Climate Dynamics

, Volume 22, Issue 2, pp 223–238

Synergistic feedbacks between ocean and vegetation on mid- and high-latitude climates during the mid-Holocene

Authors

    • Max Planck Institute for Biogeochemistry, PO Box 100164, 07701 Jena, Germany
  • S. P. Harrison
    • Max Planck Institute for Biogeochemistry, PO Box 100164, 07701 Jena, Germany
  • P. Braconnot
    • IPSL/LSCE – Laboratoire des Sciences du Climat et de l’Environnement, Unité mixte CNRS-CEA, D.S.M./Orme des Merisiers/Bat. 709, CEA/Saclay, Gif-sur-Yvette, 91191, France
Article

DOI: 10.1007/s00382-003-0379-4

Cite this article as:
Wohlfahrt, J., Harrison, S.P. & Braconnot, P. Climate Dynamics (2004) 22: 223. doi:10.1007/s00382-003-0379-4
  • 113 Views

Abstract

Simulations with the IPSL atmosphere–ocean model asynchronously coupled with the BIOME1 vegetation model show the impact of ocean and vegetation feedbacks, and their synergy, on mid- and high-latitude (>40°N) climate in response to orbitally-induced changes in mid-Holocene insolation. The atmospheric response to orbital forcing produces a +1.2 °C warming over the continents in summer and a cooling during the rest of the year. Ocean feedback reinforces the cooling in spring but counteracts the autumn and winter cooling. Vegetation feedback produces warming in all seasons, with largest changes (+1 °C) in spring. Synergy between ocean and vegetation feedbacks leads to further warming, which can be as large as the independent impact of these feedbacks. The combination of these effects causes the high northern latitudes to be warmer throughout the year in the ocean–atmosphere-vegetation simulation. Simulated vegetation changes resulting from this year-round warming are consistent with observed mid-Holocene vegetation patterns. Feedbacks also impact on precipitation. The atmospheric response to orbital-forcing reduces precipitation throughout the year; the most marked changes occur in the mid-latitudes in summer. Ocean feedback reduces aridity during autumn, winter and spring, but does not affect summer precipitation. Vegetation feedback increases spring precipitation but amplifies summer drying. Synergy between the feedbacks increases precipitation in autumn, winter and spring, and reduces precipitation in summer. The combined changes amplify the seasonal contrast in precipitation in the ocean–atmosphere-vegetation simulation. Enhanced summer drought produces an unrealistically large expansion of temperate grasslands, particularly in mid-latitude Eurasia.

1 Introduction

The high latitudes of the Northern Hemisphere have experienced significant warming during the latter part of the twentieth century, accompanied by a significant increase in ocean surface temperature, a decrease in sea-ice extent, a decrease in snow cover over the land, and an increase in vegetation cover and seasonal duration (Chapman and Walsh 1993; Martin et al. 1997; Myneni et al. 1997; Overpeck et al. 1997; Parkinson et al. 1999; Serreze et al. 2000; Zhou et al. 2001; Lucht et al. 2002). The observed changes in climate are consistent with climate-model predictions of the consequences of increased atmospheric CO2 concentrations, suggesting that greenhouse warming may already be underway in the high latitudes (Prentice et al. 2001). Climate-model predictions of the response to anthropogenic changes in atmospheric composition suggest that the Arctic is particularly sensitive to the change in radiative forcing (Cubasch et al. 2001) because of two powerful positive feedbacks: changes in the extent and duration of sea-ice cover in the Arctic Ocean (Curry et al. 1995; Hewitt et al. 2001; Vavrus and Harrison 2003), and changes in the albedo of the land surface as a consequence of changes in snow cover and the extent of forest, which masks the effect of snow cover to a large extent (Harvey 1988; Douville and Royer 1996; Bonan et al. 1992; Levis et al. 2000; Brovkin et al. 2003).

Much of our understanding of the importance of ocean and land-surface feedbacks on Arctic climates has resulted from the analysis of palaeoclimate simulations (see e.g. Foley et al. 1994; de Noblet et al. 1996; Gallimore and Kutzbach 1996; TEMPO-Members et al. 1996; Claussen and Gayler 1997; Texier et al. 1997; Broström et al. 1998; Ganopolski et al. 1998; Kubatzki and Claussen 1998; Levis et al. 1999; Vavrus 1999; Hewitt et al. 2001). These simulations have shown that ocean feedbacks and vegetation feedbacks both amplify the response to changes in orbital forcing during the mid-Holocene, and to ice-age boundary conditions (including the impact of lower atmospheric CO2 concentrations), and that synergy between ocean and land-surface feedbacks may cause a further amplification of the high latitude response. These conclusions appear to be robust across a range of experimental designs and different kinds of model, including a low-resolution but fully-coupled ocean–atmosphere-vegetation model (Ganopolski et al. 1998); further analysed by Claussen (2001). However, there is less agreement between the simulations about the relative magnitude of the two feedbacks, the strength of the synergy between them, and the degree to which the expression of the response to ocean- and land-surface feedbacks varies regionally. Thus, a more extensive analysis of the role of ocean and land-surface feedbacks and synergy appears warranted.

