally averaged temperature of the upper layers of the ocean have not completely reached equilibrium after 510 years, but the trend is below –0.2°C/500 yr at all levels (Figure 4.2). In the deep ocean an even smaller, nearly constant trend in globally averaged temperature is found both in the E,emian and in the control run. This could be explained if the deep ocean had not completely reached equilibrium before its coupling to the atmosphere model due to the existence of non-zero globally integrated fluxes used in the uncoupled spin-up run (von Storch et al., 1997), but since the spin-up run was not stored this hypothesis cannot be tested. Consistent with a small negative globally integrated freshwater flux imbalance of -0.07 Sv, a trend in globally averaged salinity was also found in both runs at all levels, but 23 24 CHAPTER 4. RESULTS WITH THE ECHAM-1/LSG COUPLED GCM mainly at the surface (ea. O.15psu/500 yr; Figure 4.3). This, however, does not have any severe influence on the ocean dynamics: the typical indices which characterize the dynamical state of the ocean (the mass transport of the Antarctic Circumpolar Current (ACC) across the Drake passage, the outflow of North Atlantic Deep Water (NADW) at 30°S, and the amplitude of the Atlantic overturning circulation, defied as the maximum value of the zonally averaged mass transport streamfunction of the meridional overturning circulation in the Atlantic Ocean) show some signs of oscil- lations subsequent to the change in the external forcing, but reach a quasi-stationary state after 200-300 years (Figure 4.4). After that the Drake passage transport is almost unchanged whereas the Atlantic Deep Water Outflow and the NADW are enhanced by 0.5 Sv and 2 Sv relative to the control run mean values of 20.5 and 35 Sv, respectively. Figure 4.1: Time series of the mean annual a) total sea ice extent (m2), b) total sea ice volume (m3), in the Northern (solid line) and SouthernHemisphere(dashed line), and c) globally averagednear-surface(2 m) temperature(’C), for the control (thin line) and Eemian run (thick line). 4.2 Large-scale atmospheric patterns This section describes the difference (Eemian minus control run) in the mean large- scale atmospheric simulated fields in JJA, DJF, and in the annual mean, averaged over the final 300 years of each run. The focus is on the simulated anomalies relative to the control run. When not shown explicitly, figures for the control run are included in the Appendix for reference. 4.2. LARGE-scALE AtmOSpheriC PATTERNS 25 a) 75m–Temperature (C) a) 75m–Salinity (psu) + 10.4. --, ~~ 10.2-——6g—~~—8~——-v——-—v—.— 900 1000 11OD 1200 Time (years) b) 250m-Temperature (C) ~ S.4. -’- -’ In 6.2. —-~- .—.—..+-..—..- .. . . Lw 1- 6.0. . . . ”. . ...’...... - . ...?...... ”-. 7.8- 600 700 6CE3 900 1000 1100 1200 Time (years) c) 10OOm–Temperature (C) 0 +.4. ~~ o R ~~~~ g 4.2. -.-.-– ------=------=-~~-—-----=- ~ 34.8...... : . ...-...... ’.” _ W.7. ------.. -— 4- 1= ~ 34.6. ,. . ..- . . . . . -. 34.5 6W 700 600 6W 102$ ““”””””’1100 1200 Time (years) Time (years) d) 2000m–Salinity (psu) a 3S.0%.9...... 0$ wB~,J.--.--:-+ ::. . : : : –--::--- + 34.6.. . . ’...... Ei*%L& ws-–~,rx-—~=~lz-z 7 600 KU 800 900 1000 t 100 Time (years) Time (years) e) 4000m–Temperature (C) e) 4000m–Salinity (psu) 3.4 o ~ 3.2. ‘-” ““ 0 R- ...-. -... .---. --. -.— ~ 34.8- . —---- ...:”::.:””:::: —.-.. .-->.. 0 3- .-.---=-~ -.--.”., -. : * w7- ~~~ ~~ , - . --: ~~ ~ 2.8- . : ...... -.- . . . ~ ~- -. .-, --- :’3 600 700 800 900 1000 1100 1200 m X0 w 9W 1OQo 1100 1200 Time (years) Time (years) Figure 4.2: Time series of the mean an- Figure 4.3: Time series of the mean nual globally averagedtemperatureof the annual globally averaged salinity of the ocean at a) 75 m, b) 250 m, c) 1000 m, d) ocean at a) 75 m, b) 250 m, c) 1000 m, d) 2000 m, and e) 4000 m depth for the con- 2000 m and e) 4000 m depth for the con- trol (dashed line) and Eemian run (solid trol (dashed line) and Eemian run (solid line) ~ C). Note that the vertical scale is line) (psu]. Note that the verticalscale is the same for all levels. the same for all levels. Due to the large sample sizes available from the multi-century simulations, a test for statistical significance of the mean differences is not meaningful, since even negligi- ble differences will outcome as statistically significant. Instead, a univariate recurrence analysis was applied to the simulated atmospheric output fields to test for physical sig- nificance of their mean differences (von Storch and Zwiers, 1999). Briefly, two rimdom variables X, Y are said to be at least p-recurrent if the probability that Y be greater than the mean value of X is greater than or equal to p (P(Y > ~) ~ p). p is therefore a measure of the overlap of the distributions of the X and Y. If ~ = ~ then p=O.5 and, as long as their distributions are of the same type and equal dispersion, their distributions are coincident and physically indistinguishable; the overlap decreases as p deviates from 0.5. Values of p close to O (1) imply that the distribution of X and Y seldom overlap and that Y is generally much larger (smaller) than X. In climate modeling applications, X usually represents the sample space (i.e., the 26 CHAPTER h. RESULTS WITH THE ECHAM-1/LSG COUPLED GCM a) Mass transport Drake passaqe (Sv) b) Atlantic Deep water Outflow (Sv) Time (years) c) Ampl;tude of Atlantic Overturning (Sv) Figure 4.4: Timeseries ofa)n~ass transport of the Antarctic Circumpolu Current (ACC) across the.Drake passage, b) outflow of North Atlantic Deep Water (NADW) at 3@S, and c) amplitude of the Atlantic overturningcirculation, defined as the maximum value of the zonally averagedmass transport streamfunction of the meridional overturning circulation in the Atlantic Ocean, for the centrol (dashed line) and the Eemian run (solid line) (Sv). list of possible outcomes) of a given variable in a control climate experiment at a given location and and Y the corresponding sample space in an anomalous climate, obtained by altered boundary conditions (in our case, the control and Eemian runs, respectively). At each time step, one realization of X and Y is obtained; for a finite time period, an ensemble of realizations is collected from which p can be estimated at a given point. If X and Y are two normally distributed random variables with mean values ~ and ~ and standard deviation a, then p is given by (4.1) where F is the distribution function of the standard normal distribution N(O,l). p represents the local level of recurrence and is a measure of of the strength of the effect of the anomalous boundary conditions. Physically non-negligible local differences correspond to extreme values of the local level of recurrence p, which is indicated by shading: light shading refers to p~O.8 or psO.2 and heavy shading refers to p~O.95 or p 4.2.1 Northern summer In JJA, the increase of insolation at all latitudes translates into an increase of surface solar radiation almost everywhere (Figure 4.5a), modified mainly through changes in cloud cover in the tropics (Figure 4.5b). This causes overall enhanced near-surface temperattu-es over land with maximum values above 3-5°C in the northern continents (North America, north Africa and Eurasia), where the insolation and surface solar radiation increases and the landmass distribution are greatest (Figure 4.5c). This response agrees with those found in an AGCM experiment with prescribed SSTS (Prell and Kutzbach, 1987) and with an EBM (Crowley and Kim, 1994). This agreement is consistent with the view that summer temperatures reflect essentially the response to radiative rather than dynamic changes (Hurrel, 1996; Mann and Park, 1996). Due to the larger heat capacity of the oceans, temperatures over the ocean change in less than 1°C except at high northern latitudes, where SST increases are amplified through the positive sea ice albedo-SST feedback: the increase of summer insolation and surface solar radiation at high northern latitudes leads to enhanced SSTS; Arctic sea ice cover is reduced (Table 4.1 and Figure 4.6), further enhancing the absorption of solar radiation at the surface (Figure 4.5a), and hence amplifying the SST response. However, some of the temperature changes at high latitudes, especially in the Pacific Ocean, do not appear to be highly recurrent. The near-surface temperature decreases by 2-3°C over central Africa and south Asia, as a consequence of enhanced cloudiness (Figure 4.5b) which, in turn, is associated with changes in southwest monsoon (see below). The planetary albedo is enhanced, and the surface solar radiation reduced (Figure 4.5a). Cooling by up to 2°C is also found over Antarctica. The Mediterranean Sea shows a very strong cooling. This is in fact an artifact of the model caused by the fact that with the current resolution of the model, the Strait of Gibraltar is not resolved and hence the Mediterranean is a closed basin. Highly recurrent sea level pressure (SLP) decreases by ca. 3 mb (i.e. 0.3%) (Figure 4.5d) take place over the northern midlatitude and southern subtropical landmasses, especially over Eurasia, north Africa and North America, and reflect the thermally direct response to the temperature changes. SLP increases over the surrounding cooler oceans, especially over the Southern Ocean and North Western Pacific. The land- ocean surface pressure gradient is accordingly enhanced, and highly recurrent changes in the low level atmospheric circulation and precipitation take place in the tropics (Figure 4.5e, f): the low level convergence is intensified over north Africq southwesterly surface wind speed increases by 1-2 ins-l (1O-2O7O)(Figure 4.5e). As a consequence of enhanced moisture transport, precipitation increases up to 2 mm day‘1 (Figure 4.5f ). In south Asia, the low level south westerly flow is shifted northwards; easterly surface wind anomalies of 1-2 ins-l appear over the Arabian Sea, south Asia and the northern Indian Ocean, while south westerly anomalies of similar magnitude are found over the Arabian Sea and north India (Figure 4.5e); moisture transport further inland is enhanced, giving rise to a decrease of precipitation in south India by 2mm day–l and a very large increase over the Tibetan region (Figure 4.5f ), reflecting a northward 28 CHAPTER 4. RESULTS WITH THE ECHAM-1/LSG COUPLED GCM migration of the large-scale convergent flow, as has been reported in previous climate simulations for 6 kBP (Hewitt and Mitchell, 1996) and for 126 kBP (de Noblet et al., 1996). Precipitation over northern South America is also enhanced (Figures 4.5e), mainly due to enhanced convection. At 200 hPa, a large region of anomalous easterly flow (negative anomalies) is found throughout low latitudes of the Southern Hemisphere and mid-low latitudes of the Northern Hemisphere, all of which are highly recurrent (Figure 4.5h). In the present climate, the summer monsoon at upper levels is characterized by the onset of an anti- cyclone over south Asia, as is seen from the 200-mb streamfunction. In addition, the 200-mb velocity potential field.x shows a region of large-scale divergence (X < O) over the western Pacific which is associated with a convection center, and large-scale con- vergence over the eastern Atlantic. These features, which are satisfactorily captured in the control run (Appendix), are perturbed in the Eemian run in several ways: the 200-mb streamfunction (Figure 4.5i) and the 200-mb velocity potential pattern (Fig- ure 4.5j) are both shifted north-westwards, and the meridional gradient of the 200-mb streamfunction is enhanced over south Asizq the result are enhanced upper-level easter- lies throughout the whole northern tropics, reflecting an intensification of the tropical easterly jet (Figure 4.5g, 4.7a). The same features were found by de Noblet (1996) in an AGCM simulation for 6 kBP and 126 kBP. Hence, the above results reflect an intensification and redistribution of the mon- soonal circulation in the Northern Hemisphere caused by the enhanced land-sea temper- ature contrast in this hemisphere, which agrees both with previous AG CM simulations of the last interglacial and with the evidence from the geological record (Petit-Maire, 1994). Changes of total precipitation in regions other than those affected by the monsoonal circulation have a much smaller amplitude and are not highly recurrent. The zonally averaged precipitation over la,nd only and over land plus ocean (Figure 4.8) reflects the aforementioned changes in the monsoon circulation in the tropics, but only minor differences in extra-tropical regions. P-E decreases over Eastern Asia and most of the ‘ Pacific and Central America (Figure 4.5g). NH SH Global JJA cover -0.1052 X 1013 0.1176 X 1013 0.0012x 1014 JJA volume -0.1795 x 101* 0.1654 X1013 -0.1630 X1014 DJF cover -0.0087 X1014 0.0696 X1013 -0.0017 x 1014 DJF volume -0.1907 x 101* 0.1294 X1013 -0.1776 X1014 Mean annual cover -0.0108 X 1014 0.0892 X1013 -0.0018 X 1014 Mean annual volume -0.1819 X1014 0.1480 X1013 -0.1671 X1014 Table 4.1: Differences (Eemia.n minus control) in mean simulated sea ice cover (m2,) and volume (m3) in JJA, in DJF, and in the annual mean for both hemispheresand global. Results indicate averages over the last 300 years of the Eemian and control runs). At mid-northern latitudes, the upper level westerly flow is reduced over the Atlantic and Pacific Oceans and over southern Eurasia, whereas positive anomalies in the zonal J.2.LARGE-SCALE ATMOSPHERIC PATTERNS 29 Figure 4.5: Mean difference (13emianminus control run) in JJA in a) net surface solar radiation (Isoline spacing (1S) = 10 Win-2); b] total cloud cover (1S = flo); c) near-surface temperature (IS = 1° C), d) SLP @S = 1 rob), e) 10-m winds (ins-l), f) total precipitation and g) precipitation minus evaporation (isolinesat &15, *1O, *5, &2, &l, &O.5mm c?ay-l); h) 200-hPa u-velocity (1S = 2 ins-l); h) 200-hPa stream fwction (IS = 3Xl@m2s-l); i) 200-hPa velocity potential (1S= l@m2s–1) Shading indicates the level of local recurrence p. Light shading: p>O.8 or p~O.2; heavy shading: p~O.95 or p< O.05. 30 CHAPTER 4. RESULTS WITH THE ECHAM-1/LSG COUPLED GCM ,re 4.5: Cont. J.2.LARGE-SCALE ATMOSPHERIC PATTERNS 31 i) Psi200 (1 O**6 m**2/s) j) Chi200 (1 O**6 m**2/s) Figure 4.5: Cont. velocity component appear in Northern Eurasia (Figures 4.5h). The former feature can be attributed to the attenuation of the meridional temperature gradient at northern midlatitudes caused by the insolation perturbation, which translates into a reduction of the zonal component of the midlat itude westerly flow at all levels. The latter one reflects a northward shift of the large-scale atmospheric circulation, caused by the northward shift of the maximum meridional temperature gradient over the Eurasian continent. Together with the aforementioned intensification of the tropical easterly jet, the re- duction of the zonal component of the midlatitude westerly flow translates into easterly anomalies in the zonally averaged zonal wind at all northern latitudes and levels? which are all highly recurrent. In the Southern Hemisphere, on the contrary, the meridional temperature gradient is slightly enhanced, giving rise to enhanced westerlies over the Atlantic Ocean and the Southern tips of south Africa and South America (Figures 4.5h), reflected in the zonal mean (Figure 4.7a). 32 CHAPTER 4. RES lJLTS’WITH THE ECHAM-1/LSG COUPLED GCM Figure 4.6: Mean annual cycle (averaged over the final 300 years of each run) of total a) Arctic and b) Antarctic sea ice cover (poleward of 65°iV and 65°S, respectively) for the control (dashed line) and Eemian run (solid line) (18 km2). a) U Eem– Control NH summer Figure 4.7: Difference (Eemianminus control run) in mean zonally averagedzonal velocity in JJA and DJF (m s–l). Shading indicates the level of local recurrence p. Light shading: p>O.8 or p :’J/./ , ...””..‘:...”,.’ .;”..””..““ EGS’MSEQ30NWN 90N !!@L!? Figure 4.8: Difference @emian minus control run) in mean zonally averagedprecipitation over land (dashed line) and over land plus ocean (solid line] in JJA, DJF and in the annuai mean (mm day–l). 4.2.2 Northern winter As in JJA, most recurrent changes of atmospheric fields in DJF take place mainly over the tropics. In DJF2 the decrease of insolation translates into reduced net surface solar radiation almost at all latitudes (Figure 4.9a). Such decreases are attenuated at southern subtropical landmasses (South Africa, Eastern Equatorial America,, Aus- tralia), where cloud cover, and hence the planetary albedo, are reduced (Figure 4.9b). Hence, maximum near-surface temperature decreases occur in the northern subtropical continents, while the southern subtropical landmasses show a more moderate cooling (Figure 4.9c). As a consequence of the enhanced summer insolation at high northern latitudes, amplified through the positive sea ice albedo-SST feedback, the sea ice re- duction is not limited to JJA, but takes place all year round (Figure 4.6). Accordingly, the absorption of solar radiation at the surface is enhanced during spring and summer, causing enhanced warming up to 2*C at high northern latitudes. Deficiencies in the performance of the sea ice component in the control run (von Storch et al., 1997) advice caution when dealing with the climate response simulated at high latitudes (Cubasch et al., 1995). However, the fact that the above mentioned features have been observed in all paleoclimate simulations of time periods with enhanced northern summer inso- lation with models of different complexity including an oceanic component (Kutzbach and Gallimore, 1988; Mitchell et al., 1988; Kim et al., 1998; Hewitt and Mitchell, 1998) indlcat e that they are not a model artifact. However, for the same reasons the focus should be on the qualitative response only. 34 CHAPTER 4. RESULTS WITH THE ECHAM-1/LSG COUPLED GCM Mid-northern latitudes alsc}show a moderate warming which reflect in part changes in northward heat transport by the ocean associated to changes in the ocean circulation (Section 4.3). Nevertheless, these do not appear to be highly recurrent. Opposite to what was seen for JJA, the Mediterranean Sea shows a very strong warming, again related to the fact that it is a closed basin. SLP (Figure 4.9d) increases over the colder subtropical landmasses of both hemi- spheres, especially over central Africa and south Asia, and decreases over the surround- ing southern oceans. As in JJA, the land-sea pressure contrast is thereby enhanced in the Northern Hemisphere, while reduced in the Southern Hemisphere. Strong outflow takes place over central Africa and south Asia (Figure 4.9e), causing enhanced low level convergence and precipitation over the Indian Ocean at about 20°S (Figure 4.9f ). At the same time, precipitation decreases over the subtropical southern continents (South America, south Africa and Australia), up to 4 mm day-l. These results reflect a weak- ening of the southern summer monsoon and a very strong northern winter monsoon response, as seen in Hewitt et al. (1996). The former is accompanied by a decrease in total cloud cover in southern subtropical landmasses which accounts for the milder cooling in this region. Precipitation is also reduced by ca. 0.5-lmm day-l over the northern mid-latitude Atlantic and Pacific Oceans, as a result of the attenuation of the storm-tracks in DJF in these regions (Section 4.4. 1). An intensification of the Icelandic low and an attenuation of the Aleutian low is found (Figure 4.9d). These changes are reflected in a spin-up and down of the corresponding cyclonic gyres, In the case of the Pacific this feature is highly recurrent. Responses of the Aleutian and Icelandic lows may well be model- dependent (Kageyama et al., 1998) but could reflect upstream feeding of one system by the other (Hoskins et al., 1983; Kageyama et al., 1998). The mid-latitude meridional temperature gradient is also reduced in winter, leading to decreased westerly flow at all levels (Figures 4.9d and 4.7b), in particular at the core of the jet stream At southern midlatitudes the opposite effect takes place, and the tropical easterly jet is reducecl, which together with the fact that circulation patterns are shifted northward translates into westerly anomalies over the tropical Atlantic ocean (Figure 4.7b). 4.2.3 Annual mean Highly recurrent mean annual features occur, again, in the tropics. Despite the large warming simulated in JJA, significant cooling occurs in winter, with the net result that the mean annual temperatures (Figure 4. 