In this work, we present analyses of a pre-existing series of mid-Holocene experiments (Braconnot et al. 1999) with the Institute Pierre Simone Laplace (IPSL) coupled ocean–atmosphere general circulation model (Braconnot et al. 2000b), with the atmospheric component of that model, and with both the atmosphere and the coupled ocean–atmosphere models asynchronously coupled to the BIOME1 equilibrium vegetation model (Prentice et al. 1992). Comparison of these simulations allows us to diagnose the relative importance of direct orbital forcing, ocean feedbacks, vegetation feedbacks, and synergies between the ocean and vegetation in explaining observed regional environmental changes in the mid- to high-latitudes of the Northern Hemisphere during the mid-Holocene (6 ka). The choice of the mid-Holocene is motivated partly because 6 ka has been a major focus for palaeoclimate modelling (Joussaume and Taylor 1995, 2000) and partly because of the existence of a new synthesis of palaeoenvironmental evidence documenting vegetation changes in the mid- to high-latitudes at 6 ka compared to today (Bigelow et al. 2003).

Observed vegetation patterns at 6 ka as shown by these data were compared with the simulations in order to determine the comparative realism of each of the experiments and thus to assess whether the model–based estimates of the magnitude of individual feedbacks are realistic. To facilitate comparison with the palaeoenvironmental data, output from each climate experiment was used to drive an offline equilibrium biogeography model. We used the most recent version of the BIOME model (BIOME 4: Kaplan et al. in press) in preference to BIOME1 because it has a better discrimination of high–latitude vegetation types, and because it uses a vegetation classification that is compatible with the classification used by Bigelow et al. (2003).

2 Methods

2.1 The coupled ocean–atmosphere model

The atmospheric component of the coupled IPSLCM1 ocean–atmosphere general circulation model (OAGCM) is version 5.3 of the Laboratoire de Météorologie dynamique (LMD) atmosphere general circulation model (Sadourny and Laval 1984; Masson and Joussaume 1997). The grid resolution is 64 points in longitude, 50 points in the sine of latitude, and 11 vertical sigma levels. The model includes the land-surface scheme SECHIBA (Ducoudré et al. 1993). There are eight vegetation types in the SECHIBA scheme. The distribution of these vegetation types is prescribed either from observations as in the control run or, using the output of a vegetation model (i.e. BIOME1). When the climate model is asynchronously coupled to the equilibrium biome model, BIOME1 is used to prescribe vegetation distribution, thus, vegetation characteristics modulate the fluxes of momentum, latent and sensible heat, and evaporation between the surface and the atmosphere. SECHIBA computes the appropriate set of land-surface parameters (e.g. seasonally varying albedo, roughness length, canopy resistance) for each grid cell as a function of vegetation type and vegetation phenology using parameter values formulated explicitly for the 17 vegetation types recognised in BIOME1 to force SECHIBA (Texier et al. 1997).

The oceanic component of the coupled model is the general circulation model OPA developed at the Laboratoire d’Océanographie dynamique et de Climatologie (LODYC) (Madec et al. 1998). The horizontal resolution is 92 points by 76 points. The horizontal mesh is orthogonal and curvilinear on the sphere. The northern point of convergence is shifted over Asia to overcome the singularity at the North Pole (Madec and Imbard 1996). There are 31 vertical levels, with 10 levels in the upper 100 m. The turbulent diffusion is isopycnal–diapycnal, with a limitation of the isopycnal slopes to 1% (Guilyardi et al. 1999). The isopycnal diffusion coefficient is 2000 m2 s–1, with no background horizontal diffusion. Momentum and heat fluxes are computed separately for sea ice and ocean (Braconnot et al. 2000b).

The sea-ice model is a diagnostic prescribed sea-ice model (Braconnot et al. 1997), where an ocean grid box is assumed to be frozen when the sea-surface temperature (SST) falls below the freezing point of sea water. The heat fluxes from the ocean to the bottom of sea ice are prescribed as –2 Wm–2 in the Arctic (Maycut and Untersteiner 1971). Once frozen, the sea-surface temperature can only warm by heat advection and diffusion. The surface temperature and albedo of the sea-ice fraction are computed using a one-layer thermodynamic model and making the assumption that sea ice is 3 m thick. The temperature at the bottom of the sea ice is prescribed as 271.2 K. The grid resolution in the ocean is greater than in the atmosphere, so the area covered by sea ice represents a fraction in an atmospheric grid box.

Freshwater fluxes to the ocean are simulated from 46 major rivers. The outflow of these rivers contributes directly to the ocean model grid at the location of the corresponding river mouth.

2.2 The equilibrium biogeography models

Models of the BIOME family are equilibrium biogeography models which simulate the distribution of major vegetation types (biomes) as a function of the seasonal cycle of temperature, precipitation, sunshine and soil moisture conditions. The climate data used to run the model can either be derived from observations or, as here, from the output of climate model simulations.

BIOME1 predicts the distribution of 17 major biomes (Prentice et al. 1992). Biomes emerge through the combination of 14 plant functional types (PFTs). The distribution of PFTs is described in terms of tolerance thresholds for cold, heat, chilling and moisture requirements. Cold tolerance is expressed in terms of minimum mean temperature of the coldest month. The chilling requirement is formulated in terms of the maximum mean temperature of the coldest month. The heat requirement is expressed in terms of growing-degree-days (GDD) above a threshold of 5 °C (for trees) or 0 °C (for non-woody plants). However, the distinction between cool and warm grass/shrub is based on the mean temperature of the warmest month. Moisture requirements are expressed in terms of limiting values of the ratio of actual to equilibrium evapotranspiration. In each grid cell, the model selects the set of PFTs which could exist in the given climate and a dominance criterion is applied. Biome types are derived through combinations of dominant PFTs.