10b) are lower than 1°C almost everywhere, and the difference (Eemian minus control) in mean annual globally averaged tempera- ture is -0.32”C. In a similar way, the strong Asian winter monsoon over-compensates its intensification in summer, and mean annual simulated features (e.g. surface winds in the Arabian sea, Figure 4.1.Oe) reflect the winter rather than the summer response. In addition, the simulated globally averaged precipitation on land is reduced by 170 (Table 4.2), with a recurrence level of 0.86. Mean annual precipitation minus evaporation (P-E) also decreases in the annual mean over most of Eastern Asia and North America (Figure 4.10h) and globally P-E over land is reduced (Table 4.2), with 4-2.LARGE-sCALE ATMOSPHERIC PATTERNS 35 Figure 4.9: Asin Figure 4.5 butin DJF. 36 CHAPTER J. RESULTS WITH THE ECHAM-1/LSG COUPLED GCM i \ Figure 4.9: Cont. .4.3. OCEAN CIRCULATION 37 Figure 4.9: Cont. the opposite increase over the oceans; hence, moisture transport from the ocean to the land decreased globally at the Eemian with respect to the control run; i.e. the intensity of the hydrological cycle is reduced. 4.3 Ocean circulation Due to the large memory of the ocean, the number of degrees of freedom, that is, of independent realizations in the 300 years available is very limited. Hence, it is not meaningful to apply a local recurrence test to the ocean simulated fields. The largest difference concerning the surface (25 m) ocean currents (Figure 4.lla) is the strong northeasterly flow along the eastern coast of North Africa reflecting an attenuation of the Somali current. In addition, the southwesterly flow across the North Atlantic and Pacific Oceans is slightly intensified. AS a consequence, in the mean horizontal barotropic (vertically averaged) streamfunction in the Pacific Ocean the 38 CHAPTER 4. RESULTS WITH THE ECHAM-1/LSG COUPLED GCM Figure 4.10: As in Figure 4.5 but in the mean annual. 4.3. OCEAN CIRCULATION 39 Figure 4.10: Cont. 40 CHAPTER J. RESULTS WITH THE ECHAM-1/LSG COUPLED GCM i) Psi200 (1 O**6 m**2/s) j) chi200 (1 O**6 m**2/s) Figure 4.10: Cont. L o Glob NHL NHO NH glob SHL SHO SH glob JJA P 0.15 -0.23” -().12* 0.24” -0.25” -0.06 -0.04 -0.21” -0.18” JJA P-E 0.04” -0.06” -0.03* 0.10” -0.02 0.03” -O.1O* -O.1O* -0.10” DJF P -0.25* 0.17” 0.05’ -0.05* 0.14’ 0.07 -0.70” 0.20” 0.04” DJF P-E -0.15* 0.10” 0.03” -0.01 0.03 0.02 -0.47’ 0.15” 0.04* AM P -0.03” -0.04” -().04* 0.04” -0.02 0.00 -0.20” -0.04” -0.07” AM P-E -0.01 0.01 0. 0.04” -0.02 0.01 -0.14” -0.02” -0.01 Table 4.2: Difference (Eemian minus control) in mean simulated precipitation and pre- cipitation minus evaporation for selected areas (land, ocean, global, Northern Hemisphere land, Northern Hemisphereocean, Northern Hemisphere global, Southern Hemisphere land, SouthernHemisphere ocean, Southern Hemisphereglobzd) in JJA, in DJF, and in the annual mean (mm day–l); * indicates those values which are above the 0.8 (or below the 0.2) level of recurrence. .4.3. OCEAN CIRCULATION 41 subtropical gyres are slightly intensified with respect to the control run (Figure 4. llb). A slight intensification (ea. 5%) of the meridional overturning circulation in the Atlantic Ocean is found at the Eemian with respect to the control run, with enhanced NADW formation compensated by increased inflow of warm waters at upper layers (Figure 4.llc). Since the ocean thermohaline circulation is controlled by the ocean meridional den- sity gradient (Rahrnstorf, 1996), which in turn depends on temperature and sali:nity,it is necessary to examine how these two fields are affected in the Eemian run relative to the control. Since the high northern latitude temperature response reflects a warming relative to the control (which translates into enhanced surface stability), the NADW re- sponse is contrary to what should be expected from the thermal response only. Hence, we need to analyze the salinity response, which is in turn controlled by changes in the freshwater flux (P-E plus runoff along the coastal boundaries). In the annual mean, a decrease of the freshwater flux is found north of 40”N in the Atlantic Ocean (Figure 4.13), which could account for the intensification of the NADW found. According to the mean annual simulated differences, the mean annual freshwater flux decrease is mainly due to a slight increase in evaporation in the entire North Atlantic ocean (by ca. 0.02 Sv at 30”N), rather than to a decrease in precipitation or runoff. Nevertheless, since it is mainly in DJF when NADW formation takes place, the winter response must be considered separately. Although in DJF, as a consequence of the attenuation of the storm-tracks over the North Atlantic ocean (Section 4.4. 1) a decrease of precipitation in these areas by 0.5 mm day–l in DJF (Figure 4.9f ), which contributes to the total reduced integrated freshwater flux in the North Atlantic. In the Pacific Ocean (Figure 4. lld), the difference in the mean overturning circu- lation shows an intensification of the Antarctic Bottom water (AABW) inflow with enhanced upwelling at the surface, a similar response to that obtained by Kim et al. (1998). In addition, an anomalous cell flow showing intermediate water formation in the North Pacific is found: convection at high northern latitudes is triggered as a result of the reduced integrated freshwater flux at all latitudes in the Pacific basin (Figures 4.13 a, b). The reduced freshwater flux, in turn, is caused both by a direct decrease in P-E over the North Pacific (e.g. at 15-20”N evaporation and precipitation contribute with -0.04 and -0.02 Sv, respectively) and by changes and redistribution of precipitation and P-E over southeast Asia related to changes associated with the Asian monsoon. In JJA, precipitation appears to decrease over most of the Pacific Ocean. Evaporation is also enhanced over Eastern Asia. In addition, the monsoonaI circulation is shifted northwards, affecting the distribution of the runoffs into the rivers of these areas. In DJF, precipitation over the Pacific Ocean is decreased by 0.5 mm day-l at mid-high latitudes. Altogether, in the annual mean the runoff into the Indian Ocean is enhanced at the expense oft he northeast ern Pacific (-0.02 Sv at 15-20°N), reducing and enhanc- ing the surface salinity, respectively, in these regions (Figure 4.12). Changes in the Indian Ocean are not shown because they are negligible. As a consequence of the changes in ocean circulation, the global northward heat transport by the ocean is enhanced at mid-latitudes, with the Pacific the main con- tributor to this change (Figure 4.14). The peak value of the change in northward heat 42 CHAPTER J. RESULTS WITH THE ECHAM-1/LSG COUPLED G(2’M ,s:::::::*3&y ““”;-;:::y.!... . ::>*::fi>~j;,..%;;;!;;;::::!;:)-“. .,- , ;:,:;:: 90- “’“““” -w..+ SON-“5;,8 Figure 4.11: Difference (Eemian minus control run, right) in mean 25-m currents (ins-l) (top left); in the horizontal barotropicstreamfunction (top right), and in the zonally averaged mass transport stream function of the meridional circulation in the Atlantic Ocean (bottom left) and in the Pacific Ocean, averaged over 300 years (bottom right) (Sv). transport (ea. 0.12 PW) is attained around 20”N for both the Pacific and global oceans, representing hence an increase by 10 and 50%, respectively at this latitude. 4.4 Climate variability y While the high-frequency atmospheric variability on time-scales up to ca. 10 days is well characterized and documented from the observational record (Blackmon, 1976), the limited length of the oceanographic observational record precludes the possibil- ity of studying the low-frequency oceanic variation using observational data. For this purpose, one can resort to the use of long coupled GCM simulations, which hence con- stitute a promising tool for the study of low-frequency climate variabilityy. In addition, the use of coupled GCMS in paleoclirnate simulations may provide information on the dependence of climate variability at different time-scales on the external forcing, zm information which cannot be obtained from the evidence from the geological record. The analysis of climate variability is herein restricted to the synoptic scale variabil- Salinity 75 m {mu) ecu 80s. ~~ >- ...... lm 120N m 0 aE 120E m .7 1 2 5 6 9b 6b 30 EQ 3i3 N 60 90 Figure 4.12: Difference (Eemian minus control run) in mean 75 m salinity (top) and in a vertical section across the Pacific Ocean (bottom] (psu). 44 CHAPTER 4. RESULTS WITH THE ECHAM-1/LSG COUPLED GCM .,’”r : -\& J___ -2’.-: E-o+””1“/”””’”’”””‘“==’;”~~” $ -’i” .V & .l.2L- : : 60S 30s EQ 30N 60N b) EEM – CONTROL T Global ~ 0.12 -—— Atlantic : -._._ ._. Pacific x 0.09 — -_— Indian 3 G= 0,06 ‘\-,,,’ — 1~:60S 30s Efl 30N WN 90N Figure 4.13: Mean integrated freshwaterflux from 90’N southward for the three ocean basinsand the global ocean for a} the control run b) the difference Eemianminuscontrol run (sV). ity for the atmosphere and tcl the large-scale oceanic variations. Future paleoclimate simulations with improved models should include the analysis of additional aspects, such as blocking phenomena, monsoon variability or inter-decadal variabilityy associ- ated to ENSO (El Niiio-Southern Oscillation), which is not resolved in such a coarse resolution model. 4.4.1 Atmospheric variabilityy: synoptic scale disturbances Due to the reasons specified above, the focus is on climate variability associated with the passage of the synoptic-scale transient midlatitude disturbances (storm-tracks): these generate variability on the 2-6 days time scale and hence can be characterized by the peak values of the intra-seasonal standard deviation of the band-pass geopotential height at 500-mb; band-pass refers to variability on time scales of 2.5-6 days (Blackmon, 1976). Although severely underestimated in the control run, their location over the North Atlantic and North Pacific and over the Southern Ocean is well captured. These disturbances originate through baroclinic instability of the basic state, whose intensity is related to the meridional temperature gradient throughout the troposphere. An 4.4. CLIMATE VARIABILITY 45 a) CONTROL 2 ————————————————-—- S 1- b) EEM - CONTROL f N Figure 4.14: Mean northward heat transport for the three ocean basins and the global ocean for a) the control run b) the differenceEemian minus control run (P W). additional consequence of the reduced meridional temperature gradient is hence an attenuation of the baroclinic instability at mid-latitudes leading to a reduction of the storm track intensity in the Northern Hemisphere in both seasons (Figure 4.15a). In the Southern Hemisphere the opposite is found, anomalies coinciding in sign and location with anomalies in the mean westerly flow at 200 mb (Figures 4.5g, 4.9g). Some areas at high northern latitudes of the Northern Hemisphere also show positive anomalies, associated with the northward shift of the meridional temperature gra&ent maximum. This has implications for the large-scale precipitation changes, since the attenuation of the storm-tracks over the North Atlantic and Pacific Oceans causes a decrease of precipitation in these areas by 0.5 mm day–l in DJF (Figure 4.9f) which in turn affects the freshwater flux balance in these regions. 4.4.2 Ocean variabilityy The focus here is to compare the global ocean variations, as simulated in the Eemian run, to those simulated in the control run, described by von Storch et. al (1997). This includes the analysis of the spatial distributions of the oceanic variations, the dominant large-scale modes of variations and the extent to which such variations in the Atlantic 46 CHAPTER J. RESULTS WITH THE ECHAM-1 /LSG COUPLED GCM a) JJA a) DJF Figure 4.15: Difference@emian minus control run) in the intra-seasonalstandard deviation of the band-pass geopotential height at 500-mb in a) northern summer (JJA) a) northern winter (DJF) (1S = 1 m). and the Pacific ocean are related to each other and to the barotropic circulation. The spatial distribution of the large-scale oceanic variations is given by the standard deviation of their yearly anomalies (Figure 4.16). As already mentioned, due to the large time response of the thermohdine circulation, the number of degrees of freedom in 300 years is very limited. Hence, power at lowest frequencies is bound to be missed, and the tot al variability will be underestimated. No considerable differences are found with respect to the location of the areas with greatest variability with respect to the control run, although as a consequence of the smaller number of years available, the variability is underestimated and appears to be lower than in the control run, calculated for 800 years. Maximum variations occur in the northern extra-tropics and alcmg the ACC in the horizontal barot ropic streamfunction occur and at the center of the meridional cells in the Atlantic and Pacific meridional circulation and around 1000 m south of the Equator in the Indian Ocean. The dominant modes of the large-scale ocean circulation are described by the first empirical orthogonal function (EOF) (von Storch and Zwiers, 1999) of the respective streamfunctions (Figure 4.17). The results are essentially the same as in the control run (von Storch et al., 1997). In all cases the first EOF represents as much as one 4.4. CLIMATE VARIABILITY 47 90- ‘ N 6Q- 30- 0- 30- 60- S 90- 2%E BkE 1*5E 1k5w 9%W 3hw 2’5E Figure 4.16: Standard deviation of the yearly horizontal barotropic stream fimction (top left), the zonally averagedmass transport stream function of the meridional circulation in the Atlantic (top right), in the Pacific (bottom left) and in the Indian Ocean (bottom right) (all in 0.05 Sv]. Barotropic Atlantic Pacific Indian Explained variance 29 29 29 42 Table 4.3: Total varianceexplained by each of the first EOFS plotted in Figure 4.17 (’%). third of the total variance. The first EOF of the horizontal barotropic streamfunction shows a westward (eastward) anomalous circumpolar flow along the Antarctic coast and clockwise (anti-clockwise) circulations in the southern Atlantic and southeast of South Africa and in the southern Pacific. The first EOF of the meridional overturn- ing circulation in the Atlantic and Pacific oceans represents essentially re-circtdations Within each ocean basin. While in the control run, the first EOF for the meridional circulation in the Indian Ocean indicated a connection to regions south of 30°S, in the Eemian run it shows a more similar behavior to the Atlantic and Pacific Oceans in the sense of the aforementioned re-circulations. Figure (Figure 4.18) shows the cor- responding principal components (Figure 4.18), to which a 15 year running mean has been superimposed. As for the control run, decadal variability is more pronounced in the Atlantic overturning circulation. No considerable correlation was found between these time series. 48 CHAPTER 4. RESULTS WITH THE ECHAM-1/LSG COUPLED GCM 90-’ N 60- 30- 0- 3D- 60- S 90- kE B’5E 3L5E Jk5n’ 9’5W 245W 2’5E Figure 4.17: First EOF of the yearly anomaliesof the horizontal barotropic stream function (top left], the zonally averaged mass transport stream function of the meridional circulation in the Atlantic (top right), Pacilic (bet tom left) and in the Indian Ocean (bottom right) (0.05 Sv). Barotropic Atlantic Pacific Atlantic 0.004 Pacific 0.130 0.194 EIndian 0.310 -0.004 -0.004 Table 4.4: Correlation between the first principal components plotted on Figure 4.18 4.5 Summary The results of the climate simulation for the LIG with the coupled atmosphere-o cean GCM ECHAM-1/LSG reflect the amplification of the seasonal cycle of temperatures in the Northern Hemisphere at this time period, relative to the control run, due to the change oft he orbital parameters. The simulated thermal response, which closely agrees with previous climate simulations for time periods periods of the Late Pleistocene with enhanced Northern Hemisphere summer insolation (Prell and Kutzbach, 1987; Kutzbach and Gallimore, 1988; de Noblet et al., 1996; Hewitt and Mitchell, 1996), has two main implications for the atmospheric circulation, namely: In the northern hemisphere, both in summer and winter, the land-sea temperature contrast is intensi- fied, and the meridional temperature gradient at mid northern latitudes is attenuated. At the same time, its maximum is shifted northwards. As a consequence, the land- 4.5. SUMMARY 49 4-$32 m 700 w Ox low 1100 1X4 lime [years] b) Time bears] lime ~eors] d) Figure 4.18: First principal component of yearly anomalies of a) the horizontal barotropic stream function, and b) the zonally averaged mass transport stream function of the meridional circulation in the Atlantic Ocean and c) in the Pacific Ocean, as simulated in the Elemian run. sea surface pressure gradient, which over the tropics responds thermally directly, is enhanced in the Northern Hemisphere in both seasons. Accordingly, the southwest monsoon circulation is intensified. The results qualitatively reflect an intensification of the southwest Asian and African monsoon circulation and precipitation in summer, consistent with the evidence from the geological record (Petit-Maire, 1994) and with previous climate simulations of warm summer orbit periods of the late Pleisl;ocene (Kutzbach and Guetter, 1986; Kutzbach and Gallimore, 1988; Mitchell et al., 1988). Features such as the northward migration of the large-scale convergent flow and the intensification of the tropical easterly jet are found as reported in AGCM simulations for 6 and 125 kBP (de Noblet et al., 1996; Hewitt and Mitchell, 1996). The reduction of the meridional temperature gradient at mid northern latitudes, translates into an at- tenuation of the mean extra-tropical (westerly) circulation throughout the troposphere and of the the baroclinic activity of the atmosphere is also reduced; the storm-tracks over the North Atlantic and Pacific are attenuated, and precipitation over these regions is reduced relative to the control run. In the Southern Hemisphere, as a consequence of the attenuation of the seasonal cycle of insolation, the land-sea thermal and pressure contrast are reduced, and the summer southern monsoon is both weakened. The same resuh was also found by Kutzbach and Guetter (1986) in their AGCM simulation for 50 CHAPTER J. RESULTS WITH THE EC’HAM-i/LSG COUPLED GCM 9 kBP. The meridional temperature gradient, in turn, is slightly intensified. Despite the fact that the largest thermal response occurs over the extra-tropics, in correspondence with the insolation perturbation, changes in precipitation over the tropics, and minor or not highly recurrent over the extra-tropics. These results are at variance with a previous climate simulations of periods with enhanced Northern Hemisphere summer insolation such as 9 kBP showing changes in these fields havering much larger amplitude (e.g. enhanced precipitation over the warmer Northern Hemi- sphere continents in northern summer, and decreased precipitation over the oceans) (Kutzbach and Guetter, 1986; Mitchell et al., 1988; Kutzbach and Gallimore, 1988). In addition, the simulated globally averaged precipitation is reduced by 1%, contrary to the idea that enhanced precipitation in the tropics due to intensified monsoons leads to enhanced precipitation on land as a whole (Keigwin and Boyle, 1985)). The large seasonal changes cancel out in large part when we consider annual mean surface temperature changes, which also reflect the previous asymmetries. Mean an- nual near-surface temperature changes are generally below 10C, and the change in global near-surface temperature is -0.3°C. As seen also by Hewitt and Mitchell (1996), strong winter cooling translates into a very strong Asian winter monsoon which over- compensates the summer response, and the mean annual simulated features reflect a reduction of the intensity of the monsoon rather than an enhancement. A slight intensification of the meridional overturning circulation is found in the Eemian, relative to the control run, both in the Pacific and Atlantic oceans, as a consequence of the decreased fresh water flux at mid and high northern latitudes. This results in an enhanced global northward heat transport by 107o at 15”N. The largest changes are found in the Pacific Ocean, and are related both to northern summer changes in the hydrological balance over the tropics, associated to precipitation and runoff redistributions over Eastern Asian, and to a reduction of precipitation associated to the storm-tracks in northern winter. Finally, the analysis of ocean variability shows essentially no changes relative to the control run. The main modes of ocean variations take the form of re-circulations within each ocean basin, whiclh are essentially independent of each other. Inter-decadal variability is, as for the control run, more pronounced in the Atlantic Ocean. CHAPTER 5 Results with the CLIMBER-2 climate system model In this chapter the results of a second climate simulation for 125 kBP with the climate system model of intermediate complexity CLIMBER-2 are compared to those with the coupled atmosphere-ocean GCM ECHAM-1/LSG, previously described in Chapter 4. The term intermediate conqdezitg refers to the fact that CLIMBER-2 fills the gap be- tween GCMS and simplified models such as EBMs. As shown below, the large-scale simulated responses of the two models agree reasonably well. However, a certainlnum- ber of disagreements is also found. The fast turnaround time of CLIMBER-2 permits carrying out further sensitivity studies which help to shed light on the possible reasons for such discrepancies. In thk way, CLIMBER-2 is introduced as a fast but still rather complex model which can be used to extend the knowledge derived from simd.ations with more sophisticated GCMS and to give hints for the direction of further investi- gations. In this line, results of several sensitivity experiments analyzing the relative influence of changes in orbital parameters and in C02 concentration, as well as in- vestigating the response of interactive vegetation to the altered boundary conditions are presented. All integrations were carried out by Claudia Kubatzki at the Potsdam- Institute for Climate Impact Research (PIK), where the global climate system model CLIMBER (for CLIMate and BiosphERe) was developed. As in chapter 4, the fo- cus is on the anomalies of the fields simulated for 125 lsBP relative to a control run representing the modern climate. 