BIOME4 is the most recent of the BIOME model family and was explicitly developed in order to resolve a larger range of high-latitude vegetation types (Kaplan et al. in press). BIOME4 distinguishes 27 biomes, including five types of tundra vegetation (low and high shrub; erect dwarf-shrub; prostrate dwarf-shrub; cushion forb, lichen and moss; and graminoid and forb). The biomes are compatible with the classification used in the most recent compilation of palaeovegetation data from the high northern latitudes (the PAIN dataset: Bigelow et al. 2003). BIOME4 differs from BIOME1 because it explicitly simulates the coupled carbon- and water-flux cycles. It is thus able to treat competition between PFTs as a function of relative net primary productivity (NPP), using an optimisation algorithm to calculate the maximum sustainable leaf area (LAI) of each PFT and associated NPP.

For diagnostic purposes, BIOME4 was run using an anomaly procedure (i.e. not directly from the climate model output as in the BIOME1 simulations). The use of an anomaly procedure facilitates comparison with observations because it goes some way to minimising the impact of model biases, which are assumed to be similar in both the control run and the experiments (Harrison et al. 1998). In the anomaly procedure, differences in the 20-year climate averages of monthly mean precipitation, temperature and sunshine between each of the 6 ka experiments and the control experiment were linearly interpolated to the 0.5° grid of the BIOME4 model and then added to a modern climatology (CLIMATE 2.2.: http://www.pik-potsdam.de/~cramer/ ). Soil properties were specified from a data set derived from the FAO global soils classification (FAO 1995). BIOME4 is calibrated for a modern CO2 concentration of 324.6 ppm (the mean value for the interval covered by the CLIMATE 2.2 data set); the CO2 level in the diagnostic biome simulations was left unchanged at this value. The CO2 concentration was left unchanged in these experiments to parallel the situation in the climate experiments. Thus, both the climate experiments and the vegetation response to the simulated climate changes show only the impact of orbital changes.

2.3 Control simulation

The control simulation (0 ka control) was made using the coupled OAGCM with prescribed modern vegetation from climatology (Braconnot et al. 1999), and with modern orbital parameters for 1950 AD and vernal equinox fixed on 21st March. The atmospheric CO2 concentration was set to 345 ppm. The simulation was run for 150 years in coupled mode. The atmospheric conditions used to initiate the coupled integration were those of January 1st of the 16th year of a pre-existing atmosphere-only simulation forced with the mean seasonal cycle of SST and sea-ice cover specified from Reynolds (1988). The ocean model was spun up with annual mean forcing of wind stress, heat fluxes, and water fluxes. Stable surface conditions in the upper ocean, with a global mean SST of 17.8 °C, were achieved after 20 years of the coupled integration. The length of the simulation is not enough to bring the deep ocean to equilibrium, but is sufficient to study changes in the seasonal cycle, which mainly affect the upper ocean. However, the drift at depth is small and similar in all simulations. The Arctic sea-ice cover built during NH winter is underestimated by about 13%, with an area of 11.7 km × 106 compared to the 13.5 km × 106 reported by Gloersen and Campbell (1991).

2.4 Simulations of the 6 ka climate

To examine the response of the climate system to changes in orbital parameters at 6 ka we use four experiments from Braconnot et al. (1999), (Fig. 1). The first experiment was a coupled OAGCM simulation (OA 6 ka) in which only the orbital parameters were changed to those appropriate for 6 ka (Berger 1978). This simulation was run for 150 years. The second experiment (A 6 ka) was made using the atmospheric component of the model alone. The A 6 ka simulation was then run for 20 years with SSTs and sea-ice distributions were specified daily using 20 years of data extracted from years 80 to 100 of the control simulation. Comparison of this experiment and the OA 6 ka experiment allows the role of the oceanic feedback at 6 ka to be assessed. The third experiment (OAV 6 ka) was a coupled ocean–atmosphere simulation in which vegetation distribution at 6 ka was specified from the last of three iterations from the BIOME1 simulation, driven by the 20 years average climate data from year 80 to 100 of the OA 6 ka experiment. This experiment was run for 50 years. The atmospheric CO2 concentration was set to the same value as in the control simulation (345 ppm) in all of the 6 ka experiments. This prescription is not realistic (the CO2 level at 6 ka was ca 270 ppm) but allows the experiments to be treated as investigations of the response to orbital forcing only. The final experiment (AV 6 ka) was made using the atmospheric component of the model alone; in this experiment, vegetation distribution at 6 ka was specified from the BIOME1 simulation used for the OAV 6 ka experiment. Daily SSTs and sea-ice distribution were specified from the control simulation, as in the A 6 ka experiment. The AV 6 ka experiment was then run for 20 years. Comparison of this experiment with the A 6 ka experiment allows the role of the vegetation feedback at 6 ka to be assessed. However because the vegetation in the AV 6 ka experiment is derived from the OA 6 ka experiment, we are unable to separate out a pure vegetation feedback signal from these experiments. Nevertheless, there is a significantly different response in the AV 6 ka experiment that is attributable to vegetation and thus it seems worthwhile to compare these experiments.
Fig. 1.

Flowchart showing the experimental set-up for the control and 6 ka climate simulations

Analyses of the simulations were based on 20-year averages. The 20-year averages were derived from years 80 to 100 of the control 0 ka and OA 6 ka experiments, and the last 20 years of the A 6 ka, AV 6 ka and OAV 6 ka experiments. The present results are based on the modern calendar. Use of celestial calendar months might provide a more accurate representation of Holocene climate changes (Kutzbach and Gallimore 1988; Joussaume and Braconnot 1997) but adoption of this calendar would not affect the conclusions of our analyses.