51 52 CHAPTER 5. RESULTS WITH THE CLIMBER-2 CLIMATE SYSTEM MODEL 5.1 The CLIMB:ER-2 climate system model CLIMBER-2has a resolution of 10° in latitude and approximately 51° in longitude. It consists of different modules. The atmospheric module is a 2.5-dimensional statistical- dynarnical model which, in contrast to AGCMS, does not resolve individual synoptic scale systems, but rather predicts their statistical characteristics, including the fluxes of heat, moisture and momentum associated with ensembles of such systems, which are parameterized as diffusion terms with a turbulent diffusivity computed from at- mospheric stability and horizontal temperature gradients. However, the large-scale circulation patt ems (e.g. the subtropical jet streams, the Hadley, Ferrel and polar cells, the monsoon and Walke:rcirculation, the tropospheric quasi-stationary planetary waves and the main high- ancl low-pressure areas) are explicitly resolved. The vertical structure of the model, which includes a planetary boundary layer, a free troposphere (including cumulus and stratiform clouds) and a stratosphere, is based on the assump- tion of a universal vertical structure of temperature and of humidity in the atmosphere. Circulation and energy and water fluxes are computed at 10 pressure levels, while long- wave radiation is calculated at 16 levels, accounting for water vapor, ozone, C02 and cloud cover. The time-step is one day. The ocean module, based on the model of Stocker et al. (1992), is zonally aver- aged with three ocean basins (Atlantic, Pacific, and Indian). It has 11 levels in the vertical including an upper mixed layer of 50 m thickness. Its time-step is 10 days. Temperature, salinity, and vertical and meridional velocity are calculated within each basin; at the circumpolar oceans, where the basins are connected, zonal velocities are determined in addition. The vorticity balance and Ekman transport are parametri- zed. A thermodynamic sea-ice model predicting the sea-ice fraction and thickness for each grid box with a simple treatment of advection and diffusion of ice is included. Hence, the ocean model resolves only the vertical circulation (the Ekman cells and the thermohaline flow); zonal gradients are ignored and the heat and salt transports are parameterized as simple diffusion terms. The terrestrial vegetation module is the Vegetation Continuous Description model VECODE (Brovkin et al., 1997). Potential vegetation is presented as a mixture of trees, grass, and desert (bare soil); the fraction of each of the former within a grid cell is a centinuous function of the growing degree days (GDD), the annual precipitation and temperature. Thus, changes in vegetation cover can be interpreted as shifts in vegetation zones smaller than the spatial resolution of the model. The three modules are cormected by the atmosphere-surface interface via the fluxes of energy, momentum, and water. The coupling does not employ any type of flux adjustment. The interface is a,strongly modified version of the Biosphere-Atmosphere Transfer Scheme (BATS) scheme (Dickinson et al., 1986). Each grid box consists of one or several of six surface types (open water, sea ice, trees, grass, bare soil, and glaciers) for which characteristics like temperature or evaporation, as well as the fluxes, are calculated separately as a function of different (prescribed or calculated) surface properties. Soil processes are described within a two-layer soil model. CLIMBER-2 was validated. under preindustrial boundary conditions (present inso- lation and a C02 concentration of 280 ppmv (Ganopolski et al., 1997; Petoukhov et al., 5.2.EXPERIMENTAL SET-UP 53 1999). CLIMBER-2 performs reasonably for modern climate and its C02 sensitivity is similar to that of coupled GCMS (3.0°C for a doubling of C02). Details of the valida- tion are described in Ganopolski et al. (1997) and Petoukhov et al. (1999), where also a more detailed description of CLIMBER-2 can be found. CLIMBER-2 aims only at the simulation of large-scale, time-averaged properties of climate; unlike GCMS it cannot simulate weather, neither many natural variabil- ity modes, nor regional features. However, due to the its low resolution and simplified equations it has a much faster turnaround time than GCMS. CLIMBER-2 has been em- ployed in previous paleoclimate simulations for the Last Glacial Maximum (Gano:polski et al., 1997) and a study of atmosphere-ocean-vegetation feedback in the mid-Holocene (Ganopolski et al., 1998). 5.2 Experimental set-up Two experiments were performed with the coupled atmosphere-ocean version of CLIM- BER-2 with conditions resembling those of the previously described experiments with ECHAM-1/LSG (Table 5.1): a control run representing the present climate (with present insolation and a C02 concentration of 346 ppmv) and an Eemian run represent- ing the climate of the last interglacial maximum, where the incoming solar radiation at the top of the atmosphere was changed by setting the Earth’s orbital parameters to their values at 125 kBP, just as in the runs with ECHAM-1/LSG, and the C02 concen- tration was set to 280 ppmv. In both runs, CLIMBER-2 was integrated to equilibrium starting from the previously mentioned run with preindustrial boundary con~ltions (Section 5.1). C02 Eccentricity Obliquity Angle of perihelion (ppmv) (degrees) (degrees) Control run 346 0.017 23.45 282.16 125 kBP 280 0.040 23.79 127.27 Table 5.1: Boundary conditions for runs with CLIiW3E13-2 The C02 concentration values used in the Eemian and control runs with CLIMBER- 2 are not identical to the corresponding runs with ECHAM-1/LSG. But, since the dependence of radiative forcing from C02 concentration is logarithmic they result in approximately the same C02 anomalous forcing at the Eemian with respect to the control run, and the value of 280 ppmv for the Eemian is consistent with C02 estimated levels (Barnola et al., 1987). 5.3 Temporal evolution Due to the fast turnaround time of CLIMBER-2, integrating this model until quasi- stationary equilibrium is achieved does not pose any problem. After changing the 54 CHAP TER 5. RESULTS WITH THE CLIMBER-2 CLIMATE SYSTEM MODEL external forcing from the preindustrial run to the present control and Eemian runs, equilibrium was attained in all systems (sea-ice, atmosphere and ocean) after several hundred years. To demonstrate this, the time series of global surface-air temperature, sea ice volume in the Northern and Southern Hemispheres and the amplitude of the Atlantic overturning for the control and Eemian run, as simulated by CLIMBER-2, are shown (Figure 5.1). From the results, we can conclude that the change from the preindustrial to present COZ (left) has greater consequences than the change in insolation forcing (right) in the annual mean. b) T c) NH ice d) NH ice ]; :. .. . :’”- ,“.:: :,...:/ .:: .. :::: .: :.: :,,:” ‘- 1:: ,“: ‘j;::’:: ,“:; ;’:’: :“” ;:’; ,’,.”: :“. “:”,, ,. Ii+ .~ II: if: ~ .. :,’: ‘,.”: if: :3W:”6;’.’:: ‘:””:’ :::,.. , ! 900 1200 lio 1800 2;00 “2;0 270; 3W0 300 &o” ’900 ttio “l&” 1“2CU’21W Z+& 27W”” k f) SH ice —.—.—..——__——__——_ j::.:: 1{ ~. .:.::.:.:;..::.. :::.:,:::;::::::.:: a.:.:..”:::’1o”:”””- : .:::..... , 300 300 900 1200 15C4 1803 2?00 2403 2700 3000 g) NADW h) NADW ( :! :;,;;::::::”>::::::: :; I lo.::::;:::::::.:: : :::. :: :.1 . . 1{3:I~: ““”::::”. ;“ ;.’::.:;::;’:,: ,1 15. ..-’. , ...... ’ ...... 1 113”.::iw:6w ..:: :.::.:, :::::.: :.,:.4 900 1200 1590 1800 2100 2400 27oO 3000 Figure 5.1: Time series of the mean annual globally averaged surface air temperature a) in the control run, b) in the Eemian run ~ C) ; Arctic sea ice cover c) in the control run, d) in the Eerr&n run (l& lcm2);Antarctic sea ice cover e) in the control run, f) in the Eemian run (Id km2); amplitude of NADW g] in the control run, h) in the Eemian run (Sv). 5.4 Comparison of mean simulated fields In order to compare their simulated responses to the imposed changes in radiative forcing, differences in the mean fields (Eemian minus control run) as simulated by ECHAM-1/LSG and by CLIMBER-2 averaged over the final 300 and 100 years, re- spectively, are considered. The focus is on the comparison of model results. Results with ECHAM-1/LSG were sm,oothed to obtain a spatial resolution comparable to that of CLIMBER-2 in order to circumvent discrepancies related to this feature. 5..4. COMPARISON OF MEAN SIMULATED FIELDS 55 5.4.1 Near surface temperature The near surface (2m) temperature responses simulated by the two modeIs (Figure 5.2) are very similar both in JJA and DJF, and reflect essentially the response to the sea- sonal insolation perturbation (Figure 3.2), modulated by the different heat capacities of land and ocean. The main common features include: (i) warming in JJA over north- ern land-masses and southern subtropical land-masses (Figure 5.2a~b), and cooling in DJF (Figure 5.2c, d), and in general a weaker response over the ocean; (ii) a reduced meridional temperature gradient in the Northern Hemisphere both in JJA and DJF; (iii) high northern latitude warming through most of the year. The latter is a result of the decreased sea-ice area due to the enhanced summer and annual insolation at these latitudes and the positive sea ice albedo-temperature feedbacks as simulated by both models (Figures 4.6 and 5.3). Nevertheless, the remarkable decrease in the thickness of Northern Hemisphere sea ice as found in ECHAM-1/LSG is not seen in CLIMBER-2; hence, the thermal response at high northern latitudes differs quantitatively. The agreement of the models is best in JJA, (Figure 5.2a, b) which is consistent with the view that summer temperatures are essentially determined by insolation (Hurrel, 1996; Mann and Parkj 1996). The only important discrepancy is found over Antarctica and the Southern Ocean, where CLIMBER-2 simulates a strong cooling by ca. -3° C together with a pronounced increase in sea-ice area (Figure 5.3) which is found to a much lesser degree in ECHAM-1/LSG (where the sea-ice area increase is much weaker, Figure 4.1). Howeverj since this area is a critical one for ECHAM-1/LSG results should be considered with caution. In DJF (Figures 5.2c, d), dynamicd processes play a more important role in deter- mining temperatures, so that disagreement in the thermal responses can be partially attributed to differences in how the two models simulate the latter processes. 13ssen- tially, CLIMBER-2 shows a stronger winter cooling than ECHAM-1/LSG: warming in results with CLIMBER-2 is restricted to high northern latitudes, while in those with ECHAM-1/LSG it extends farther south, especially over the mid northern latitude oceans. In the annual mean (Figure 5.2e, f), both models simulate changes in the near surface temperature below 1°C almost everywhere, except at high northern latitudes which due to the positive sea ice albedo-SST feedback show a strong warming. The discrepancies in the simulation of the Southern Hemisphere sea-ice result in the fact that CLIMBER-2 shows a large cooling at high southern latitudes which is not found in ECHAM-1/LSG. The globally averaged near surface temperature change for CLIMBER- 2 and ECHAM-1/LSG are -0.8 and -0.3°C respectively (Table 5.2). Hence, results with both models indicate that the simulated globally averaged temperature at the 13emian was lower than in the present (Montoya et al., 1998). 5.4.2 Precipitation The precipitation response (Figure 5.4) shows qualitative agreement of the two models in an intensification of the northern summer monsoon (caused by the enhanced land- sea pressure contrast, which strengthens the onshore inflow and moisture transport 56 CHAPTER 5. RESULTS WITH THE CLIMBER-2 CLIMATE SYSTEM MODEL a) T2 JJA ECHAM-1/LSG (C) b) T2 JJA CLIMBER-2 (C) w m m m !4 m . .mI 9 ;~b . t C) T2 DJF ECHAM-1/LSG (C) e) T2 ANN ECHAM- l/LSG (C) fi T2 ANNCLIMBER-2 (C). . .: n ‘1 I .1 m. ,< ,pg~. ::;@::& .,<.;’”+... :“ :. E@ ml . .. .T“”’””q...... % ~... . Figure 5.2: Difference (Eem minus present control run) in mean near-surface temperature as simulated by ECHAM-1/LSG smoothed to CLIMBER-2’s spatial resolution (a, c, e), and by CLIMBER-2 (b, d, f), respectivel~ in JJA (a, b), in DJF (c, d), and in the annual mean (e, f). Isoline spacing: I“C’. 5.J. COMPARISON OF MEAN SIMULATED FIELDS 57 a) Arctic 20 18...... -...’ ~~~~ . ..., ...... ~ 16. * E a o L ~ ~. . . :....’.., a 4. . “.: ...... -.-”. ‘m m MM APR U4’f JUN JUL AUG SW OCT NW DEC 20W Months b) Antarctic 20——————_——._——_—-_—. 18. ~ 16...... * E ~ 12. m o = al .;/,,’ :..,..?. . . . *. ..,<;-..,.<, ..:. { ...... , ., 0. ;oM& FEE MM #PR WY JUN JUL AM SW OCT NW CKC Months Figure 5.3: Mean annual cycle of total a) Arctic and b) Antarctic sea ice cover (poleward of 65°N and 65°S, respectively) for the control (dashed line) and Eemian run (solid line,)(l@ kn+’). GLOBAL NHL NH SH SHL N-AFRL ECHAM-1/LSGl -0.3 -0.3(2.2) -0.1(1.1) -0.5 (0.3) -0.19 (-0.05) 0.17 (0.88) CLIMBER-21 -0.8 -0.7 (2.9) -0.6 (1.5) -1.0 (-0.4) -0.18 (0.08) 0.38 (1.33) A - CL2 0.4 (4.5) 0.3 (2.6) 0.1 (0.9) -0.07 (0.18) 0.46 (1.59) AO - CL2 0.4 0.6 (4.3) 0.5 (2.6) 0.3 (0.9) -0.03 (0.17) 0.46 (1.56) AV-CL2 0.8 (5.2) 0.5 (3.0) 0.1 (0.9) -0.07 (0.18) 1.09 (3.14) AOV-CL2 1.1 1.7 (5.7) 1.4 (3.6) 0.8 (1.6) 0.04 (0.22) E 1.22 (3.40) Table 5.2: Annually (JJA) averaged temperature (in “C) and precipitation (in mm day-l) over land, and land plus ocean for selected areas (AH Northern Hemisphere; SH: Southern Hemisphere; N-AFl?: North Africa between 10 and 3@’N; subscript L indicates averages over land only). ll?emian - present ; 2Eemian - preindustrial CLIMBER-2 (C02 = 280 ppmv) 58 CHAPTER 5. RESULTS WITH THE CLIMBER-2 CLIMATE SYSTEM MODEL into land, section 4.2) manifested in an increase of precipitation over North Africa and South Asia in JJA (Figure 5.4a, b), and an attenuation of the summer monsoon circulation in the Southem Hemisphere, reflected in an increase of precipitation over the Indian Ocean and a decrease over South America, South Africa, and Australia in DJF (Figure 5.4c, d). However, precipitation changes simulated by ECHAM-1/LSG have generally a larger amplitude and a smaller spatial scale than in CLIMBER-2. a) P JJA ECHAM-1 /LSG (mm/day) c) P DJF ECHAM- 1/LSG (mm/day) J ) , :~ -“. — I I e) P ANN ECHAM-1/LSG (mm/day) a Figure 5.4: Difference (Eem minus present control run) in mean precipitation as simulated by ECHAM-3/LSG smoothed to CLIMBER-2’s spatial resolution (a, c, e), and by CLIMBER- 2 (b, d, f), respectively, in JJA (a, b), in DJE’ (c, d), and in the annual mean (’e, f). Isolines at +15, +10, +5, +2, +1, +0.5, +0.1 mm day–l. Zonally averaged seasonal changes in precipitation over land (Figure 5.5) show a very satisfactory agreement: i.n JJA, both models show an increase in the subtropics reflecting the aforementioned intensification of the summer monsoon, while the op- posite occurs in DJF. Both mlodels simulate slight decreases of the globally averaged precipitation of -0.05 mm day-l and over land (-0.03 mm day-l in ECHAM-1/LSG 5.4. COMPARISON OF MEAN SIMULATED FIELDS 59 and -0.01 mm day-l in CLIMBER-2). a) JJA 1.5J 1.2. 0.9- > 0.6- : 0.3- ~------\ 0- . .-. .—- — E -0.3. v E -0.6- -0.9- -1.2. 66s 30s <0 3M 66N b) DJF 1.5+ 1.2. 0.9. ~ 0.6- -0 0.3. ,.-.. — \ o- . E -o.3- E -0.6. -0.9. -1.2. -1.5? 60s 30s El 313N Em c) Annual Mean $ 0.8- -O 0.3- E -0.3. E -0.G. -0.9 4 I -1.24 I -1.5, Figure 5.5: Mean difference @emian minus control) in zonally averagedprecipitaticm over land in JJA, in DJF, and in the annual mean (bottom) as simulated by ECHAM-1/LSG (solid line) and CLIMBER-I (dashed line) (mm day-l). 5.4.3 Atmospheric Circulation Changes in the zonally averaged zonal wind velocity simulated by the two rnodeis (Figures 4.7 and 5.6) agree by and large in JJA, whereas they show a larger degree of disagreement in DJF, mainly in the Northem Hemisphere. Both models show an intensification of the tropical easterly jet in JJA related to the intensification of the southwest summer summer monsoon circulation (see section 5.4.2), and a decrease in the intensity of the zonally averaged westerly flow at all levels, both in DJF and JJA, as a result of the decrease in the meridional temperature gradient at all levels throughout the troposphere (section 4.2). However, as was already mentioned (Section 5.4.1), in DJF CLIMBER-2 shows a warming only at high northern latitude, sea- ice covered regions, whereas to the south a temperature decrease of about 2°C is found. Thus, the greatest temperature gradient decrease in DJF and hence the most pronounced changes in the zonal circulation take place at about 60”N. In ECHAM- l/LSG, where the Northern Hemisphere temperature changes more gradually from north to south, changes in the zonally averaged atmospheric circulation are nc)t only restricted to the high latitudes and also reveal a northward shift of the westerlies at 60 CHAPTER 5. RESULTS WITH THE CLIMBER-2 CLIMATE SYSTEM MODEL mid-latitudes. Hence, this discrepancy may point to a inability of CLIMBER-2 to resolve the meridional temperature change simulated in ECHAM- 1/LSG due to its coarse resolution. 0.5 0.9 6C.5 30s m .3a)“:em SON ‘& Figure 5.6: Mean difference (Eemian minus control) in zonally averaged zonal velocity for a) JJA and b) DJF, as simulated by CLIMBER-2 (m .s–1,. 5.4.4 Mean ocean circulation Before analyzing the simulated response of the thermohaline circulation, it is worth mentioning that the differences in the strength of the mean overturning circulations simulated in the corresponding control runs by EC HAM-1 /LSG and CLIMBER-1 are considerably large (Figures 9.6 (Appendix) and 5.7). The response of the merid- ional overturning circulation simulated by ECHAM- l/LSG and CLIMBER-2 in each ocean basin differs in several ways. In the Atlantic Ocean, ECHAM-1 /LSG shows a slight intensification of about 2 Sv, i.e. by less than 6% (Figures 4.11 bottom), while CLIMBER-2 shows a weakening up to 3.5 Sv, i. e. by 20% (Figure 5.7c). The response simulated by CLIMBER-2 can be explained by the enhanced high northern latitude warming and freshwater flux into the North Atlantic, resulting both in an increase of surface stability and a decrease of the ocean meridional density gradient, which reg- ulates the strength of the overturning (Rahmstorf, 1996). In the Pacific Ocean while CLIMBER-2 shows essentially no differences (Figure 5.7d), in ECHAM-1/LSG the meridional overturning circulation is intensified, and there is an anomalous intermedi- ate flow of up to 2 Sv triggered by convection at high northern latitudes (see Section 4.3, Figure 4.11). As was mentioned in section 4.3, the decrease in freshwater flux into the North Pacific as simulated by ECHAM-1/LSG is partly a consequence of changes 5.5. SENSITIVITY EXPERIMENTS 61 in the monsoon circulation, and partly of an attenuation of the storm-tracks, which are not explicitly resolved in CLIMBER-2: while ECHAM-1/LSG shows a decrease of precipitation over the North Pacific of ca. 0.5 mm day–l, the corresponding decrease simulated by CLIMBER-2 is just 0.1 mm day–l. a) Atlantic Control (Sv) o b) Pacific Control (Sv). . ml 1 d) Pacific EL-CH (Sv) o- m-,. - ~~~~ == h7 - *W Ixa /“” ~= ~zx ..’... ~x.xb \ ‘J I 1[1\ I w’““ ““””’””””’ i Figure .5.7: Zonally averaged mass transport stream function of the meridional overturning circulation as simulated by CLIMBER-2 in the Atlantic Ocean and Pacific Ocean in the control run (a, b), and for .Eemminus the control run with present COZ concentration (c, d) (sV). The different simuIated oceanic responses have different consequences on mid-high latitude temperatures. The changes in oceanic circulation simulated by ECHAM- l/LSG, mainly in the Pacific, contribute to an increase of the northward global heat transport by 10% which affects temperatures at mid-high northern latitudes and might contribute to the slightly enhanced midlatitude warming in DJF (Section 4.3). In con- trast, the weakened Atlantic overturning in CLIMBER-2 results in the fact that north- ward heat transport is reduced by more than 10’ZOup to 60”N (Figure 5.8a). This is not balanced by changes in the meridional heat transport in the other oceans (Figure 5.8b, c) neither in the atmosphere (Figure 5.8d) and thus contributes together with decreased atmospheric heat transport to the north to the mid-latitude cooling in DJF as observed in this model. 5.5 Sensitivity experiments Thefast turnaround time of CLIMBER-2 in comparison with GCMS permits to per- form a number of sensitivity experiments in order to explain the possible reasons for 62 CHAPTER 5. RESULTS WITH THE CLIMBER-2 CLIMATE SYSTEM MODEL a) Atlantic ’11~ 30S EQ W 60N b) Pacific 41EE5EI60S 30S EQ W 6.3N c) Indian Figure 5.8: Mean difference (Eemian minus control) in northward heat transport a) in the Atlantic Ocean, b] in the Pacific Ocean, c) in the Indian Ocean (bottom] d) by the atmosphere, as simulated by CLIMBER-2 (PW). the discrepancies between the two model responses. First, in order to explain the reasons and analyze the consequences of the different responses of the thermohaline circulation, CLIMBER-2 was forced with the freshwater flux anomalies obtained in ECHAM-1/LSG (section 5.5.1). Secondly, the effects of altered insolation and C02 were analyzed separately (Section 5.5.2). Finally, CLIMBER-2 was used to study the effects of interactive changes in vegetation, ignored in the previous runs (Section 5.5.3). 5.5.1 Prescribed freshwater flux run Since the main difference between the model responses is that of the thermohaline circulation, which is ultimately cbiven by temperature and salinity, and given that the temperature response from bckh models at high northern latitudes reflects a warming rather than a cooling throughout the year, the discrepancies in the thermohaline cir- culation must be due to differences in the simulated freshwater fluxes. To test this hypothesis, an additional run with CLIMBER-2 was performed in which Eemian ex- ternal forcing was imposed but the freshwater flux was prescribed as the sum of the mean fluxes from the CLIMBER-2 control simulation plus the mean freshwater flux anomalies as simulated by ECHAM- l/LSG for the Eemian run, relative to the control run. Hence, the freshwater flux was constant in time except for the seasonal cycle. All other ocean characteristics were calculated interactively. In the Atlantic Ocean the overturning circulation is intensified in comparison with 5.5. SENSITIVITY EXPERIMENTS 63 the previous Eemian run (Figure 4.13a), as surface salinity is increased due to the imposed ECHAM-1/ LSG freshwater flux differences, which demonstrates the potential of this forcing. However, it is still weaker than for the control run (Figure 5.7a), at variance with the response simulated by ECHAM- l/LSG. The decreased freshwater flux into the North Pacific as imposed from the ECHAM-1/LSG differences (Figure 4.13) results in a similar anomalous circulation pattern (Figure 5.9b) to that found in ECHAM-1/LSG (see section 5.4.4, Figure 4.11): the Pacific meridional overturning circulation is intensified in comparison with the control run, showing an anomalous intermediate flow of about 3 Sv. At the same time AABW is stronger and penetrates farther north. Differences near the Antarctic are difficult to interpret as they seem to be affected by the deficiencies in the sea-ice simulation of the Southern Hemisphere within ECHAM-1/LSG (section 4.2). Matching all this, within CLIMBER-2 the main differences in the ocean merid- ional heat transport due to the fixing of the freshwater fluxes from the atmosphere can be observed in the Pacific (Figure 5.10), where the increased northward heat trans- port might contribute to the slight warming of some parts of the middle and higher northern latitudes, bringing the temperature response of CLIMBER in these regions slightly closer to that of ECHAM-1/LSG: the difference in near-surface temperature with respect to the previous Eem simulation shows a warming by 0.2°C (Figure 5.11). 5.5.2 Radiative forcing The sole effect of insolation on the coupled atmosphere-ocean system have been in- vestigated by comparing the differences between the Eemian run and a control with equal C02 concentration of 280 ppmv (hereafter, AO). This is more than an academic exercise, since the present-day climate is not in equilibrium with the currently increas- ing C02 levels and paleodata reflect mean values over at least several hundred years. Thus, for comparison with the paleodata, it is more reasonable to use preindustrial conditions, for which the assumption of such an equilibrium is more valid, as control simulation instead of the former control run with the higher C02 level. Temperature differences over the Northern Hemisphere continents reach values of +6°C in summer, whereas winter time cooling does not exceed -2°C (Figure 5.12a, b, to compare with Figures 5.2c, d). Harrison et al. (1995) find in their simulations with a GCM coupled to a mixed-layer ocean summer warming of up to 8°C and winter cooling of about 2°C (but warming for western Eurasia and eastern North America) over the Northern Hemi- sphere continents. Although patterns over the ocean differ, both models show at high northern latitudes a very pronounced warming over the whole year. In the Southern Hemisphere, JJA temperatures now are higher for the Eemian than for the preindus- trial climate (Figure 5.12a), in accordance with the differences in insolation (Figure 3.2). The annual mean temperature differences show a warming for the Eemian over nearly the whole globe (Figure 5.12c), with a globally averaged change of 0.4”C. Annual precipitation is higher than for the preindustrial climate in most areas, especially in the regions of increased summer monsoons in north Africa, south-west Asia, and east Asia (Figure 5.13a), and global annual precipitation increase is 0.12 mm day-]. The Atlantic thermohaline circulation is even weaker and shallower at the Eemian, while 64 CHAP’TER 5. RESULTS WITH THE CLIMBER-2 CLIMATE SYSTEM MODEL a) Atlantic FRESH-CH (Sv) b) Pacific FRESH–CH (SV) l“~~“’‘;lwhi’11 d)–, Pacific--- - EL–CL—— —- (Sv)--- 0= 1 — 1. :1: ‘“mRJ~ e) Atlantic AOV-CL (Sv) f) Pacific AOV-CL (Sv) 0 D ?“~=---.__?: — I xQ - lm. - ? - ~~ ,% I am 0 $.’ .?’ . :: “~ I l.g ; \~ Om W3 m m ml Figure 5.9: As in Figure 5.7 but for: Eem with prescribed freshwater flux minus the control run with preindustrial C02 concentration (a, b); Eem simulated with the atmosphere-ocean- vegetation version minus the control run with preindustrial C02 concentration (c, d); Eem simulated with the atmosphere- vegetation version (e, f) (Sv). 5.5. SENSITIVITY EXPERIMENTS 65 k AILANTIC MHF & ATLANTIC UHF 0.1 o --:--:-:~~ :“ -O,f . . . . .;-:...... -,...... ~~~~: ; :. :. I ? 90S 60S 30S EO 30N 60N 90N 90S 60S 30S EO 3DN 60N 90N m& INDIC MHF & lNDIC UHF (Jef ...... :. . I ‘t+-+-+ 0 ‘- - : : +.~...... : ..:...... I +,f ...... :. \ I ,.. 1 :.. .: 90S 60S 30S EO 30N 60N 90N 90S 60S 30S EO 30N 60N 90N E PACIFIC MHF ~ PACIFIC MHF , m o+~‘1--?iKw...... ~.1,m...... -“’w90S 60S 30S EO 30N WN 90N Figure 5.10: Northward heat transport response simulated by CLIMBER-2 in the Atlantic, Indian and Pacific Oceans for the experiment with prescribed freshwater flux with respect to the control run (left) and with respect to the Eemian run (right) (PW). 66 CHAPTER 5. RESULTS WITH THE CLIMBER-2 CLIMATE SYSTEM MODEL a) T2 JJA FRESH-C346 (C) m- M! C) T2 ANN FRESH-C346 (C) . . EO 3s a , Figure 5.11: Differencein mean near-surface temperature a) in JJA, b,) DJF, and c) the annual mean as simulated by CLIMBER-2 for Eem with prescribed freshwater flux minus present control run. Isoline spacing: 1° C. AABW is stronger and penetrates farther north (Figure 5.9c). 5.5.3 The role of interactive vegetation Up to now, vegetation characteristics in all simulations have been fixed to their modern state, which is not a realistic assumption since vegetation will also change under altered climatic conditions, with the potent ial for further tiecting the climate response. To investigate the role of vegetation during the Eemian, a simulation AOV in which both the ocean and the vegetation were allowed to react to the changed orbital forcing at the Eemian was performed, and its results compared against those of the preindus- trial control run. In addition, in order to disentangle the roles of atmosphere, ocean and vegetation, two more experiments were carried out: a simulation with the cou- 5.5. SENSITIVITY EXPERIMENTS 67 a) T2 JJA EEM–CL (C) . X# m m v e6’ ...... m b) T2 DJF EEM-CL (C) C) T2 ANN EEM–CL (C) . ‘~~~~~<“”’””””””’l””fir’”’””” Figure 5.12: Difference in mean near-surface temperature in a) JJA, b) DJF, and c) the annual mean as simulated by CLIMBER-2, for Eem minus preindustrial control run. Isoline spacing: 10C. pled atmosphere-vegetation version of CLIMBER-2 (AV) with ocean characteristics prescribed to their modern values, whereas vegetation was computed within the terres- trial vegetation model, and a simulation with the atmosphere-only model (A) in which both ocean as well as vegetation characteristics were fixed to their modern values. AOV, AV and A all refer to anomalies relative to the preindustrial (COZ=280 ppmv) control run. The role of interactive vegetation can be analyzed by inspecting the results of simu- lation AOV. The main differences concerning the thermal response can be observed over the mid and especially high latitude continents of the Northern Hemisphere (Figures !5.14, 5.22): because of warmer summers and longer growing seasons during Eemian, the forests in AOV expand farther northward at the expense of tundra (which means grass in CLIMBER-2). The rougher surface in AOV leads, due to its lower albedo, to 68 CHAPrER 5. RESULTS WITH THE CLIMBER-2 CLIMATE SYSTEM MODEL a) P JJA EEM-CL (mm/dav).,, , m m w s MS wk. 6m o a ,Za m ,m~ 7 ,m . 0 Rf ,= ,- c) P ANN CLIMBER-2 (mm/dav).,. , al m m ‘1 “’SW’””’““”’”””’u ~ -=+ Figure .5.13: Difference in mean precipitation in a) JJA, b) L?JF, and c) the annual mean as simulated by CLIMBER-2 for Eem minus present control run. I.solines at A15, &lO, +5, &2, Al, +0.5, AO.1 mm day–l. additional warming and thus 10 earlier snow melting which again contributes to an ad- ditional temperature increase, which in turn favors forest expansion (vegetation-snow- albedo feedback). In turn, the warmer conditions also lead to a significant decrease of the sea-ice area which due to the lower albedo of open water leads to additional warming (sea ice-albedo feedback). Thus, both the vegetation-snow-albedo feedback and the sea ice-albedo feedback in this region act together in the same direction, lead- ing to a strong amplification of the individual warming signals (Table 5.2). Altogether, the area of tundra and polar desert is reduced by some 5.3 x 106 kmz while the area covered by forests is enhanced! (Figures 5.15, 5.16a). In addition, parts of the Southern Hemisphere to the south of 30°S are warmer than in the modern climate cwer the whole year (Figure 5.14c). This reflects partly the fact that this region has a net gain in annual insolation in comparison to today; 5.5. SENSITIVITY EXPERIMENTS 69 a) T2 JJA AOV-CL (C) b) T2 DJF AOV-CL (C) Figure 5.14: Difference in mean new-surface temperature in a] JJA, b) DJF, and c) the annual mean as simulated by CLIMBER-2 for Eem with the atmosphere-ocean-vegetation version minus preindustrial control run. I.solinespacing: 1° C. as a result, the amount of sea ice decreases leading to some additional temperature feedback. As this is also valid for AO (section 5.5.2) in which warming was found only in austral winter and spring, there must be other mechanisms resulting from interactive vegetation: in AOV a further weakening of the meridional overturning circulation in the Atlantic Ocean by up to 2 Sv (7 Sv with respect to the preindustrial control run, Figure 5.9e) is found in comparison with simulation AO and less heat is transported into the North Atlantic, warming the Southern Oceans. A warming effect of the reduced NADW on the Southern Hemisphere during the last interglacial was already discussed by Crowley and Kim (1992) although the inferred strength of thk mechanism might be different. The main differences in the precipitation response can be observed in the subtropics, especially in North Africa (Figure 5.17 and Table 5.2) where the increase of summer 70 CHAPTER 5. RESULTS WITH THE CLIMBER-2 CLIMATE SYSTEM MODEL Figure 5.15: DifferencesbetweentheEemianclimatesimulationwith the atmosphere-cacean- vegetation model AOV and the preindustrial control run in the fraction of area covered by forest and by desert. monsoon precipitation during the Eemian leads to an expansion of vegetation in AOV. Subsequent changes in the radiation balance lead to a surplus in available energy which is used for evaporation. The increase of atmospheric total energy (sensible plus latent heat) results in enhanced convection (Charney et al., 1976) and enhanced precipitation favoring in turn the growth of vegetation (Claussen, 1997; Claussen and Gayler, 1997; Kubatzki and Claussen, 1998), As a result, 97% of North Africa is covered by vegetation (Figures 5.15 and 5.16). The increase of annual precipitation in comparison with the modern climate in the interactive vegetation runs is about two times stronger than in the fixed vegetation runs (Table 5.2), suggesting that the inclusion of interactive vegetation is a crucial factor determining the strength of the North African summer monsoon. In all, results of simulation AOV show strong synergistic effects in the coupled atmosphere-ocean-vegetation system. Interaction between these three climate subsys- tems results in a pronounced additional warming (Table 5.2). In comparison with the modern climate, summer temperature differences over the Northern Hemisphere conti- nents rise to values of up to 7“C, in winter differences now do not exceed - 1.5°C. Thus in the former simulations, only changes in vegetation might provide the mechanism necessary to account for overall warmer than present conditions at the last interglacial. 5.6.FEEDBACK ANALYSIS 71 0.8 0.6 0,4 W2 II 1 0.8 0,6 0.4 0.2 #k *I!II , 0 I 30S EQ 30?4 60N Figure 5.16: Zonally averaged trees (a) and desert (b) fraction for the CLIMBER-2 con- trol run (grey), and the Eemian climate simulations with the atmosphere-vegetation model AV (white) and with the atmosphere-ocean-vegetation model AOV (black] (fraction). All simulations run with preindustrial C02 concentration. 5.6 Feedback Analysis To assess the contribution of the different climate sub-systems and of their interaction in the total response of the coupled atmosphere- ocean-vegetation system, a feedback analysis has been performed. This analysis shall show to which extent the processes in each of the climate sub-systems and the interactions among them (synergisms) amplify or damp the response of the atmosphere-only model when changing from modern to Eemian conditions. The approach followed herein an extension of the linear feedback analysis presented by Peixoto and Oort (1992). We assume that the climate state, which is defined in terms of a series of variables S, depends on the external forcing E such as insolation, C02, etc. and internal processes Hi: S = S (E, Hi). A change in the external forcing AE will translate into a change in the state of the system AS. Hence, AS = G AE~where G is a sensitivity factor generally referred to as gain. Without any feedbacks, the response of the system would simply be ASO = Go AE. With feedbacks, however, implicit in the response AS of the climate system are internal mechanisms Hi which are triggered by AS and whose response ~fll Hi is fed into the system so that: 72 CHAPTER 5. RESULTS WITH THE CLIMBER-2 CLIMATE SYSTEM MODEL a) P JJA AOV-CL [mm /dtw) . 1- ,CdE ,Za . 0 !& ** m Figure 5.17: Difference in mean precipitation in a) JJA, b) DJF and c) the annual mean as simulated by CLIMBER-2 fclr Eem with the atmosphere-ocean- vegetat ion version minus preindustrial control run, Iso]ines at &15, +10, &5, *2, +1, +0.5, +0.1 mm day–l. AS= GOAE + Go ~ Hi As, (5.1) ()i=l where each product fi = Gro Hi is called feedback. Hence: AS= GOAE + ~ fi AS. (5.2) ()‘i=1 This analysis is based on the assumption that there are no synergisms, i.e. inter- action among feedbacks. To extend this analysis to the case in which synergisms do exist one may write: 5.6.FEEDBACK ANALYSIS 73 a) T2 JJA AV–CL (C) . m m s ...... ,.,,, 1 b) T2 DJF AV-CL (C)., I n I 4 I Figure 5.18: Difference in mean near-surface temperature a) JJA, b) DJF, and c) the annual mean as simulated by CLIMBER-2 for Eem with the atmosphere-vegetation version minus preindustrial control run. I.soline spacing: 1° C. NN AS= GOAE + ~ fi AS+ ~~f ~j) AS+ .... (5.3) (;=1 ‘i=l j=l where fij represents the synergism between feedbacks Hi, Hj. Positive feedbacks and synergisms amplify the response while negative ones damp it. The feedbacks and synergisms involved in the atmosphere-ocean-vegetation system at the Eemian can be calculated by making use of the results of the four CLIMBER-2 simulations A, AO ~AVj and AOV, if we assume that the only significant synergism is that which occurs between the ocean and the vegetation. As an example, let AT be the response of the mean annual globally averaged temperature at an Eemian run performed with the atmosphere only version of CLIMBER-2 (A) with respect to the 74 CHAPrER 5. RESULTS WITH THE CLIMBER-2 CLIMATE SYSTEM MODEL O) T2 JJA A-CL (C) . . m Qs =5 ,m lm . 0 m $= b} T2 DJF A–CL (C) Figure 5.19: Difference in mean near-surface temperature in a) JJA, b) DJI?, and c) the annual mean as simulated by CLIMBER-2 for Eem with the atmosphere version minus prein- dustrial control run. Isoline spacing: 1° C. preindustrial control run (no feedbacks). Then: AT = ATA, (5.4) which is known from experiment A. We now perform a second simulation AO with the atmosphere-ocean version of CLIMBER-2. The global mean temperature response is then: ATAO = ATA + foATAo, (5.5) where f., representing the temperature feedback due to the internal process char- acterizing the the atmosphere-ocean interaction, can be calculated from the known quantities ATA, ATAO: 5.6. FEEDBACK ANALYSIS 75 . m lm m e m ,Zm ,m b) P AV-CL CLIMBER-2 (mm/day) J n a., –. 1 1 m Figure 5.20: Difference in mean precipitation in a) JJA, b) DJF’, and c) the annual mean as simulated by CLIMBER-2 for Eem with the atmosphere-vegetation version minus prein- dustrial control run. Isolines at +15, +10, +5, &2, *1, +0.5, &O.1 mm day–l. f. = I – ATA/ATAo (5.6) Analogously, if instead only the atmosphere vegetation (AV) version of CLIM13ER-2 is employed: ATAV = ATA + fvAT~v, (5.7) and .fv= 1 – ATv/ATAv. (5.8) f. and fA calculated in this way are given in Table 5.3, for both temperature and precipitation. Both feedbacks have the effect of enhancing Northern Hemisphere warm- 76 CHAPTER 5. RESULTS WITH THE CLIMBER-2 CLIMATE SYSTEM MODEL a) P JJA A-CL (mm/day) m m b) P A–CL CLIMBER–2 (mm/day) ~— n I ,*r?:lam . . me ~,?4 C) P ANN CLIMBER–2 (mm/day] Figure 5.21: Difference in mean precipitation in a) JJA, b), DJF, and c) the annuaJmean as simulated by CLIMBER-2 for Eem with the atmosphere version minus preindustrial control run. Isolines at +15, +10, +5, +2, +1, +0.5, +0.1 mm day–l. ing. The ocean contribution is mainly due to its capacity to store heat, affecting SSTS through the positive sea ice a3bedo-SST feedback. The contribution of vegetation is greatest at mid-high northern latitudes land-masses, where tundra and polar desert are replaced by taiga, and hence contribute to reduce the surface albedo and over North Africa, where it plays a major role in the monsoonal response as shown in section 5.5.3. If no synergisms existed between the ocean and the vegetation, then the global mean temperature response would be given by: ATAOV = AT~ + foATAov+ fvA1’Aov (5.9) Table 5.3 shows the solution of this equation (as well as those for precipitation), labeled AS. However, these results do not equal the corresponding output-values of experiment AOV (AS*), due to the existence of a synergism between the ocean and 5.7.SUMMARY 77 Table 5.3: Feedbacks and synergisms due to ocean and vegetation involved in annually averaged temperature (in 0C) and precipitation (in mm day– 1) changes at the Eemian over land plus ocean for selected areas (AW: Northern Hemisphere; SH: Southern Hemisphere; N-AFR: North Africa between 10 and 30”N; L indicates averagesover land only). NHL NH SH NHL SHL N-AFRL fcl 0.33 0.40 0.67 0. -1.33 0 f~ 0.50 0.40 0. 