2.5 Analytical protocol for the derivation of the feedback and synergy components of the simulated 6 ka climatic changes

The relative importance of ocean and vegetation feedbacks and synergy on the seasonal cycle of temperature and precipitation over the continents (north of 40°N) was assessed by comparing the simulated mean climate (based on the 20-year averages), from each of the 6 ka climate simulations (Stein and Alpert 1993; Kubatzki et al. 2000; Berger 2001) as follows:

The indexf is the sum of all components, thus:
$$({\text{A}}6{\text{ka}} - {\text{ctr}}) = {\text{dA}} = \Delta {\mathbf{A}}_{{\mathbf{f}}} $$
$$({\text{OA}}6{\text{ka}} - {\text{ctr}}) - {\text{A}}6{\text{ka}} - {\text{ctr}}) = {\text{dA}} + {\text{dO}} + {\text{dOA}} - {\text{dA}} = {\text{dO}} + {\text{dOA}} = \Delta {\mathbf{O}}_{{\mathbf{f}}} $$
$$ (\text{AV}6\text{ka} - \text{ctr}) - (\text{A}6\text{ka} - \text{ctr}) = \text{dA} + \text{dV} + \text{dAV} - \text{dA} = \text{dV} + \text{dAV} = \Delta \mathbf{V}_{\mathbf{f}} $$
$$ \begin{aligned} & ({\text{OAV}}6{\text{ka}} - {\text{ctr}}) - ({\text{A}}6{\text{ka}} - {\text{ctr}}) - (({\text{OA}}6{\text{ka}} - {\text{ctr}}) - ({\text{A}}6{\text{ka}} - {\text{ctr}})) \\ & \quad - (({\text{AV}}6{\text{ka}} - {\text{ctr}}) - ({\text{A}}6{\text{ka}} - {\text{ctr}})) = ({\text{dA}} + {\text{dO}} + {\text{dV}} + {\text{dOA}} + {\text{dOV}} + {\text{dAV}} + {\text{dOAV}}) \\ & \quad - ({\text{dA}}) - ({\text{dO}} + {\text{dOA}}) - ({\text{dV}} + {\text{dAV}}) = {\text{dOAV}} = \Delta {\mathbf{S}}_{{\mathbf{f}}} \\ \end{aligned} $$

2.6 Palaeovegetation data for 6 ka

The Pan–Arctic Initiative (PAIN) has made reconstructions of vegetation patterns at 6 ± 0.5 ka across the high northern latitudes (north of 55°N) based on pollen records from individual sites using a standard procedure (Bigelow et al. 2003). This reconstruction represents the most extensive compilation of palaeovegetation data from high northern latitudes currently available, and has the merit of using a classification scheme that was designed to be compatible with the scheme used in BIOME4. There are 493 sites in the PAIN 6 ka reconstruction. Reconstructions of vegetation patterns to the south of the PAIN window have been made as part of the Palaeovegetation Mapping Project (BIOME 6000: Prentice and Webb III 1998; Prentice et al. 2000). There are reconstructions from 757 sites between 40° and 55°N in the BIOME 6000 data set (Prentice et al. 1996; Tarasov et al. 1998; Edwards et al. 2000; Williams et al. 2000). Although the BIOME 6000 data set does not discriminate different tundra types, the classification of temperate and boreal forest types used in the BIOME 6000 data set is compatible with the forest classification used in PAIN. Thus, in the region between 40° and 55°N it is possible to amalgamate the two data sets without needing to make adjustments to the biome designations. There are 1250 sites in the combined data set (Fig. 2). Some 245 of these sites are not used in the data-model comparisons: some of these sites (125) are poorly dated (DC = 7 using the COHMAP dating control scheme: Yu and Harrison 1995) and some sites lie along the coast, on islands or in inland lake areas (147) where the BIOME4 model has sea, lakes and/or rivers and thus there is no simulated vegetation for comparison. Three sites that, according to the original vegetation reconstruction, were misclassified as tundra (Prentice et al. 2000) have also been omitted from the comparison.
Fig. 2.

Vegetation distribution a today and b at 6 ka, as reconstructed from pollen records at individual sites. The data north of 55°N are derived from the PAIN data set (Bigelow et al. 2003) and the data between 40–55°N from the BIOME 6000 data set (Prentice et al. 2000)

These data were used to evaluate the 6 ka palaeoclimate simulations by visual comparison. A more quantitative evaluation of the simulations was made by counting the number of matches between observed and simulated vegetation for the 0.5° grid cell within which each data point is located. In order to simplify this comparison, we adopted a classification into major biomes proposed by Harrison and Prentice (2003) and summarised in Table 1.
Table 1.

Definition of a simplified scheme of major biomes used for comparisons between simulated and observed biomes at 6 ka. The scheme is derived from Harrison and Prentice (2003)

Biome classification used in BIOME4

Major biome classification used for summary statistics in this work

 

Tropical evergreen broadleaf forest

Tropical forest (not present in region N of 40°N)

 

Tropical semi-evergreen broadleaf forest

 

Tropical deciduous broadleaf forest and woodland

 

Warm-temperate evergreen broadleaf and mixed forest

Warm-temperate forest

 

Temperate deciduous broadleaf forest

Temperate forest

 

Temperate evergreen needleleaf forest

 

Cool mixed forest

 

Cool evergreen needleleaf forest

 

Cool-temperate evergreen needleleaf and mixed forest

 

Cold evergreen needleleaf forest

Boreal forest

 

Cold deciduous forest

 

Tropical savanna

Savanna and dry woodland

 

Temperate sclerophyll woodland and shrubland

 

Temperate deciduous broadleaf savanna

 

Temperate evergreen needleleaf open woodland

 

Tropical xerophytic shrubland

Grassland and dry shrubland

 

Temperate xerophytic shrubland

 

Tropical grassland

 

Temperate grassland

 

Desert

Desert

 

Barren

 

Graminoid and forb tundra

Dry tundra

 

Low and high shrub tundra

Tundra

 

Erect dwarf-shrub tundra

 

Prostrate dwarf-shrub tundra

 

Cushion-forb tundra

 

3 Results

3.1 The seasonal cycle of air temperature

The atmospheric response (ΔAf = (A 6 ka – 0 ka ctrl)) to orbital forcing produces a warming of the mid- and high-latitude (N of 40°N) continents of 1.15 °C during the summer (June, July, August = JJA) and a cooling during the rest of the year (Fig. 3; Table 2). The cooling is largest during winter (–0.61 °C) and spring (–0.66 °C). As a result, there is no change in mean annual temperature compared to today.
Fig. 3.