0.42 0. 0.58 fo~ -0.07 -0.01 0.21 0.11 4.08 0.05 AS 2.4 1.5 0.3 0.33 -0.03 1.09 A S* 1.7 1.4 0.8 0.41 0.04 1.22 the vegetation systems and other unverified assumptions in the analysis scheme. To account for this we write: ATAOV = ATA -I-foATAov + fVATAOV + fOVATAOV. (5.10) from where fov can be calculated as: fov = ~ – ATA/ATAov – f. – fv. (5.11) Results show although vegetation and ocean feedbacks in the Northern Hemisphere are both positive, their interaction in this region represents a negative synergism. The reason for this is the further weakening of the meridional overturning circulation in the Atlantic Ocean in AOV (Figure 5.9e), which translates into a reduction of the northward heat transport. Hence, although the net effect of ocean, atmosphere and vegetation over northern land is a warming, the ocean-vegetation synergism partially damps the response respect to that which would be expected only from the sums of the contributions of the ocean and vegetation feedbacks, with no interaction between them. These results, mainly concerning temperature, vary from those obtained for the mid-Holocene (Ganopolski et al., 1998): while at the Eemian, the ocean introduces a positive feedback over northern landmasses in the mean annual, for the Holocene it represents a negative feedback. In addition, the ocean-vegetation synergism in the Northern Hemisphere is negative at the Eemian while positive at the mid-Holocene; the signs are reversed for the Southern Hemisphere. 5.7 Summary A second climate model has been employed to simulate climatic conditions at the LIG maximum: the climate system model of intermediate complexity CLIMBER-2. Com- parison of the results with ECHAM-1/LSG and CLIMBER-2 reveals abroad agreement in most large-scale features (warming over northern land-masses in summer anti cool- ing in tinter, intensified northern summer monsoons, an all-year-round warming at 78 CHAPTER 5. RESULTS WITH THE CLIMBER-2 CLIMATE SYSTEM MODEL 6 5 /? 4 / 0-.,... “\ ... 3 /.. ~ ‘:L <:., 2 >.<, ,-.. >= m- ti~--” 1 . . ..._. _ ~_ ..., o -i LY ,A -6!1 -N o xi 60 7 6 5 4 3 ...... 2 .. 1 .. 0 . . ..j> ----- ,,, ,., _ ,,+ ,- & .’+ — -t . _,-— L- .-4 -611 -30 0 30 w Figure 5.22: Zonally averaged temperature differenceswith respect to the preindustrial control run) for JJA (upper panel) and DJF (lower panel) for the four runs with CLIMBER- 2: Eem with the atmosphere-ocean-vegetation version minuspreindustrial control run (solid line]; Eem minus preindustrial control run (dotted line); Eem simulated with the atmosphere- vegetation version minus preindustrial control run (dashed line); Eem simulated with the atmosphere only version minus preindustrial control run (dashed-dotted), ~ C). high northern latitudes, and reduced midlatitude westerly flow throughout the tropo- sphere), but certain disagreements are also found. Essentially, the models differ in two points, namely: the sea-ice response in the Southern Hemisphere, and the response of the ocean circulation to the altered boundary conditions. The first of these fea- tures is probably due to the cleficiencies in the sea-ice model in ECHAM-1/LSG, and is reflected in the underestimated sea-ice thickness and extent in the control run in ECHAM-1/LSG, which leads to a smaller increase of sea-ice at high southern latitudes and therefore higher mean annual temperatures than in CLIMBER-2. With respect to the ocean response, in CLIMBER-2 the meridional overturning circulation is at- tenuated in the Atlantic and is not altered in the Pacific while in ECHAM-1/LSG it is intensified both in the Atlantic and Pacific. This translates into differences in the northward heat transport andl hence in discrepancies concerning the thermal response of the climate system and the atmospheric circulation, especially at mid-high northern latitudes. The disagreement in the Pacific Ocean can be attributed to different fresh- water flux anomalies simulated by the models. The freshwater fluxes are ultimately dependent on the parameterization of evaporation, precipitation and the runoff scheme employed. This study demonstrates the importance of the accurate simulation of such processes due to their potential for affecting the global climatic response. The decrease in freshwater flux into the North Pacific as simulated by ECHAM-1/LSG is partly a consequence of changes in the monsoon circulation, and partly of an weakening of the storm-tracks, which are not explicitly resolved in CLIMBER-2. The previous results refer to deviations with respect to a control run with present 5.7.SUMMARY 79 C02 levels. However, due to the rapid increase of the atmospheric COZ concentration in the past hundred years, the present climate is not in equilibrium with such high levels, and paleodata rather reflect deviations with respect to the last several hundred years. Thus, for the comparison against paleodata it is more appropriate to study the deviations with respect to a control run with preindustrial C02 concentration. Anomalous fields simulated In CLIMBER-2 show a more satisfactory agreement with the evidence from paleodata than the aforementioned Eemian run relative to a control run with present C02 levels. Whereas comparison with the present-day C02 control run resulted in an annual cooling at mid latitudes of the Northern Hemisphere for the Eemian, comparison with the modern C02 control run reveals a warming, as is also suggested by geological data (van der Hammen et al., 1971; Miller et al., 1983; Pens et al., 1992). On the other hand, such a warming was also found in the ECHAM-1 /LSG model, in part as a result of the aforementioned changes in the ocean circulation. Sea- sonal patterns of the temperature changes as indicated by the paleodata vary regionally. In addition, comparison with present-day C02 control run showed a slight decrease of total precipitation for the Eemian, climate is generally wetter when comparing with the modern C02 control run: mainly in the Northern Hemisphere total precipitation and precipitation over land are slightly enhanced, in better agreement with the assumption that enhanced summer monsoon precipitation at time periods of maximum precession translates into enhanced total precipitation over land in the annual mean, and with geological evidence suggesting enhanced carbon storage on land at time periods of maximum precession, as would be the case for the Eemian (Keigwin and Boyle, 1985). In addition to a further decrease in NADW formation, comparison with the m~odern C02 control run also results in increased AABW. The CLIMBER-2 simulated change in global annual temperature is only about O.#C. Nevertheless, significant additional warming is obtained through the inclusion of interactive vegetation. Firstly, at high northern latitudes warming due to tlhe sea ice-albedo feedback is strongly amplified through the vegetation-snow-albedo feedback. As a result, the sea-ice area further decreases and the area of boreal forests increases. Second, in the sub-tropics extended vegetation strongly amplifies the monsoonal re- sponse to the insolation forcing. The model shows large parts of the modern Sahara covered by vegetation. In all, the response of the atmosphere-ocean-vegetation sys- tem to the changes in insolation results in summer temperature differences over the Northern Hemisphere continents which rise to values of up to 70C, while in winter the differences do not exceed -1 .5°C. Strong synergisms are found to operate between the ocean and vegetation systems. In particular, their interaction in this region represents a negative synergism. The reason for this is the further weakening of the meridional overturning circulation in the Atlantic Ocean caused by the enhanced surface stability due to the large warming at high northern latitudes, which translates into a reduction of the northward heat transport causing a further warming of the Southern Hemisphere. Altogether, the change in globally averaged annual temperature in this case is l..l°C. CHAPTEIR 6 Comparison against reconstructed SSTS Model-data comparisons for the LIG are hindered by the fact that a limited number of quantitative reconstructions exist. One of the most complete ones is the SST re- construction by the CLIMAP Project Members (1984), against which we have tested against the SST response simulated both by ECHAM-1/LSG and CLIMBER-2 at the LIG. 6.1 CLIMAP SSTS Underlying the methodology employed in the SST reconstruction is the high degree of differentiation of organisms according to their physical environment. In this particular case, the reconstruction is based on the relationship existing between summer and winter SSTS at a given location and the biot a (planktonic foraminifera, radiolaria, or coccoliths) population at that site, which is expressed mathematically by means of a transfer function. The empirically derived transfer function is calibrated against the present observed SSTS by making use of the biota population at the top of the core, which roughly represents mean present conditions. By applying the transfer functions to the census counts of a given biota component along the depth of the sedliment core, seasonal SSTS can be reconstructed throughout thousand of years back in time. Although the CLIMAP SST fields for the LGM have been challenged on a number of grounds (Guilderson et al., 1994), the general similarity of the LIG and the present biota suggests that the transfer function technique on which the reconstruction is based can be applied with more confidence to the LIG. 81 82 CHAPTER 6. COMPARISON AGAINST RECONSTRUCTED SSTS 6.2 Model-data intercomparison For the sake of simplicity, the model-data intercomparison is limited to the best-guess obtained with each of the models, namely: the response simulated by ECHAM- 1/LSG and the SST response simulated by CLIMBER-2 in its ocean-atmosphere-vegetation, relative to preindustrial C02 against CLIMAP reconstructed SST anomalies with their errors. As done by Broccoli a]ndMarciniak (1996) for the LGM, to avoid uncertainties arising from an interpolation procedure, EC HAM-1 /LSG simulated anomalies, relative to the control run, and CLIMAP reconstructed Eemian SST anomalies, relative to modern SSTS (see below), were first compared only at those model grid-points where CLIMAP cores are located. Tlhisapproach was not pursued with CLIMBER-2 due to its coarse resolution. At those grid boxes which cent ain more than one core, the average SSTS were computed to calculate the anomaly. This approach poses two problems a priori: First, since the dynamical resolution of a GCM is greater than a single grid point, the informational value of a grid box value is somewhat questionable (von Storch, 1995). Second, the size of the area represented by a single grid point is larger than that represented a deep-sea core. However, due to the smooth behavior of the SST field, in this particular case the response over a larger number of model points will not differ much from the response at a single point, neither will the regional climate substantially differ from the large-scale response, so that the comparison based on a single point is appropriatee. Since the proxy-data used. to reconstruct SSTS provide only estimates of summer and winter values, mean annual anomalies (Eemian minus present mean SSTS) were estimated by averaging winter and summer SSTS both for the reconstructions and the simulations. Instead of presenk SST data, core-top estimates (that is, the result of the application of the derived transfer function to the calibration data set) were chosen as modern CLIMAP SST eStimateS. In this way systematic errors in the SST estimates cancel out when computing their anomdles (see below). At those cores where a biotic census for different fauna groups the average over all SST estimates was taken. Errors for the CLIMAP reconstructed SST anomalies were calculated from the standard errors of seasonal estimates. These represent a lower estimate of the total uncertainty in the SST estimates since they only include transfer function errors, which are between 1- 2°C, with an average value of 1.5”C. One problem that has to be addressed with respect to the climate model-data com- parison is the following: Whereas simulated SST anomalies represent deviations with respect to the control run, which is representative of the mid to late 20th century, reconstruct ed SST anomalies represent deviations with respect to the core-top SST es- timates, which correspond to average climatic conditions of the last 1-2 kyr. Indeed, the core-tops have been regressed against mean SSTS over the period 1930-1980 (CLIMAP Project Members, 1984), that is, a time interval that was probably warmer than the last several thousand year. Hence, the transfer functions derived in this way contain a systematic error. As a consec[uence, the core-top estimates are biased towards higher (more positive) values. However, utilization of the same transfer functions throughout the core maintains the same bias; hence, systematic errors cancel out in the CLIMAP reconstructed SST anomalies relative to the core-top estimates. In order to define the 6.2.MODEL-DATA INTERCOMPARISON 83 same reference interval for the SST simulated anomalies, these must be adjusted by the calculated COz-induced temperature warming between preindustrial and mid 20th century levels. The difference in radiative forcing between two levels of C02 Cl ~C2 is approximately given by AF = 6.3 log Cl /C2 Wm– 2. For small radiative perturbations, the corresponding change in global temperature is approximately linear: AT = ~AF. Making use of the fact that the sensitivity of the ECHAM-1/LSG model is 2.6°C for a doubling of COZ the temperature correction is AT= O.79”C. This value was uniformly added to the simulated anomalies. Finally, the yearly standard deviation of the control and Eemian runs were to estimate the error in the simulated anomalies. On a point-by-point basis the model agrees with the observations within 1 standard deviation for ca. 77% of the cores (Figure 6.1). Discrepancies between the climate model and observations primarily occur in regions along eastern boundary currents and subtropical/subpolar frontal zones. These could reflect inadequacies in the model both in terms of physics and resolution (e.g., problems in the simulation of sea ice, strong gradients in these regions and changes in upwelling which are not resolved by the model etc.), but they may also be an indication of problems with transfer functions in such regions (CLIMAP Project Members, 1984; Ravelo et al., 1990), long time- scale (> 2000 years) leads and lags in the climate system that are not captured by an equilibrium simulation (CLIMAP Project Members, 1984; Crowley, 1990) or non- synchronous stratigraphic picks, bioturbation, dissolution, and a possible biasing of some samples by a brief 290 ppmv C02 excursion in the early Eemian (Barnola et al., 1987). The zonally averaged SST responses simulated by ECHAM-1/LSG (adjusted) by CLIMBER-2 in its ocean-atmosphere-vegetation, relative to preindustrial C02 show a remarkable agreement with the zonally averaged CLIMAP reconstructed SST anoma- lies (Figure 6.2a; errors in the model simulations are omitted for the sake of clearness). The number of the selected core sites used at each point of the CLIMAP curve is shown in Figure 6.2b. Both ECHAM-1/LSG and CLIMBER-2, in their best approximations, agree in showing positive anomalies north of 30”N, the largest occurring at mid-high northern latitudes, although the magnitude of such warming is higher in the dlata is located slightly southwards. At low latitudes, the agreement between the simulated and reconstructed curves is satisfactory, with differences below or about 1°C. Disagree- ments are important at southern mid latitudes, where CLIMAP reconstructed SST anomalies reflect a considerable warming which is not reflected by ECHAM-1/LSG or is located to far south in CLIMBER-2. Nevertheless, the fact that the simulated curves show a much smoother behavior than the reconstructed SST curve indicates the more physical behavior of the former and the fact that the latter might be affected by additional errors. The agreement in the total response of the near-surface temperature (including near-surface temperatures over land) simulated by each model is less satisfactory than over the SST response. Figure 6.3 shows the zonally averaged surface temperature change (Eemian minus control run) as simulated by ECHAM-1/LSG and CLIMBER- 2. For additional comparison, the response obtained by a linear EBM (Kim et al., 1998) is shown. As expected, both model responses are considerably warmer than 84 CHAPTER 6. COMPARISON AGAINST RECONSTRUCTED SSTS CLIMAP–GCM 90N 60N 30N .:. EQ 30s 60S -1 90: Figure 6.1: CLIMAP minus model-simulated SST anomalies (difference between Eemian and present summer and winter average SSTS) at CLIMAP core locations. Thin (bold) digits indicate there is (no) significant difference at 1-o level between the CLIMAP and simulated SST anomalies ~C). would be predicted by a linear model, but results with CLIMBER-2 are systematically warmer than for ECHAM- 1/LSG, in particular at high latitudes. Several mechanisms may account for these discrepancies. The increased summer warming and melting of Arctic sea ice in the ECHAM-1/LSG and CLIMBER-2 runs results in warmer winter temperatures than simulated by the EBM. In CLIMBER-2 this warming is further enhanced at high northem latitudes, as a consequence of the effects of positive sea- ice albedo and vegetation-albedo feedbacks. As explained in section 5, this warming translates into enhanced surface stability which in turn causes an attenuation of NADW formation, reduced meridional overturning circulation, and hence reduced northward heat transport, and hence a warming at southern latitudes. 6.3 Global temp~erature differences from present The difference in globally averaged near-surface temperature simulated by ECHAM- l/LSG, corrected to refer to the preindustrial baseline, is ca. 0.5° C (Table 6.1). This results is at variance with the common assumption of considerably warmer than present global temperatures at the Eemian. Nevertheless, such inferences are based in the ev- idence from paleodata, which for the LIG are mainly concentrated over northwestern Europe. Hence, the difference in simulated mean annual near-surface temperature (Eemian minus control run) has been calculated for several selected areas (Table 6.1): 6.3. GLOBAL TEMPERA TUREDIFFERENCES FROM PRESENT 85 4.0 a) 2.0 --- .’ ------.’ ---- g 0“0 _> !$ -2.0 2 1- Cn (n -4.0 -6.0 4 t 1 1 I c , I , 3 z 2 1 0 L--_n 60S 30s E(2 30N 60N Latitude Figure 6.2: a) Zormlly averaged mean annual SST anomalies estimated from CLIMAP SST reconstructions> with their errors calculated from the standard errors of seasonal estimates (solid line) and simulated SSTanomalies by ECHAM-1/LSG only at grid points which contain CLIMAP cores. corrected in order to refer them to the average climatic conditions of the last 1500 years (GCM, dotted line), and by CLIMBER-2, in its ocean-atmosphere-vegetation, relative to preindustrial C02 (AOV-CL, dashed-pointed line) f’ C). Note that errors in the zonally averaged simulated SSTS have been omitted; b,)Latitudinal distribution of the number of cores for which core-top estimates em”st. global, Northern Hemisphere, O-30”N, 30-90”N, and 30-90”N over land only (also in northern summer). The corrected values for the selected areas show that the warmer model temperatures occur in the same region where most of the paleodat a are located, specially in northern summer. Thus, higher than present simulated mean armua,ltem- peratures in the Northern Hemisphere are, at least qualitatively, consistent with the evidence from the paleodata from this region, as well as with reconstructed SSTS, with- out necessarily implying a much higher globally averaged temperature. This would suggest that the proxy temperature estimates could be geographically and seasonally biased. On the other hand, CLIMBER-2 does simulate overall warmer than present SSTS, and a globally averaged temperature higher than its preindustrial control run by 1.2°C. 86 CHAPTER 6. COMPARISON AGAINST RECONSTRUCTED SSTS q ...... Figure 6.3: Zonally averagednear surface temperature change @emian minus modern) as simulated by ECHAM-1/LSG (scdialine), by CLIMBER-2 in its atmosphere-ocean-vegetation version, and by (dashed line) a ;!inearEBM (dotted line), in 0C. Global NH O-30”N30-90”N30-90”N,L 30-90”NL,NS AT (“C) -0.32 -0.15 -0.60 0.27 0.20 3.0 AT~~j(“G) 0.47 0.64 0.19 1.10 0.99 3.79 Table 6.1: Difference (Eemian minus control) in mean near-surface temperature for se- lected areas (global, Northern Hemisphere, O-3( TN,30-9@N, 30-90’N over land only, and for northern summer, as simulated by ECHAM-1/LSG, corrected values to compensate for the different C02 concentrations of our Eemian and control runs (2nd row). 6.4 Summary Comparison of the SST response obtained with the best-guess simulations with ECHAM- l/LSG andCLIMBER-2withCLIMAP(CLIMAPProjectMembers,1984)recon- structedSSTS shows a satisfactory agreement of both models with the paleodata. Dis- crepancies could reflect inadequacies in the model but they may also be an indication of problems with transfer functions in certain regions, pinpointing regions that could be the subject of further modeling studies, for example, regions of the midlatitude northern hemisphere where paleo data are slightly warmer than the climate model (Terasmae, 1960; van der Hammen et al., 1971; Woillard, 1978; Miller et al., 1983; de Vernal et al., 1986; Mangerud and Svendsen, 1992; Pens et al., 1992; Guiot et al., 1993). Despite the SST agreement, the response over land as simulated by the two models differs significantly, offering two different views of the climate at the LIG. In ECHAM- l/LSG, the large seasonal changes cancel out in large part when we consider annual mean surface temperature changes. The results, nevertheless, are consistent with the evidence for warmth from northwestern Europe, and suggest there might be a seasonal and geographical bias in the interpretation of the proxy data. The simulated change in global temperature with respect to the preindustrial baseline (ea. 0.5°C) implies 6.4. SUMMARY 87 that the mid 20~ century may already have had global temperatures higher than.those of the past 200 kyr, and therefore outside the range of the natural climate variability of the most recent past. On the contrary, the response simulated by CLIMBER-2 including interactive vegetation indicates that while the SST response of the EC HAM- l/LSG and CLIMBER-2 models are consistent with each other and with the data, the land, and land plus ocean response is considerable larger than that simulated by ECHAM-1/LSG. CHAPTEIR 7 Conclusions and Discussion Coupled GCMS are the state-of-the-art tools to simulate climate under given boundary conditions and to study the climatic response to changes in these. They constitute a valuable tool for the study of climate variability, in particular the large-scale, low- frequency ocean variability, otherwise not feasible due to the lack of observational data. Coupled GCMS have been used in the past decade to assess potential climate change in response to increased anthropogenic atmospheric greenhouse gas levels. A necessary previous step is to test such climate models under present boundary conditions, in order to establish whether they can adequately simulate the present climate. However, since all models include a number of adjustments to reproduce the present observations, an accurate simulation of the present climate does not guarantee a correct performance of the model under different boundary conditions. The validation of climate models under independent boundary conditions is pro- vided by the simulation of past climates and the comparison of the results to the evidence from the geological record. This approach may additionally provide explana- tions of the reconstructed climatic features in terms of specific physical processes. In this line, the coupled ocean-atmosphere GCM ECHAM-1/LSG has been used to simulate climatic conditions at 125 kBP, the maximum of the last interglacial (LIG) pe- riod (the Eemian, w 120 — 130 kBP). In addition, a climate simulation for the Eemian with identical boundary conditions has been performed with the global climate system model of intermediatee complexity CLIMBER-2. The large-scale simulated responses of the two models agree reasonably well. Howeverj a certain number of disagreements is also found. As a GCM, ECHAM-1/LSG should, within its limitations, be able to simulate as realistically as possible the main processes which significantly aflect the global climate. However, the fast turnaround time of CLIMBER-2 permits carrying out further sensitivity studies which help to shed light on the possible reasons for such discrepancies. In this way, CLIMBER-2 is introduced. as a fast but still rather com- 89 90 CHAPTER 7. CONCLUSIONS AND DISCUSSION plex model which can be used. to extend the knowledge derived from simulations with more sophisticated GCMS ancl to give hints for the direction of further investigations. Results of several sensitivity experiments analyzing the relative influence of changes in orbital parameters and in C02 concentration, as well as investigating the response of interactive vegetation to the altered boundary conditions have been presented. Fi- nally, the simulated best-guess responses obtained by each of the two models have been compared with one another and against CLIMAP reconstructed SST for the LIG. 7.1 ECHAM-1/IAG The results of the climate simulation for the LIG with the coupled atmosphere-ocean GCM ECHAM-1/LSG reflect the amplification of the seasonal cycle of temperatures in the Northern Hemisphere at this time period, relative to the control run simulating the present climate, due to the different configuration of the orbital parameters. A large warming is found over the northern landmasses in northern summer, but severe cooling takes place in winter. This response, which closely agrees with those simulated in previous climate simulations for time periods of the Late Pleistocene with enhanced northern summer insolation (warm summer orbit periods) (Prell and Kutzbach, 1987; Kutzbach and Gallimore, 198/3;de Noblet et al., 1996; Hewitt and Mitchell, 1996), has two main implications for the atmospheric circulation, namely: In the northern hemi- sphere, both in summer and winter, i) the land-sea temperature contrast is intensified; ii) the meridional temperature gradient at mid northern latitudes is attenuated. and its maximum is shifted nortlrwards. As a consequence of the former result, the land- sea surface pressure gradient, whose response to the temperature perturbation in the tropics is thermally direct, is enhanced in the Northern Hemisphere, both in summer and winter. This affects features of the atmospheric circulation which are driven by differential heating, such as the monsoon circulation. The results qualitatively reflect an intensification of the southwest Asian and African monsoon circulation and precipi- tation in summer, consistent with the evidence from the geological record (Petit-Maire, 1994) and with previous climate simulations of warm summer orbit periods of the late Pleistocene (Kutzbach and C]uetter, 1986; Kutzbach and Gallimore, 1988; Mitchell et al., 1988). In addition, features such as the northward migration of the large-scale convergent flow and the intensification of the tropical easterly jet are found as reported in AGCM simulations for 6 and 125 kBP (de Noblet et al., 1996; Hewitt and Mitchell, 1996). The reduction of the meridional temperature gradient at mid northern latitudes, in turn, translates into an attenuation of the mean extra-tropical (westerly) circulation throughout the troposphere, in particular of the jet streams. In addition, the baroclinic activity of the atmosphere is also reduced: the storm-tracks over the North Atlantic and North Pacific are weakened, and precipitation over these regions is reduced relative to the control run. In the Southern Hemisphere, as a consequence of the attenuation of the seasonal cycle of insolation, the land-sea thermal contrast is reduced and the meridional temperature gradient is intensified. The former result was also found by Kutzbach and Guetter (1986) in their AGCM simulation for 9 kBP. To summarize, in northern summer the simulated features reflect enhanced warm- 7.1.ECHAM-1/LSG 91 ing over most landmasses, stronger monsoons and enhanced precipitation over tropical landmasses and over land as a whole. However, due to the compensating effects in tint er, most of the large simulated seasonal differences cancel out in the annual mean. For instance, mean annual near-surface temperature differences are generally below l“C, and the change in global near-surface temperature is -0.3”C; strong winter cool- ing translates into a very strong Asian winter monsoon which over-compensates the summer response, and, as reported by Hewitt and Mitchell (1996), the mean annual simulated features reflect a reduction of the intensity of the monsoon rather than an enhancement, the simulated globally averaged precipitation is reduced by 170. The aforementioned results contrast with the widespread assumption of higher than present global temperatures and precipitation at the LIG. The assumption of higher global temperatures is based on paleodata evidence mainly from northwestern Europe. The assumption of enhanced precipitation is based on the correlation found between enhanced carbon storage on land and precession (Keigwin and Boyle, 1985). Duplessy et al. (1984) present deep-sea 613C data that challenge this second conclusion, suggest- ing less carbon stored on land during the Eemian. However, this does not constitute a straight-forward validation of our results for two reasons. First, the l% decrease in precipitation simulated by ECHAM-1/LSG might not be detectable in the marine &3C record. For example, terrestrial carbon decreases during the LGM, around 20-2570 or larger, lead to a 0.3 – 0.4°/00 change in marine J13C (Crowley, 1995). Accordingly, the change of O.1°/00 in marine J13C found by Duplessy et al. (1984) at the Eemian would translate into about a 570 change in carbon storage on land. If there is any linkage between total precipitation and total carbon storage (a reasonable hypothesis), a 1% decrease in precipitation should therefore cause only about a 0.02°/00 change in the marine 613C record. This small change cannot be detected at a significant level in the marine record. Second, our model utilized a C02 level about 60 ppmv less than the control. Hence, all or part of our l% decrease could be a consequence of the lower C02 levels in the Eemian run. As a consequence of the decreased fresh water flux at mid and high northern lati- tudes, a slight intensification of the meridional overturning circulation is found in the Eemian relative to the control run, both in the Pacific and Atlantic oceans, resulting in an enhanced ocean global northward heat transport by 10VOat 15”N. The largest circulation differences are found in the Pacific Ocean, and are related both to northern summer changes in the hydrological balance over the tropics, associated to precipitation and runoff redistributions over Eastern Asian, and to a reduction of precipitation as- sociated to the storm-tracks in northern winter. The evidence from paleodata (mainly &3C and Cd/Ca) regarding the large-scale ocean circulation differences at the Eemian is ambiguous. Data are sparse over the Pacific basin. In the Atlantic Ocean some of the initial data recovered suggest either reduced NADW formation or enhanced Antarctic Bottom Water (AABW), which in turn could be coupled responses (Duplessy et al., 1984; Duplessy and Shackleton, 1985). However, more recent high-resolution records from the North Atlantic (Oppo et al., 1997; Adkins et al., 1997) suggest mid-Eemian (mid MIS 5e) NADW levels similar to Holocene levels. Hence, the data available at present preclude extracting firm conclusions concerning ocean circulation at this time 92 CHAPTER ‘7.CONCLUSIONS AND DISCUSSION period. In any case, it appears from the paleodata that circulation differences rela- tive to present might have hacl a small amplitude, which would be consistent with the changes simulated by EC HAM- 1/LSG. 7.2 CLIMBER-2 A second climate model has been employed to simulate climatic conditions at the LIG maximum: the climate system model of intermediate complexity CLIMBER-2. For comparison with the results of ECHAM-1 /LSG, a first simulation was performed imi- tsting boundary conditions used with this model. Most of the large-scale atmospheric simulated features found previously are reproduced with CLIMBER-2: warming over northern land-masses in summer and cooling in winter, intensified northern monsoons, an all-year-round warming at high northern latitudes, reduced midlatitude westerly flow throughout the troposphere in the Northern Hemisphere, and opposite changes in the Southern Hemisphere. While the meridional overturning circulation shows es- sentially no differences in the Pacific Ocean, a weakening by ca. 20% takes place in the Atlantic Ocean as a result of the enhanced all-year round warming translating in enhanced surface surface stability and reduced NADW formation. Hence, the main discrepancy concerning the two aforementioned models regards the response of the ocean thermo.haline circulation. To analyze the reasons for such dis- agreements, an integration was performed with CLIMBER-2 in which Eemian bound- ary conditions were imposed but the freshwater flux was prescribed as the sum of the mean fluxes from the CLIMBER-2 control simulation plus the mean freshwater flux anomalies as simulated by ECHAM-1/LSG for the Eemian run, relative to the con- trol run: Given that the temperature response from both models at high northern latitudes reflects an all-year-round warming relative to present, the discrepancies in the thermohaline circulation must be due to differences in the simulated freshwater fluxes. Indeed, the decreased freshwater flux into the North Pacific as imposed from the ECHAM-1/LSG differences results in a similar anomalous circulation pattern to that found in ECHAM-1/LSCr: the Pacific meridional overturning circulation is inten- sified in comparison with the control run, showing an anomalous intermediate flow of about 3 Sv. At the same time AABW is stronger and penetrates farther north. The freshwater fluxes are ultimately dependent on the parameterization of evaporation, pre- cipitation and the runoff scheme employed. This study demonstrates the import ante of the accurate simulation of such processes due to their potential for affecting the global climatic response. The decrease in freshwater flux into the North Pacific as simulated by ECHAM-1/LSG is partly a consequence of an attenuation of the storm-tracks, which are not explicitly resolved in CLIMBER-2. The previous resuhs refer to deviations with respect to a control run with present C02 levels. However, due to the rapid increase of the atmospheric C02 concentration in the past hundred years, the present climate is not in equilibrium with such high levels, and paleodata rather reflect deviations with respect to the last several hundred years. Thus, for the comparison against paleodata it is more appropriate to study the deviations with respect to a control run with a preindustrial C02 concentration. 7.2.CLIMBER-2 93 Anomalous fields simulated by CLIMBER-2 show in this case a more satisfactory agree- ment with the evidence from paleodata than the aforementioned Eemian run relative to a control run with present C02 levels. For instance, whereas comparison with the present-day control run resulted in an mean annual cooling at mid latitudes of the Northern Hemisphere and a in slight decrease of total precipitation for the Eemian, comparison with the modern C02 control run reveals a warming, as is also suggested by geological data (van der Hammen et al., 1971; Miller et al., 1983; Pens et al., 1992) and climate is generally wetter when comparing with the preindustrial contrcd run. In addition, due to additional warming at high northern latitudes, NADW formation is further reduced while AABW is enhanced, in agreement with some the evidence from changes in deep water circulation from 13C data (Duplessy et al., 1984; Duplessy and Shackleton, 1985). The CLIMBER-2 simulated change in global annual tenmpera- ture is only about 0.4”C. This results is still at variance with widely held assumption that the Eemian was significantly warmer than present in terms of globally averaged temperature. Finally, to investigate the role of interactive vegetation at the Eemian, a simulation with the full version (atmosphere-ocean-veget ation) of CLIMBER-2 was performed, and its results compared against those of the preindustrial control run. In addition, in order to disentangle the roles of the atmosphere, ocean and vegetation, two more experiments were carried out: a simulation with the coupled atmosphere-vegetation version of CLIMBER-2, and a simulation with the atmosphere-only CLIMBER-2 com- ponent, in which both ocean as well as vegetation characteristics were &ed tc} their modern values. The main differences concerning the thermal response can be observed over the mid and especially high latitude continents of the Northern Hemisphere. Because of warmer summers, the forests expand farther northward at the expense of tundrab.The rougher surface leads, due to its lower albedo, to additional warming and thus to earlier snow melting which again contributes to an additional temperature increase, which in turn favors forest expansion (vegetation-snow-albedo feedback). In turn, the warmer conditions also lead to a significant decrease of the sea-ice area which due to the lower albedo of open water leads to additional warming (sea ice-albedo feedback). Thus, both the vegetation-snow-albedo feedback and the sea ice-albedo feedback in this region act together in the same direction, leading to a strong amplification of the warming. The results concerning vegetation changes are in line with the findings of Grichuk (1992) for the LIG, showing the Northern Hemisphere up to about 70”N mainly covered by different forest formations of boreal type, and an absence of polar deserts as well as a more limited distribution of tundra. In the sub-tropics extended vegetation strongly amplifies the monsoonal response to the insolation forcing. The model shows large parts of the modern Sahara covered by vegetation. For the L]G, Le Hou4rou (1997) gives Mediterranean forest and tropical savanna, and Grichuk (1992) gives a mixture of savanna and open tropical forests, grass and bush formations of desert type and bush-arboreal formations in the Sahara. Mean annual warming is also found relative to the control run at high southern latitudes. The reason for this is the further weakening of the meridional overturning circulation in the Atlantic [ocean 94 CHAPTER 7. CONCLUSIONS AND DISCUSSION caused by the enhanced surfa,ce stability due to the large warming at high northern latitudes, which translates into a reduction of the northward heat transport causing a further warming of the Southern Hemisphere. In all, the response of the atmosphere- ocean-vegetation system to the changes in insolation results in summer temperature differences over the Northern Hemisphere continents which rise to values of up to ~C, while in winter the differences do not exceed -1.5°C. Strong synergisms are found to operate between the ocean and vegetation systems. In particular, their interaction over northern hemisphere landmasses represents a negative synergism. The reason for this is the aforementioned weakening of the meridional overturning circulation in the Atlantic Ocean which translates into a reduction of the northward heat transport. Altogether, the change in globally averaged annual temperature in this case is l.l°C. This could provide a reconciliation between the widespread evidence for warmth at the LIG and the assumption of a globally warmer world. 