Changes in the seasonal cycle of temperature (°C) and precipitation (mm/day) as shown in each of the 6 ka experiments compared to the control OA experiment

Table 2.

Summary of seasonal changes in temperature (°C) and precipitation (mm/day) by latitude zones

Region

Season

A 6 ka

OA 6 ka

AV 6 ka

OAV 6 ka

         

T

P

T

P

T

P

T

P

         

N of 40°N

DFJ

–0.61

–0.01

–0.24

0.03

–0.31

–0.01

0.57

0.08

         

MAM

–0.66

–0.03

–0.77

–0.02

0.29

–0.01

0.44

0.05

         

JJA

1.15

–0.01

1.30

–0.01

1.50

–0.04

2.13

–0.05

         

SON

–0.28

–0.01

0.51

0.06

–0.10

–0.02

1.00

0.09

         

Annual

–0.10

–0.02

0.20

0.02

0.34

–0.02

1.04

0.04

         

N of 70°N

DFJ

0.00

–0.01

0.39

0.02

–0.33

–0.02

0.92

0.06

         

MAM

–0.24

–0.01

–0.54

–0.01

0.16

0.02

0.07

0.04

         

JJA

0.71

0.02

0.97

0.04

0.79

–0.01

1.29

–0.01

         

SON

0.05

0.00

0.89

0.09

0.08

0.00

1.21

0.08

         

Annual

0.13

0.00

0.43

0.04

0.18

0.00

0.87

0.04

         

60–70°N

DFJ

–0.61

–0.02

–0.41

0.03

–0.46

–0.01

0.24

0.05

         

MAM

–0.61

–0.05

–0.92

–0.02

0.52

0.00

0.47

0.07

         

JJA

1.24

–0.01

1.25

0.01

1.56

0.03

2.09

0.07

         

SON

–0.46

–0.01

0.33

0.09

–0.27

0.02

0.76

0.14

         

Annual

–0.11

–0.02

0.06

0.03

0.33

0.01

0.89

0.08

         

50–60°N

DFJ

–0.93

–0.02

–0.61

0.01

0.07

–0.02

0.77

0.11

         

MAM

–0.92

–0.05

–0.94

–0.02

0.74

–0.01

1.06

0.09

         

JJA

1.44

0.00

1.53

0.00

1.97

–0.05

2.83

–0.07

         

SON

–0.68

–0.01

0.14

0.04

–0.32

–0.01

0.73

0.11

         

Annual

–0.27

–0.02

0.03

0.01

0.61

–0.02

1.35

0.06

         

40–50°N

DFJ

–1.12

0.01

–0.49

0.06

–0.53

0.00

0.29

0.14

         

MAM

–1.04

0.00

–0.67

–0.03

–0.37

–0.08

0.22

–0.01

         

JJA

1.34

–0.05

1.57

–0.11

1.90

–0.16

2.62

–0.23

         

SON

0.01

–0.06

0.67

–0.02

0.13

–0.11

1.33

0.01

         

Annual

–0.20

–0.03

0.27

–0.03

0.28

–0.80

1.11

–0.02

         
Ocean feedbacks (ΔOf = (OA 6 ka – 0 ka ctrl l) – (A 6 ka – 0 ka ctrl)) amplify the response to 6 ka orbital forcing in spring (–0.11 °C) and summer (0.14 °C) and counteract the direct response to orbital forcing in autumn and winter (Fig. 4). This response largely reflects the time lag (1–2 months) introduced by the thermal inertia of the oceans (Le Clainche 2000). The warming caused by ocean feedback in the autumn is large (+0.79 °C), resulting in conditions +0.51 °C warmer in the OA 6 ka simulation than in the 0 ka control simulation. The winter warming caused by ocean feedback is smaller (+0.37 °C): as a result winters remain slightly colder than present (–0.24 °C) in the coupled ocean–atmosphere simulation (although warmer than in the atmosphere-only simulation). As a result of the ocean feedback, mean annual temperature over the high northern latitudes is +0.30 °C warmer in the OA 6 ka experiment than in the A 6 ka simulation and +0.20 °C warmer than in the control.
Fig. 4.

The relative contribution of the atmospheric response, ocean and vegetation feedbacks, and the synergy between these feedbacks, to the changes in the seasonal cycle of temperature (°C) and precipitation (mm/day) at 6 ka as simulated in the OAV experiment

Vegetation feedback (ΔVf = (AV 6 ka – 0 ka ctrl) – (A 6 ka – 0 ka ctrl)) results in enhanced warming in all seasons (Fig. 4) compared to the atmospheric response to orbital forcing. The largest changes due to vegetation feedback occur in spring (+0.95 °C), the smallest (+0.17 °C) in the autumn. This reflects the fact that changes in vegetation, and particularly the simulated extension of forest cover, modify snow albedo and affect the timing of snow melt. As a result of the vegetation-induced enhancement, simulated temperatures in the AV 6 ka experiment are warmer than today in spring and summer (Fig. 3; Table 2). Thus, vegetation feedbacks amplify orbital forcing in summer and offset the orbitally induced cooling in the spring, but have little impact during the remaining seasons.

Synergy (ΔSf = (OAV 6 ka – 0 ka ctrl) – (A 6 ka – 0 ka ctrl) – ((OA 6 ka – 0 ka ctrl) – (A 6 ka – 0 ka ctrl)) – ((AV 6 ka – 0 ka ctrl) – (A 6 ka – 0 ka ctrl))) between the ocean and vegetation feedbacks results in warming in all seasons (Fig. 4). The magnitude of the warming caused by this synergy is largest in winter (+0.52 °C) and in summer (+0.49 °C). Indeed, in these two seasons the warming due to synergy is as large as the warming due to the combination of ocean and vegetation feedbacks when considered separately. The combined effect of ocean and vegetation feedbacks considered independently is to warm winter by 0.66 °C and summer by 0.48 °C.

The combined effects of orbital forcing, ocean feedback, vegetation feedback and the synergy between the ocean and vegetation feedbacks produce temperatures about 0.5 °C warmer than present during the winter and spring, ca 2 °C warmer than present during summer, and ca 1 °C warmer than present during autumn in the OAV 6 ka simulation (Fig. 3; Table 2). The importance of the contribution of the various effects to the overall warming varies from season to season. Thus, summer warming is dominated by the direct response to orbital forcing, spring warming is dominated by the impact of vegetation feedback, autumn warming is dominated by the impact of ocean feedback, and winter warming is dominated by the synergy between the ocean and the vegetation feedbacks (Fig. 5).
Fig. 5.

Seasonal contribution of the atmospheric response, ocean feedback, vegetation feedback, and the synergy between ocean and vegetation feedbacks to a the overall year-round warming (°C) and b the seasonal changes in precipitation (mm/day) shown in the OAV 6 ka experiment. The outer boundary of the grey area represents the simulated seasonal values of temperature and precipitation in the control simulation

3.2 The seasonal cycle of precipitation

The direct response to orbital forcing results in a decrease in precipitation in the mid- and high northern latitudes in every season (Fig. 3; Table 2). The largest changes, however, occur in spring (–0.03 mm/day). Ocean feedback leads to increased precipitation in the autumn, winter and spring (Fig. 4), offsetting the orbitally-induced tendency towards aridity (Le Clainche 2000), but has little effect in summer. As a result, mean annual precipitation is higher (0.04 mm/day) in the OA 6 ka simulation than in the A 6 ka simulation. Vegetation feedback increases precipitation in spring (+0.02 mm/day) and reduces precipitation in summer (–0.03 mm/day) but has no discernible effect on autumn and winter precipitation (Table 2). Thus, the impact of vegetation feedback is to partially offset orbitally-induced drying in spring and to augment orbitally induced drying in summer (Fig. 5b). As a result, mean annual rainfall in the AV 6 ka simulation is –0.02 mm/day less than today but similar to the mean annual rainfall as a result of orbital-forcing alone. The synergy between ocean and vegetation feedbacks has a large positive impact on precipitation in autumn (+0.03 mm/day), winter (+0.06 mm/day) and spring (+0.05 mm/day) and a smaller negative impact on summer precipitation (–0.01 mm/day) (Fig. 5b). As a result, the seasonal contrast in precipitation is much larger in the OAV 6 ka simulation (Fig. 3; Table 2) than today, with drier summer conditions and more precipitation during the remainder of the year.

3.3 Regional temperature and precipitation responses to changes in orbital forcing and feedbacks

The simulated temperature changes in response to orbital forcing, ocean and vegetation feedbacks, and the synergies between these feedbacks are not zonally uniform across the high northern latitudes (Fig. 6; Table 2). The strength of the orbitally induced winter cooling is maximal in the mid–latitudes (40–50°N). The winter warming caused by vegetation feedback is also large in this latitude band (+0.59 °C) although maximal (+0.99) between 50–60°N. Winter warming as a result of ocean feedback is relatively large (+0.39 °C) along the Arctic coast (N of 70°N) but larger (+0.63 °C) between 40–50°N. The impact of the synergy between ocean and vegetation feedbacks on winter warming is largest in the far north. As a result, the latitudinal pattern of winter warming in the OAV 6 ka simulation has two maxima (Fig. 6), with a warming of +0.92 °C occurring in the high northern latitudes along the Arctic coast and a secondary maximum between 50–60°N (+0.77 °C).
Fig. 6.

The relative contribution of the atmospheric response, ocean and vegetation feedbacks, and the synergy between these feedbacks, to the changes in the seasonal cycle of temperature (°C) and precipitation (mm/day) at 6 ka as simulated in the OAV experiment for different latitudinal bands. The change in the seasonal cycle of temperature and precipitation in the OAV 6 ka experiment compared to the modern control for each latitudinal band is shown for comparison

The situation in summer is simpler (Fig. 6; Table 2). Direct orbital forcing, vegetation feedback, and the synergistic effects of ocean and vegetation feedbacks all produce maximum warming in the latitude band between 50–60°N. The contribution of the ocean to summer warming, which is maximal at high northern latitudes, is small (+0.27 °C). As a result, the maximum summer warming in the OAV 6 ka experiment (+2.83 °C) occurs between 50–60°N.