7.3 Model-data i,ntercomparison We have compared the simuli~tedSST response of each of the best-guess simulations with ECHAM-1/LSG and CLIIMBER-2. For ECHAM-1/LSG, this means the SST sim- ulated response was corrected, to account for the different C02 concentrations used in the Eemian and control runs, in order to refer simulated anomalies to the preindustrial baseline. For CLIMBER-2, SSTS obtained from the full atmosphere-ocean-vegetation version were used. Comparison of these against the CLIMAP (CLIMAP Project Mem- bers, 1984) reconstructed SSTS shows a satisfactory agreement of both models with the paleodata. Discrepancies could reflect inadequacies in the model, both in terms of physics and resolution (e.g. problems in the sea-ice performance in ECHAM-1/LSG, non resolution of strong gradients, etc.), pinpointing regions that could be the subject of further modeling studies, for example, regions of the midlatitude northern hemi- sphere where paleo data are slightly warmer than the climate model (Terasmae, 1960; van der Hammen et al., 1971; Woillard, 1978; Miller et al., 1983; de Vernal et al., 1986; Mangerud and Svendsen, 1992; Pens et al., 1992; Guiot et al., 1993). Despite the SST agreement, the response over land as simulated by the two models differs significantly, offering two different views of the climate at the LIG. In ECHAM- l/LSG, the large seasonal changes cancel out in large part when we consider annual mean surface temperature differences. The results, nevertheless, are consistent with the evidence for warmth from northwestern Europe, and suggest there might be a sea- sonal and geographical bias in the interpretation of the proxy data. The simulated change in global temperature with respect to the preindustrial baseline (ea. 0.5”C) implies that the mid 20~ century may already have had global temperatures higher than global temperatures over the past 200 kyr (Montoya et al., 1998), and therefore outside the range of the natural climate variabilityy of the most recent past. On the contrary, the response simulated by CLIMBER-2 including interactive vegetation indi- cates that while the SST response of the ECHAM-1/LSG and CLIMBER-2 models are consistent with each other and with the data, the land, and land plus ocean response is considerable larger. 7.4.OUTLOOK 95 7.4 Outlook The idea of performing a paleoclimate simulation with the coupled atmosphere-ocean GCM ECHAM-1/LSG emerged as a further step after the use of this model in the millenia integration in a control run. Hence, as is often the case in climate research, this study was built starting from preexisting work, namely, the control run. This run was designed to simulate unperturbed present mean climatic conditions, and was performed with a constant equivalent C02 concentration of 330 ppmv. This value seems appropriate for a comparison against present climatologies. However, when dealing with features having long time scale responses such as the thermohaline ocean circulation, we cannot consider the present climate being in equilibrium. Paleodata rather reflect deviations with respect to mean values over at least several hundred years. Thus, for comparison with the paleodata, it would be more reasonable to use preindustrial conditions as control simulation instead of the former control run with the higher C02 level. Hence, this should be taken into account for future paleoclimate simulations. Although the former caveat cannot be neglected, and even if, as a consequence, some of the simulated processes might be model dependent, results indicate the direction of further investigations. For instance, ECHAM- l/LSG simulates a weakening of the storm-tracks at the Eemian, which translates into reduced precipitation over the North Atlantic and Pacific Oceans, affecting their freshwater flux balance, which controls the thermohaline circulation. Even if the differences simulated by ECHAM-11/LSG are small, the storm-tracks in the control run are as well severely underestimated. It would be interesting to test whether larger changes, with the potential to affect global climate through changes in the thermohaline circulation would be obtained by means of improved GCMS. This would constitute an interesting test of the validity of CLIMBER-2, in which feedbacks involving the storm-tracks are disregarded. Since the informational value of a grid box value is somewhat questionable (von Storch, 1995), except in this particular case in which the comparison was restricted to the SST response, making use of simulated fields at single grid points is of limited value. These problems are usually dealt with by the use of dynamic and empirical downscaling (Zorita and von Storch, 1997). In the case of empirical downscaling, the main idea is to establish an empirical relation between the anomalies of a large-scale field L and a regional scale field R. In this way, a statistical model is created which translates L anomalies can be translated into R anomalies. An assumption in this approach is the fact that the relationship established between the large and regional scale fields L and R is maintained in an altered climate, which does not hold necessarily. In recent years, quantitative paleoclimate reconstruction are increasingly being produced. .Atthe LIG, one of the most ambitious approaches has been the work by Aalbersberg and Litt (1998). In addition, similar databases for precipitation over northwestern Europe are in preparation (Ren6 Isarin, private communication). Resorting to statistical downscaling would help bridge the gap between GCMS or other climate models with comparable or coarser resolution and the local features evidenced by the proxy data. Comparing simulated distributions with interpolated local geological evidence is the commonly used method of assessing the skill of paleoclimatic simulations. We suggest, 96 CHAPTER 7. CON(5’LUSIONS AND DISCUSSION however, to expand GCMS such that model output directly simulate variables which can be compared to the proxy-data. Additionally, paleodat a should include a measure of the uncertainties in their reconstruction. Full paleoclirnate simulaticms will eventually require inclusion of the effects of vege- tation changes on climate. Current models differ in their quantification of this feedback, While CLIMBER-2 suggests that vegetation feedbacks could additionally raise global temperatures by 0.5°C (Ganopolski et al., 1997) at the Eemian or the early Holocene, two GCM simulations (Texier et al., 1997; Crowley and Baum, 1997) suggest global temperature changes are only on the order of 0.05-O.l°C for the early Holocene and LGM. Hence, additional GCM simulations including interactive vegetation should be performed to establish whether the sensitivity of CLIMBER-2 is overestimated. It should as well be taken into account that climate can be described as a random process whose parameters are determined by the external forces, but each realization of the climate trajectory, as simulated by an AOGCM, is a random realization of this process. That is, it should not be expected that the simulation reproduce in detail the paleo climatic states reconstructed from paleo evidence. Therefore, simulations should be done in ensemble mode reflecting the inherent uncertainty of the climatic process. e.g. by varying the initial state or the forcing within its range of uncertainty. The observed paleoc/imatic state should be a credible member of the ensemble. An additional question would be, in fact, whether the latter is or not well de- fined. For instance, climate during the Eemian interglacial in particular was not stable. Rather, the initial warming phase was followed by a rapid transition to colder condi- tions as paleodata from western and central Europe (Cheddadi et al., 1998) as well as from the Norwegian Sea (Cortijo, 1994) show. This transition started mound the time of 125 kBP. A robust chronology for the Eemian pollen records is still needed (Cheddadi et al., 1998). Future interests should include e.g. the low-frequency climatic variations during the Holocene, rapid events like the Younger Dryas, the inception/termination of glaciation up to the full glacial/interglacial cycles. Thus, paleoclimate simulation for time-slices of the past should be replaced by more realistic transient experiments, CHAPTER 8 Acknowledgments I am very gratefrdto Hans von Storchfor accepting me as aPh. D. student at the Meteorologisches Institut, aswellas forhiscontinuous assistance andsupport, and to Thomas J. Crowley for his constant supervision and for his patience with my often questions and inquiries. I am indebted to Dirk Schriever and Michael Lautenschlager for the simulations for the last interglacial, S.-Y. Kim for providing us with the EBM results, Reinhard Voss and Ulrich Cubasch for discussion concerning technical details of the model, and Jin-Song von Storch for her help in the analysis of the ocean data. Special thanks to Hermann Kuhn for technical assistante with the work-stations. It is a pleasure to thank Claudia Kubatzki for many stimulating discussion, for her readiness to collaborate, and for carefully reading parts of this work. Special thanks also to Stefan Rahmstorf for his constant advise and discussion. I am grateful to Ute Luksch for her constant encouragement and disposition to discuss problems; to Victor Ocaiia, Patrick Heimbach, Ulrich Kilian, Juan Pedro Mont6vez, Fidel Gonzidez, Traute Krueger, Peter Brandt, Dennis Bray, and Mar- tina Junge for providing a most friendly atmosphere; to Matthias Dorn for being so implacable with my German, to Slava Kharinj whose help during my first months in Hamburg was invaluable, and very specially, to Eduardo ‘Zorita, who was always ready to listen to problems, answer questions and give advise. Financial support from the Direcci6n General de Investigaci6n Cientifica y T4cnica (Spain) is gratefully acknowledged. Finally, special thanks to Chema Hernindez and Gemma de la Varga for their friendship, encouragement, patience, comprehension and sense of humor. 97 CHAPTER 9 Appendix Figures 9.1-9.10 shown the mean fields as simulated in the control run in by ECHAM- l/LSG as a reference for the anomalous changes shown in Chapter 4. Mean fields are averaged over the final 300 years of the integration. !39 100 CHAPTER 9. APPENDIX Figure 9.1: Mean difference in the control run in JJA in a) net surface solar radiation @soline spacing (1S) = 10 Win-2]; b) total cloud cover (1S = We,); c) near-surfacetemperature (1S = 1° C), d) SLP (1S = 1 rob,), e) 10-m winds (ins-I), f) total precipitation and g) precipitation minusevaporation (isolines at &15, %1 O, &5, +2, %1, +0.5 mm day– 1,; h) 200-hPa u-velocity (1S = 2 ins-l); h) 200-hPa streamfimction (1S = 3x l@m2s-l); i) 200-hPa velocity potential (IS = l@m2s-1) Shading indicates the level of local recurrence p. Light shading: p>O.8 or p Figure 9.2: Cont. 102 CHAPTER 9. APPENDIX i) Psi200 (1 O**6 m**2/s) j) Chi200 (1O**6 m**2/s) a) Surf. solar rad. (W/m* *2) b) Total cloud cover (%) L-’---- — 602. 30 ~ .- 180 l’bw WI b & 120E 100 d) SLP (rob) p i-l W $5 et 104 CHAPTER 9. APPENDIX Figure 9.3: Cont. 105 i) Psi200 (1 O**6 m**2/s) .j) Chi200 (1 O**6 m**2/s) EON Xfl E4 m m 180 lZW 8LW 0 60E lZOE Iau Figure 9.3: Cont. 106 CHAPTER 9. APPENDIX Figure 9.4: As in Figure 9.1 but in the mean annual. WN m Eo % m I u . I 12aw 6CIW h) U200 (m/s) i) Psi200 (1 O**6 m**2/s) -— -.c6.=D+——~ — .- .- I ta ..xyy@!.. 0 \_lo...... ,, “1 ,_ \ ...... 0 10 10 10 :/ .. X 20 20 \ .?O .\’._. ,’}. - 30s0 30 ~~ 30 kxg{’-~e~20 *. 20 108 CHAPTER 9. APPENDIX i) Psi200 (1 O**6 m**2/s) ------“---”?m:~ . . .—------, -90------...... —---80.. --...... _...--. 0-..—.. ----i+’--”e~ i) Chi200 (1O**6 m**2/s) 6W w m 3CU m .( ‘“ .\. \. %.7---7--- /.. I ‘“!. “. ‘-’---:----:- Figure 9.4: Cont. 109 a) U Ctrl NH winter (m/s) b) U Ctrl NH summer Figure 9.5: Mean zonally averaged zonal velocity in JJA and DJF as simulated in the control run (m s–l). 110 CHAPTER9. APPENDIX STREAMFUNCTION 10*=6 M==3/S :“IL..~ AZE5Y7 D.J 60- 30- 0- 30- 60- S ~’# +---+$ -w i5E d5E 1*5E 145w 9’5w 3kw $5E o.i5 Figure 9.6: Mem25-m currents (ms-l)(top left) ;horizontd bmotropicstream fmction (top right), and zonally averaged mass transport stream function of the meridional circulation in the Atlantic Ocean (bottom left) and in the Pacific Ocean, as simulated in the control run, averaged over 300 years (bottom right) (Sv). 111 b) Figure 9.Y Intra-seasonal standard deviation of the band-pass geopotential height at 500- mb in a) northern summer (JJA] a) northern winter @JF) as simulated in control run (1S = 5 m]. 112 CHAPTER 9. APPENDIX STREAMF”NCTION , 0..6 M.. 3/s 90.’ N 60- 30- 0- 30- 60- S 90~ L 215E 8kE 145E fksw 9’!tw 3% 23E Figure 9.8: Standard deviation of the yearly horizontal barotropic stream function (top left), the zonally averaged mass transport strearnfunctionof the meridional circulation in the Atlantic (top right), in the Pacific (bottom left) and Indian Ocean (bottom right) as simulated in the control run (al;!in 0.05 Sv). 113 90 N 60 30 0 30 60 s 90 5E Figure 9.9: First EOF of the yearly anomaiies of the horizontal barotropic stream function (top left, in Sv), the zonally averaged mass transport stream function of the meridional cir- culation (in Sv) in the Atlantic (top right), Pacific (bet tom left) and Indian Ocean @ot tom right) (Sv). I 114 CHAPTER9. APPENDIX a) 3 2 “i’ 1 ;< !:.:1 j (.$p -3 E 500 600 700 800 900 1000 Tloo 1200 Time [years] b) c) 3 21 -; -2 -3 500 600 700 eoo 900 1000 1100 1202 Time bears] d) 3’ 2’ o1 .N.”. _, :. -2 -3 ~~ k500 600 700 800 900 1000 1100 120Q Figure 9.10: First principal component of yearly anomalies of a] horizontal barotropic streamfunction, and b) zonally averaged mass transport stream function of the meridional circulation in the Atlantic Ocean and c) in the Pacific Ocean, as simulated in the control run. References Aalbersberg, G. and Litt, T. 1998. Multiproxy climate reconstructions for the Eemian and Early Weichselian. J. Quat. Sci., 13, 367–390. Adkins, J. F., Boyle, E. A., Keigwin, L. and Cortijo, E. 1997. Variability of the North Atlantic thermohaline circulation during the last interglacial period. Nature, 390, 154-156. An, Z., Kukla, G. J., Porter, S. C. and Xiao, J. 1991. Magnetic susceptibility evidence of monsoon variation on the loess plateau of central China during the last 1:30,000 \ years. Quat. Res., 36, 29–36. Anderson, D. A. and Prell, W. L. 1993. A 300 Kyr record of upwelling off Oman during the Late Quaternary: evidence of the Asian southwest monsoon. Pateoceanogr., 8, 193–208. Bakan, S., Chlond, A., Cubasch, U., Feichter, J., Graf, H. F., Grassl, H., Hassehnann, K., Kirchner, I., Latif, M., Roeckner, E., Sausen, R.j Schlese, U., Schriever, D., Schult, I., Schumann, U., Sielmann, F. and WeIke, W. 1991. Climate response to smoke from the burning oil wells in Kuwait. Nature, 351, 367–371. Barnola, J. M., Raynaud, D., Korotkevich, Y. S. and Lorius, C. 1987. Vostok ice core provides 160,000-year record of atmospheric COZ. Nature, 329, 408–414. Behre, K. E. 1989. Biostratigraphy of the last glacial period in Europe. Quat. Sci. Rev., 8, 25-44. Berger, A. L, 1978. Long-term variations of daily insolation and Quaternary climatic changes. J. Atmos. Ski., 35, 2362–2367. Blackmon, M. 1976. A climatological spectral study of the 500 mb geopotential height of the Northern Hemisphere. J. Atm. Sci., 33, 1607–1623. 115 116 REFERENCES Bloom, A. L., Chappel, J. M. A., Broecker, W. S., Matthews, R. K. and Mesolella, K. J. 1974. Quaternary sea-level fluctuations on a tectonic coast: New zsOTh/zsAU dates from the Huon peninsula. Quat. Res., 4, 185-205. Bond, G., Broecker, W., Johnson, S., Jouzel, J., Labeyrie, L., McManus, J. and Bonani, G. 1993. Correlations between climate records from North Atlantic sediments and Greenland ice. Nature, 365, 143–147. Broccoli, A. J. and Marciniak, E. P. 1996. Comparing simulated glacial climate and paleodata: A reexamination. Paleoceanography, 11, 3–14,. Brovkin, V., Ganopolski, A. and Svirezhev, Y. 1997. A continuous climate-vegetation classification for use in climate-biosphere studies. Eco. Model., 1997, 251–261. Budyko, M. I. and Izrael, Y. A. 1991. Anthropogenic Climatic Ghange. Arizona Univ. Press. Bush, A. B and Philander, G. H. 1998. The role of ocean-atmosphere interactions in tropical cooling during the last glacial maximum. Science, 279, 1341–1344. Charney, J. G., H., Stone P. and J., Quirk W. 1976. Reply. Science, 191, 100-102. Cheddadi, R., Mamakova, K., Guiot, J., de Beaulieu, J.-L., Reille, M., Andrieu, V. and W. Granoszewski, O. Peyron, 1998. Was the climate of the Eemian stable? A quantitative climate reconstruction from seven European pollen records. Palaeo- geography, Palaeoclimatology, Palaeoecoiogy, 143, 73–85. Chen, J. H., Curran, H. A., White, B. and Wasserburg, G. J. 1991. Precise chronology of the last interglacial period: 234U - 230Th data from fossil coral reefs in the Bahamas. Geol. Sot. Am. Bull., 103, 92–97. Claussen, M. 1997. Modeling bio-geophysical feedback in the African and Indian Mon- soon Region. Clim. Dyn.,, 13, 247–257. Claussen, M. and Gayler, V, 1997. The Greening of the Sahara during the Mid- Holocene: Results of an Interactive Atmosphere-Biome Model. Glob, Ecol. Bio- geog. Let., 6, 369–377. Clemens, S., Prell, W., Murray, D., Shimmield, G. and Weedon, G. 1991. Forcing mechanism of the Indian monsoon. Nature, 353, 720–725. CLIMAP Project Members. 1984. The last interglacial ocean. (&at. Res,, 21, 123-224. Cortijo, E. 1994. Eemian cooling in the Norwegian Sea and the North Atlantic Ocean preceding continental ice sheet growth. Nature, 372, 446–449. Crowley, T. J. 1990. Are there any satisfactory analogs for a future greenhouse warm- ing? J. Clim., 3, 1282–1’292. REFERENCES 117 Crowley, T. J. 1994. Potential reconciliation of Devils Hole anddeep-sea Pleistocene chronology. Paleoceanogr., 9, 1–5. Crowley, T. J. 1995. Ice age terrestrial carbon changes revisited. Glob. Biogeochem. Cycles, 9, 377-389. Crowley, T. J. and Baum, S. K. 1997. Effect of vegetation on an ice age climate model simulation. J. Geophys. Bes., 102, 16463–16480. Crowley, T. J. and Kim, K. Y. 1992. Complementary roles of orbital insolation and North Atlantic Deep water during Late Pleistocene interglacial. Paleoceanogr., 7, 521-528. Crowley, T. J. and Kim, K.-Y. 1994. Milankovitch forcing of the last interglacial sea level. Science, 265, 1566-1568. Crowley, T. J. and North, G. R. 1991. Paleoclimatology. Oxford Univ. Press, 349 pp. Cubasch, U., Hasselmann, K., Hock, H., Maier-Reimer, E., Mikolajewicz, U., Santer, B.D. and Sausen, R. 1992. Time-dependent greenhouse warming computations with a coupled ocean-atmosphere model. C?im. Dyn., 8, 55–69. Cubasch, U., Santer, B. D., Hellbach, A., Hegerl, G. C., Hock, H., Maier-Reimer, E., Mikolajewicz, U., Stossel, A. and Voss, R. 1994. Monte Carlo climate change forecasts with a global coupled ocean-atmosphere model. Clim. Dyn., 10, 11-19. Cubasch, U., Hegerl, G. C., Hellbach, A., Hock, H., Mikolajewicz, U., Santer, B. D. and Voss, R. 1995. A climate change simulation starting from 1935. CWn. Dyn., 11, 75-84. Dansgaard, W., Johnsen, S. J., Clausen, H. B., Dahl-Jensen, D., Gundestrup, N. S., Hammer, C. U., Hvidberg, C. S., Steffensen, J. P., Sveinbj6rnsdottr, A. E., Jouzel, J. and Bond, G. 1993. Evidence of general instability of past climate from a 250-kyr ice-core record. Nature, 266, 218–220. de Beaulieu, J. L. and Reille, M. 1984. A long upper Pleistocene pollen recorcl from Les Echets, near Lyon, France. Boreas, 133, 111-132. de Beaulieu, J. L. and Reille, M. 1984b. The pollen sequence of Les Echets (France): a new element for the chronology of the Upper Pleistocene. Geogr. Phys. Quat., 38(l), 3-9. de Beaulieu, J. L. and Reille, M. 1992. The last climatic cycle at La Grande Pile (Vosges, France). A new pollen profile. Quat. Sci. Revs., 11, 431-438. de Beaulieu, J. L. and Reille, M. 1992b. Long Pleistocene pollen sequences from the Velay Plateau (Massif Central, France), I. Veget. Hist. Archaeobot., 1, 233-242. 118 REFERENCES de Noblet, N., Braconnot, P., Joussaume, S. and Masson, V. 1996. Sensitivity of simulated Asian and African summer monsoons to orbitally induced variations in insolation 126, 115 and 6 kBP. Clim. Dgn., 12, 589–603. de Vernal, A., Causse, C., Hillaire-Marcel, L., Mott, R. J. and Occhietti, S. 1986. Pa- lynostratigraphy and Th/U ages of upper Pleistocene interglacial and interstadial deposits on Cape Breton Island, eastern Canada. Geology, 14, 554-557. de Vernal, A., Miller, G. H. and Hillaire-Marcel, L. 1991. Paleoenvironments of the last interglacial in the northwestern North Antlantic region and adjacent mainland Canada. Quat. In-t., 10-12, 95-106. Dickinson, R.E., HenderssonSellers, A., Kennedy, P.J. and Wilson, M.F. 1986. Biosphere-Atmosphere Transfer Scheme (BATS) for the NCAR CCM, NCAR/TN- 275-STR. National Center for Atmospheric Research, Boulder, Colorado, 69 pp. Dong, B. and Valdes, P. J. 1998. Simulations of the last glacial maximum climates using a general circulation model: prescribed versus computed sea surface temperatures. Clim. Dyn., 14, 571-591. Duplessy, J. C. and Shackleton, N. J. 1985. Response of global deepwater circulation to Earth’s climatic change 135,000-107,000 years ago. Nature, 316, 500–506. Duplessy, J. C., Shackleton, N,., Matthews, R. K., Prell, W., Ruddiman, W. F., Caralp, M. and Hendy, C. H. 1984, 13Crecord of benthic foraminifera in the last interglacial ocean: implications for the carbon cycle and the global deep water circulation. Quat. Res., 21, 225-243. Edwards, R. L. and Gallup, C. D. 1993. Technical comments: Dating of the Devils Hole calcite vein. Science, 259, 1626–1626. Edwards, R. L., Chen, J. H., Ku, T.-L. and Wasserburg, G. J. 1987. Precise timing of the Iast interglacial perioci from mass spectrometric determination of Thorium-230 in corals. Science, 236, 1547–1553. Egger, J. 1997. Flux correction: tests with a simple ocean-atmosphere model. Clim. Dyn., 13, 285-292. Emeis, K. C., Anderson, D. A., Doose, H., Kroon, D. and Schulz-Bull, D. 1991. Sea- surface temperatures and the history of Monsoon upwelling in the northwest Ara- bian Sea during the last 500,000 years. Quat. Res., 43, 366-361. Emiliani, C. 1955. Pleistocene temperatures. J. Geology, 63, 535-578. Field, M. H., Huntley, B. and Muller, H. 1994. Eemian climate fluctuations observed in a European pollen record. Nature, 371, 779–783. Follieri, M. and Sadori, L. 1988. 250,000 year pollen record from Vane di Castiglione (Roma). Pollen Spores, 30,329-356. REFERENCES 119 Fronval, T. and Jansen, E. 1996. Rapid changes inocean circulation and heat flux in the Nordic seas during the last interglacial period. Nature, 383, 806–810. Fung, I., Lacis, A., Rind, D., Lebedeff, S., Ruedyand, R. and Russell, G. 1988. Global climate changes as forecast by Goddard Institute for Space Studies three- dimensional model. J. Geophys. Res., 93, 9341-9364. Gallup, C. D., Edwards, R. L. and Johnson, R. G. 1994. The timing of high sea levels over the past 200,000 years. Science, 263, 796–800. Ganopolski, A., Rahmstorf, S., Petoukhov, V. and Claussen, M. 1997. Simulation of modern and glacial climates with a coupled global climate model. Nature, 391, 350-356. Ganopolski, A., Kubatzki, C., Claussen, M., Brovkin, V. and Petoukhov, V. 1998. The influence of vegetation-atmosphere-ocean interaction on climate during the mid-Holocene. Science, 280, 1916–1919. Grichuk, V.P. 1992. Vegetation during the Last Interglacial. Page 11 of: B. Frenzel, M. Pesci, A. A. Velichko (cd), Atlas of Paleoclimates and Paleoenvironments of the Northern Hemisphere. Late Pleistocene - Holocene, vol. 85. Gustav l?ischer Verlag, Stuttgart. GRIP Members. 1993. Climate instability during the last interglacial period record in the GRIP ice core. Nature, 364, 203–207. Grootes, P. M., Stuiver, M., White, J. W. C., Johnsen, S. and Jouzel, J. 1993. Com- parison of oxygen isotope records from the GISP2 and GRIP Greenland ice cores. Nature, 366, 552–554. Guilderson, T. P., Fairbanks, R. G. and Rubenstone, J. L. 1994. Tropical temperature variations since 20,000 years ago: Modulating interhemispheric climate change. Science, 263. Guiot, J. 1990. Methodology of the last climatic cycle reconstruction in France from pollen data. Palaeogeography, Palaeoclimatology, Palaeoecology, 80, 49-69. Guiot, J., Pens, A., de Beaulieu, J. L. and Reille, M. 1989. A 140,000-year continental climate reconstruction from two European pollen records. Nature, 338, 309–313. Guiot, J., de Beaulieu, J. L., David, F., Ponel, P. and Reille, M. 1993. The climate in Western Europe during the last Glacial/Interglacial cycle derived from pollen and insect remains. Palaeogeography, Palaeoclimatology, Palaeoecology, 103, 73--93, Hall, N. M. J. and Valdes, P. J. 1997. A GCM simulation of the climate 6000 years ago. J. Ch-n., 10, 3–17. 120 REFERENCES Harrison, S. P., Kutzbach, J. E., Prentice, C. E., BeMing, P. J. and Sykes, M. T. 1995. The response of Northern Hemisphere extratropical climate and vegetation to orbit ally induced changes in insolation during the last interglaciation. Quat. Res,, 43, 174–184. Hellerman, S. and Rosenstein, M. 1983. Normal monthly wind stress over the World ocean with error estimates. J. Phys. Oceanogr., 13, 1093–1104. Hewitt, C. D. and Mitchell, J. F. B. 1996. GCM simulations of the climate of 6kyr BP: Mean changes and interdecadal variability. J. CZim., 9, 3505-3529. Hewitt, C. D. and Mitchell, J. F. B. 1998. A fully coupled GCM simulation of the climate of the mid-Holocene. Geophgs. Res. Lett., 25, 361–364. Hoskins, B. J., James, I. N. anti White, G. H. 1983. The shape, propagation, and mean- flow interaction of large-scale weather systems. J. Aim. Sci., 40, 1595-1612. Houghton, J. T., Callander, E. A. and Varney, S. K. 1990. The 1P(X’ Scientific As- sessment. Cambridge University Press, Cambridge. Hurrel, J. W. 1996. Influence of variations in extratropical wintertime teleconnections on Northern Hemisphere temperature. Geophys. Res. Lett., 665–668. Imbrie, J. 1993. Milankovitch theory viewed fro Devils Hole. Nature, 363, 531-533. Imbrie, J., Hays, J. D., Martinson, D. G., McIntyre, A., Mix, A. C., Morley, J. J., Pisias, N. G. and Shackleton, N. J. 1984. The orbital theory of Pleistocene climate: Support from a revised chronology of the marine J180 record. Pages 269-305 of Berger, A., Imbrie, J., Ha,ysa,J., Kukla, G. and Salzmann, B. (eds), IUWankovitch and Climate. Reidel, Hin:gham, Massachusetts. Joussaurne, S. and Braconnot, P. 1997. Sensitivity of paleoclimate simulation results to season definitions. J. Geophys. Res., 102, 1943–1956. Jouzel, J., Lorius, C., Petit, J. R., Genthon, C., Barkov, N. I., Kotlyakov, V. M. and Petrov, V. M. 1987. Vostok ice core: A continuous isotope temperature record over the last climatic cycle (160,000 years). Nature, 329, 403–408. Jouzel, J., Barkov, N. I., Ba,rnola, J. M., Chappellaz, J., Ghenton, C., Kotlyakov, V. M., Lipenkov, V., Lorius, C., Petit, J. R., Raynaud, D., Raisbeck, G., Ritz, C., Sowers, T., Stievenard, M., Yiou, F. and Yiou, P. 1993. Extending the Vostok ice-core record to the penultimate glacial period. Nature, 364, 407–412. Kageyarna, H., Valdes, P. J., Ramstein, G., Hewitt, C. and Wyputta, U. 1998. N“orth- ern hemisphere storm tracks in present-day and last glacial maximum climate simulations: A comparison of the European PMIP models. J. Clim. REFERENCES 121 Keigtin, L.md Boyle, E. A.1985. Carbon isotopes indeep-seaforminifera: precession and changes in low-latitude biomass. In: Sundquist, E. T. and Broecker, W. S. (eds), The Carbon cycle and atmospheric COZ: Natural variations Archean to present. Geophys. Mono., no. 32. Am. Geophys. Union, 319-328. Keigwin, L. D., Curry, W. B., Lehman, S. J. and Johnsen, S. 1994. The role of the deep ocean in North Atlantic climate change between 70 and 130 kyr ago. Nature, 371, 323-326. Kellog, T. B. 1980. Paleoclimatology and paleoceanography of the Norwegian and Greenland seas: glacial-interglacial contrasts. Boreas, 9, 115-137. Kellogg, W. 1977. Eflects of Human Activities on Global Climate. Rept. 486, World Meteorological Org. Kim, S. Y., Crowley, T. J. and St6ssel, A. 1998. Local orbital forcing of Antarctic climate change during the last interglacial. Science, 280, 728–730. Klein, R., Loya, Y., Gvirtzman, G., Isdale, P. and Susie, M. 1990. Seasonal rainfall in the Sinai Desert during the Late Quaternary inferred from fluorescent bamds in fossil corals. Nature, 345, 145-147. Ku, T. L., Kimmel, M. A., Easton, W. H. and O’Neil, T. J. 1974. Eustatic sea level 120,000 years ago on Oahu, Hawaii. Science, 183, 959-962. Kubatzki, C. and Claussen, M. 1998. Simulation of the global bio-geophysical interac- tions during the Last Glacial Maximum. Clim. Dyn., 14, 461471. Kutzbach, J. E. and Gallimore, R. 1988. Sensitivity of a coupled atmosphere/rnixed layer ocean model to changes in orbital forcing at 9000 BP. 3, Geophys. Res., 93, 801-821. Kutzbach, J. E. and Guetter, P. J. 1986. The influence of changing orbital parameters and surface boundary conditions on climate simulations for the past 18,000 years. J. Atm. Sci., 43, 1726-1759. Kutzbach, J. E. and Liu, Z. 1997. Response of the African monsoon to orbital forcing and ocean feedbacks in the middle Holocene. Science, 278, 440–443. Kutzbach, J. E., Gallimore, R. G. and Guetter, P. J. 1991. Sensitivity experiments on the effects of orbitally-caused insolation changes on the interglacial climate of high northern latitudes. Quat. Int., 223-229. Kutzbach, J. E., Bonan, G., Foley, J. and Harrison, S. P. 1996. Vegetation and soil feedbacks on the response of the African monsoon to orbital forcing in the early to middle Holocene. Nature, 384, 623–626. Le Hou6rou, H.N. 1997. Climate, flora and fauna changes in the Sahara over the past 500 million years. Journal of Arid Environment, 37, 619-647. 122 REFERENCES Lehman, S. J. and Keigwin, L. D. 1992. Sudden changes in North Atlantic circulation during the last deglaciation. Nature, 365, 757-762. Levitus, S. 1982. Climatology atlas of the world ocean. NOAA Prof. Paper, 173 pp. LIGA Members. 1991. Report of 1st discussion group: the last interglacial in high latitudes of the northern hemisphere: terrestrial and marine evidence. Quat. Int., 10-12, 9–28. Ludwig, K. R., Simmons, K. R., Szabo, B. J., Winograd, 1. J., Landwehr, J. M., Riggs, A. C. and Hoffman, R. J. 1992. Mass spectrometric 230Th–234U –23s Th. Science, 258, 284–287. Maier-Reimer, E., Mikolajewicz, U. and Hasselmann, K. 1993. Mean circulation of the Hamburg LSG OGCM and its sensitivity to the thermohaline surface forcing. J. Phys. Oceanogr., 23, 731--757. Manabe, S., Spelman, M. J., Stoutler, R. J. and Bryan, K. 1991. Transient responses of a coupled ocean-atmosphere model for gradual changes of atmospheric C02. Part I: annual mean response. J. (Mm., 4, 785-818. Manabe, S., Spelman, M. J. and Stouffer, R. J. 1992. Transient responses of a coupled ocean-atmosphere model for gradual changes of atmospheric C02. Part II: seasonal response. J. Clim., 5, 105–126. Mangerud, J. 1989. Correlation of the Eemian and Weichselian with deep sea oxygen stratigraphy. Quat. Int., 3-4, 1–4. Mangerud, J. and Svendsen, J. 1. 1992. The last interglacial-glacial period on Spits- bergen, Svalbard. Quat. Ski. Revs., 11, 633–664. Mangerud, J., S@nstegaard, IE., Sejrup, H. P. and Haldorsen, S. 1981. A continu- ous Eemian-Early Weichselian sequence containing pollen and marine fossils at Fjosanger, Western Norway. Boreas, 10, 137-208. Mann, M. E. and Park, J. 1996. Joint spatio-temporal modes of surface tempera- ture and sea level pressure variability in the Northern Hemisphere during the last century. J. (Xim., 9, 213’7–2162. Martinson, D. G., Pisias, N. G., Hays, J. D., Imbrie, J., Moore, T. C. and Shackleton, N. J. 1987. Age dating and the orbital theory of the ice ages: development of a high-resolution Oto 300,000 year chronostratigraphy. Quat. Res., 27, 1-29. McManus, J. F., Bond, G. C., Broecker, W. S., Johnsen, S., Labeyrie, L. and Higgins, S. 1994. High-resolution climate records from the North Atlantic during the last interglacial. Nature, 371, 326–329. Mesolella, K. J, Matthews, R. K., Broecker, W. S. and Thurber, D. L. 1969. The astronomical theory of climatic change: Barbados data. J. Geol., 77, 250–274. REFERENCES 123 Miller, G. H., Sejrup, H. P., Mangerud, J. and Andersen, B. G. 1983. Amino-acid ratios in Quaternary mollusks and foramitifera from western Norway: Correlation, geochronology andpaleotemperature estimates. Boreas, 12, 107-124. Mitchell, J. F. B., Grahame, N. S. and Needham, K. J. 1988. Climate simulations for 9000yearsbeforepresent:SeasonalvariationsandefiectsontheLaurentideice sheet. J. Geophys. Res., 93, 8283–8303. Monin, A. S. 1986. Introduction to the Theory of Climate. Reidel, 261 pp. Montoya, M., Crowley,T. J. andvonStorch,H. 1998. Temperaturesat tlhe last interglacial simulated by means of a coupled generalcirctiationmodel.Paleo- ceanography, 13, 170-177. Oppo, D. W., Horowitz, M. and Lehman, S. J. 1997. Marine core evidence for a sharp decrease in deep water production during Termination II and a relatively stable MIS 5e (Eemian). Paleoceanogr., 12, 51–64. Peel, D. 1995. Ice cores: profiles of the past. Nature, 378, 234-235. Peixoto, J. P. and Oort, A. H. 1992. Physics of Climate. American Institute of Physics, New York. Petit-Maire, N. 1993. Past global changes climatic changes and the tropical arid/semiarid belt in the North of Africa. Pages 551–560 of: Thorweihe, lJ. and Schandelmeier, H. (eds), Geoscientific research in northeast Africa. Balkema, Rot- terdam (NL). Petit-Maire, N. 1994. Natural variability of the Asian, Indian and African monsoons over the last 130 ka. In: Desbois, M. and D6salmand, F. (eds), Global Precipi- tations and Climate Change. NATO ASI Series, vol. 1, no. 26. Springer-Verlag, 3-25. Petit-Maire, N., Casta, L., Delibrias, G. and Gaven, C. 1980. Preliminary data on quaternary paleolacustrine deposits in the Wadl ash Shati, Libya. Pages 797-807 of: Salem, M. J. and Busrewil, M. T. (eds), The geology of Libya. Academic Press, London. Petit-Maire, N. (cd.). 1982. Le Shati, lac Pleistocene du Fezzan (Libye). CNRS, iMarseille/Paris {France), 118 pp. Petoukhov, V., Ganopolski, A., Brovkin, V., Claussen, M., Eliseev, A., Kubatzki, C. and Rahmstorf, S. 1999. CLIMBER-2: A climate model of intermediate complex- ity. Part 1: Model description and performance for present climate. Clim. Dyn., submitted. Pokras, E. M. and Mix, A. C. 1985. Eolian evidence for spatial variability of Late Quaternary climates in tropical Africa. Quat. Res., 24, 137-139. 124 REFERENCES Pens, A. and Reille, M. 1988. The Holocene andupper Pleistocene pollen record from Padul (Granada, Spain). Palaeogeography, F’alaeochmatology, Palaeoecology, 66, 243-263. Pens, A., Guiot, J. L., de Beaulieu, J. L. and Reille, M. 1992. Recent contributions to the climatology of the last glacial-interglacial cycle based on French pollen sequences. Quat. Sci. Revs., 11, 439–448. Prell, W. L. 1984a. Monsoonal climate if the Arabian sea during the Late Quaternary: a response to changing solar radiation. Pages 349–366of:~~l~~~ov~~clzand Climate. Reidel, Hingham, Massachusetts. Prell, W. L. 1984b. Variation of monsoonal upwelling: a response to changing solar radiation. Pages 48–57 of: Hansen, J. and Takahashi, T. (eds), Climate processes and climate sensitivity. AGU, Reidel. Prell, W. L. and Kutzbach, J. E. 1987. Monsoon variability over the last 150,000 years. J. Geophys. Res., 92, 8411-8425. Rahmstorf, S. 1996. On the Freshwater Forcing and Transport of the Atlantic Ther- mohaline Circulation. Clim. Dyn., 12, 799–811. Ravelo, A. C., Fairbanks, R. G. and Philander, S. G. H. 1990. Reconstructing tropicaI Atlantic hydrography using planktonic forarninifera and an ocean model. Paleo- ceanography, 5, 409–431. Reille, M. and de Beaulieu, J. L. 1988. The end of the Eemian and the pre-Wurm interstadial as evidenced for the first time in the French Massif Central from pollen analysis. CR. Acad. Sci. U, 306, 1205–1210. Roeckner, E., Arpe, K., Bengtsson, L., Brinkop, S., Duemenil, L., Esch, M., Kirk, E., Lunkeit, F., Ponater, M., Rockel, B., Sausen, R., Schlese, U., Schubert, S. and Windelbzmd, M. 1992. Simulation of the present day climate with the ECHAM model: Impact of model physics and resolution. Max-Planck-Institut fuer Mete- orologic Rep. 93. (Available from Max-Planck-Institut fuer Meteorologie, Bun- desstrasse 55, D-20146, Germany). Rossignol-Strick, M., Nesterofl’, W., Olive, P. and Vergnaud-Grazzini, C. 1982. After the Deluge: Mediterranean stagnation and sapropel formation. Nature, 295, 105– 110. Rostek, F., Ruhland, G., Bassinet, F. C., Muller, P. J., Labeyrie, L. D., Lancelot, Y. and Bard, E. 1993. Reconstructing sea surface temperature and salinity using 6180 and alkeonone records. Nature, 364, 319-321. Royer, J. F., Deque, F. and Pestiaux, P. 1984. A sensitivity experiment to astronomical forcing with a spectraI GCM for the simulation of July 125 kBP years ago. ln: REFERENCES 125 Berger, A. L. and Nicolis, C. (eds), Newperspectiues in Climate modeling. De- velopments in Atmospheric Science. NATO ASI Series, vol. 16. Springer-Verlag, 269-285. Salanville, P. 1992. Changements climatiques clans la p6ninsule Arabique durant le Pleistocene sup&ieur et l’Holocene. Pal&orient, 18, 193-196. Sausen, R., Barthel, K. and Hasselmann, K. 1988. Coupled ocean-atmosphere lmodels with flux corrections. Clim. Dgn., 2, 154–163. Shabalova, M. and K&men, G. P. 1995. Climate change scenarios: comparison of paleoreconstructions with recent temperature changes. Clim. Change, 291, 409– 428. Shackleton, N. J. 1969. The last interglacial in the marine and terrestrial records. Proc. R. Sot., B174, 135–154. Shackleton, N. J. 1993. Last interglacial in Devils Hole. Nature, 362, 596. Slowey, N. C., Henderson, G. M. and Curry, W. B. 1996. Direct U-Th dating of marine sediments from the two most recent interglacial periods. Nature, 383, 242--244. Stocker, T. F., Wright, D.G. and Mysak, L.A. 1992. A zonally averaged, coupled ocean- atmosphere model for paleoclimate studies. J. C/ire., 5, 773–797. Taylor, K. C., Hammer, C. U., Alley, R. B., Clausen, H. B., Dahl-Jensen, D., Gows, A. J., Gundestrup, N. S., Kipfstuhl, J., Moore, J. C. and Wadington, E. D 1993. Electrical conductivity measurements from the GISP2 and GRIP Greenland ice cores. Nature, 366, 549–552. Terasmae, J. 1960. A palynological study of Pleistocene interglacial beds at Tclronto, Ontario. Geol. Surv. Can. Bull., 56, 23–41. Texier, D., de Noblet, N., Harrison, S. P., Haxeltine, A., Jolly, D., Joussaume, S., Laarif, F., Prentice, I. C. and Tarasov, P. 1997. Quantifying the role of biosphere- atmosphere feedbacks in climate change: coupled model simulations for 6000 years BP and comparison with paleodata for northern Eurasia and northern Africa. Clim. Dyn., 13, 865-882. Thouveny, N. et al. 1994. Climate variations in Europe over the past 140kyr deduced from rock magnetism. Nature, 371, 503-506. Turon, J. L. 1984. Direct land-sea correlations in the last interglacial complex. Nature, 309, 673–676. Tzedakis, P. C. 1993. Long-term trees population in northwest Greece through multiple Quaternary climatic cycles. Nature, 364,437-440. 126 REFERENCES van der Hammen, T., Wijmstra, T. A. and Zagwijn, W.H. 1971. Thej70ral record of the Late Cenozoic of Europe in The Late Cenozoic Glacial Ages. K. K. Turekian, Yale Univ. Press, New Ha,ven,Corm. von Storch, H. 1995. Inconsistencies at the interface of climate impact studies and global climate research. Jleteorol. Z, 4 NF, 72-80. von Storch, H. and Zwiers, F, 1999. Statistical Analysis in Climate Research. Cam- bridge University Press. 528 pp. von Storch, J.-S. 1994. Interdecadal variability in a coupled model. Tellus, 46A, 419-432. von Storch, J.-S., Kharin, V., Cubasch, U., Hegerl, G. C., Schriever, D., von Storch, H. and Zorita, E. 1997. A 1260-year control integration with the coupled ECHAM1/LSG general circulation model. J. C/ire, 10, 1525-1543. Watts, W. A. 1985. A long pcjllen record from Lagui di Monticchio, southern Italy: A preliminary account. J. Geol. Sot. Lend., 142, 491–499. Wijmstra, T. A. 1969. Palynology of the first 30 meters of a 120 m deep section in Northern Greece. Acts Bet. Neerl., 18, 511-527. Winograd, I. J., Coplen, T. B., Landwehr, J. ML, Riggs, A. C., Ludwig, K. R., Szabo, B. J., Kolesar, P. T. and. Revesz, K. M. 1992. Continuous 500,000 year climate record from vein calcite i:nDevils Hole, Nevada. Science, 258, 255-260. Woillard, G. M. 1978. Grande Pile peat bog: A continuous pollen record for the last 140,000 years. Quat, Res., 9, 1–21. Woodruf, S. D., Slutz, R. J., Jeme, R. L. and Steurer, P. M. 1987. A comprehensive ocean-atmosphere data set. Bull. Amer. Meteor. Sot., 68, 1239–1250. Xu, J. S., von Storch, H. and van Loon, H. 1990. The performance of four spectral GCMS in the Southern Hemisphere: The January and July climatology and the semiannual wave. J. Clim., 3, 53–70. Zagwijn, W. H. 1996. An analysis of Eemian climate in western and central Europe. Quat. Sci. Revs., 451-46!3. Zorita, E. and von Storch, H. 1997. A survey of statistical downscaling techniques. Tech. rept. 97/E/20. GKSS-Forschungszentrum. Zubakov, V. A. and Borzenkcwa, I. I. 1990. Global Pa/aeoclimate of the late Cenozoic. Elsevier, Amsterdam, 453 pp.