The simulated precipitation changes in response to orbital forcing, ocean and vegetation feedbacks, and the synergies between these feedbacks are also not zonally uniform (Fig. 6; Table 2). Precipitation changes in response to orbital forcing alone (A 6 ka) are negligible in the far north. However, precipitation is decreased compared to present further south. Between 50–70°N, the overall decrease in mean annual precipitation (–0.02 mm/day) results from changes during winter and spring while in the temperate zone (40–50°N) the overall decrease (–0.03 mm/day) occurs largely as a result of decreased rainfall during the summer and autumn. Ocean feedback enhances precipitation everywhere north of 50°N and in most seasons, with the largest changes in autumn. However, although the ocean feedback enhances autumn and winter rainfall in the temperate zone, this tendency is offset by decreased rainfall during the spring and summer. Thus, the impact of the ocean feedback is to create wetter conditions over most of the region while amplifying spring (–0.03 mm/day) and summer (–0.06 mm/day) drought in the temperate zone (40–50°N). The impacts of vegetation feedback on precipitation are spatially strongly differentiated (Fig. 6). In the far north, vegetation feedback enhances precipitation in spring (+0.03 mm/day) but decreases precipitation in summer (–0.02 mm/day); as a result, there is little change in annual mean precipitation. Between 60–70°N, vegetation feedback enhances precipitation in all seasons leading to an increase in mean annual precipitation (+0.03 mm/day). To the south (50–60°N), seasonal changes in precipitation as a consequence of vegetation feedback offset each other and there is no overall impact on mean annual precipitation. The largest changes resulting from vegetation feedback occur in the temperate zone where precipitation is decreased in all seasons and very strongly during the spring (–0.07 mm/day), summer (–0.10 mm/day) and autumn (–0.06 mm/day). These changes are caused by the large extent of grassland simulated in response to the pronounced spring and summer drought in the OA 6 ka experiment and used as input in the AV 6 ka experiment. Synergy between ocean and vegetation has a positive impact on precipitation at all latitudes in autumn, winter and spring. However, in the summer months, this synergy tends to produce a small reduction in precipitation across most latitude bands.

Despite the apparent complexity of the changes in precipitation in response to individual components, the overall result of orbital forcing, ocean and vegetation feedbacks, and the synergy between these feedbacks (as shown in the OAV 6 ka simulation), is relatively straightforward (Fig. 6; Table 2). Autumn and winter precipitation are increased at all latitudes, spring precipitation is also increased except in the extreme south, summer precipitation is increased between 60–70°N but is decreased elsewhere and substantially decreased south of 60°N.

3.4 Changes in vegetation patterns in response to the simulated climate changes

Changes in vegetation patterns in response to simulated climate changes were derived using BIOME4 and an anomaly procedure described already. There are large changes in vegetation distribution in response to the orbitally induced changes in climate simulated by the atmosphere-only model (Table 3). About 30% of the area north of the 40°N has a different vegetation type compared to the modern situation. The changes affect three major aspects of Northern Hemisphere vegetation patterns: the location of the tundra-forest boundary, the northern limits of temperate and cool forests, and the extent of grassland and shrubland in the interior of the continents. These features of the Northern Hemisphere vegetation distribution are governed by different aspects of climate. The limit between tundra and forest is controlled by accumulated temperatures during the growing season, the northern limits of temperate and cool forests by minimum temperature in winter, and the extent of grassland versus forest by growing-season aridity (Woodward 1987; Prentice et al. 1992; Kaplan et al. in press). In response to orbital forcing, the tundra-forest boundary is shifted northward across much of the Arctic and the overall area of tundra is reduced by 22.7% compared to the modern state. This change is a result of the simulated summer warming and longer growing season in the A 6 ka experiment. The northward shift is most pronounced in eastern and central Siberia, and in central and eastern Canada, where the high-latitude summer warming is greatest. There is no discernible change in the tundra-forest boundary in Beringia, because the high-latitude summer warming there is negligible. The northern limits of temperate and cool forests are further north than today, particularly in Europe and western Siberia and in eastern North America. The regions affected lie in the zone of maximum winter warming between ca 50–60°N (Fig. 6). The impact on Europe and western Siberia is larger than elsewhere because warm-air advection from the Atlantic produces a large increase in winter temperature across this region. The largest vegetation changes shown in the A 6 ka simulation are associated with expansion of grassland vegetation (91.6% increased compared to today) in the interior of Eurasia. According to these simulations, grassland replaced temperate and boreal forest in much of central and northern China, across central Asia and in the region north of the Caspian and Aral Seas. The expansion of grassland reflects the reduction in precipitation during spring, summer and autumn in the mid-latitudes, and particularly in the Asian sector where mean annual precipitation is reduced by 0.04 mm/day.
Table 3.

Areas of major biomes (in 106 km2) in the modern BIOME4 simulation and as a result of each of the 6 ka experiments

Modern

A 6 ka

OA 6 ka

AV 6 ka

OAV 6 ka

     

Key individual biomes

     

  Cold evergreen needleleaf forest

13.39

12.59

12.56

11.04

10.02

     

  Cold deciduous forest

4.00

4.75

4.13

4.23

3.39

     

  Temperate deciduous broadleaf forest

2.40

3.15

2.96

3.04

3.46

     

  Temperate grassland

3.14

6.01

6.43

9.39

11.25

     

Major biomes

     

  Warm-temperate forest

12.12

10.64

10.65

9.61

10.03

     

  Temperate forest

0.25

0.21

0.21

0.23

0.23

     

  Boreal forest

17.39

17.33

16.69

15.28

13.41

     

  Savanna and dry woodland

0.75

0.76

0.86

0.68

0.93

     

  Grassland and dry shrubland

5.21

8.43

8.91

12.20

14.21

     

  Desert

1.70

1.63

1.61

1.63

1.53

     

  Dry tundra

0.23

0.22

0.18

0.15

0.13

     

  Tundra

6.77

5.19

5.31

4.63

3.95

     

Summary

     

  All tundra

7.00

5.41

5.49

4.79

4.07

     

  All forest

29.76

28.18

27.55

25.12

23.67

     

There are only small changes in vegetation distribution as a result of ocean feedback: only 12% of the area north of the 40°N has a different vegetation type in the OA 6 ka simulation from the A 6 ka simulation. There is little discernible change in the position of the tundra-forest boundary, except in the central part of eastern Beringia where the area of tundra was slightly larger than it is today because ocean feedback produces a cooling (–0.49 °C) in this region. The comparatively small changes in the tundra-forest boundary reflect the fact that the major impact of ocean feedback in the high northern latitudes occurs during winter; the simulated cooling in spring and warming in summer are small and do not substantially affect the length of the growing season. The impact of ocean feedback on the northern limits of temperate forest is spatially heterogeneous: in eastern North America, the simulated winter warming results in a more northerly position of the northern limit of temperate deciduous broadleaf forest than today. In contrast, the northern limit of temperate deciduous broadleaf forest is further south than today in Europe as a consequence of the expansion of cool mixed forest. Changes in the extent of cold deciduous forest also reflect changes in winter temperature: cold deciduous forest is reduced in extent in eastern Siberia because of the simulated warmer winters while in Beringia, where simulated winters are considerably colder than today, the cold deciduous forest is more extensive in the OA 6 ka experiment. In the mid–continent, and particularly in the Asia sector, there is a further expansion of grassland in the OA 6 ka experiment (105% increased compared to modern) compared to the A 6 ka experiment (91.6% increased compared to present-day). This expansion reflects the fact that ocean feedback amplifies summer drought in the temperate zone (40–50°N).

Vegetation feedback (ΔVf) has a much larger impact on simulated vegetation patterns than ocean feedback (ΔOf). Over 43% of the area north of the 40°N has a different vegetation type compared to the modern situation (ca 25% of the area is different from the A 6 ka simulation). The tundra-forest boundary is further north than today in central Siberia and in Europe in the AV 6 ka simulation. However, there is no significant difference in the position of this limit compared to the A 6 ka simulation. This reflects the fact that vegetation feedback has no impact on summer and autumn temperatures at high northern latitudes (Fig. 6). The simulated northern limits of temperate and cool forests, particularly in the European sector, are further north than they are today, and further north than they would be in response to orbital forcing alone. This situation reflects the fact that the impact of vegetation feedback on winter temperature is maximal between 50–60°N. However, the most noticeable impact of vegetation feedback is on the extent of grassland. Grasslands cover an area of 9.3*106 km2 in the AV 6 ka simulation (compared to 3.1*106 km2 today: Table 3), extending north of 60°N in central Asia and in central North America. The expansion of grassland vegetation in eastern North America creates a disjunction between the temperate and boreal forests in the AV 6 ka simulation. This expansion results from the considerable decrease in summer precipitation south of 60°N in consequence of vegetation feedback, which is further amplified in the temperate zone (40–50°N) by decreases in spring and autumn precipitation.

Vegetation changes occur over 52% of the area north of 40°N in response to the climate changes simulated in the OAV 6 ka simulation. Most of the additional change (compared to the AV 6 ka simulation) is due to a further expansion of grassland (258% increased compared to present-day, whereas the AV 6 ka increased 200% relative to the modern situation). This expansion reflects the synergy between ocean and vegetation feedbacks which suppresses precipitation in the region south of ca 60°N and particularly in the Asia sector. The northern limit of the temperate deciduous broadleaf forest in Europe is at its most northerly position in all the simulations, reflecting the additive effect of orbital forcing, ocean and vegetation feedbacks, and the synergy between them on winter temperatures in Europe between 50–60°N.

3.5 Comparison of observed and simulated vegetation patterns at 6 ka

Palaeoenvironmetal reconstructions show the tundra-forest boundary was north of its present position in some regions at 6 ka but the pattern of this shift was strongly asymmetrical around the North Pole, with the largest northward shift in central Siberia (ca 200 km), a smaller northward shift in Fennoscandia, little change in Beringia and a southward shift of ca 200 km in Keewatin and Labrador (Fig. 7; Bigelow et al. 2003). More extensive geographic changes are seen in the boundaries among forest types south of the Arctic tree line, including a northward displacement of temperate deciduous forest in eastern North America and especially in Western Europe, and a more restricted distribution of cold evergreen needle-leaf forest towards the western edges of the continents (Western Europe and eastern Beringia). Expansion of drought-tolerant vegetation, including temperate grasslands and xerophytic shrublands occurred in North America (Harrison et al. 2003). However, the pollen-based reconstructions of 6 ka vegetation show no expansion of drought-tolerant biomes in the continental interior of Eurasia (Fig. 7; Tarasov et al. 1998). This finding is consistent with independent evidence based on geomorphic and biostratigraphic records of changes in lake status that show little or no change in the regional water balance of central Eurasia between the mid-Holocene and present (Harrison et al. 1996).
Fig. 7.

Vegetation patterns simulated using BIOME4 a under modern climate conditions and as a consequence of the changes in climate simulated in response to b orbital forcing only (A 6 ka), c orbital forcing and ocean feedbacks (OA 6 ka), d orbital forcing and vegetation feedbacks (AV 6 ka) and e orbital forcing, ocean and vegetation feedbacks and their synergy (OAV 6 ka). The simulated vegetation patterns can be directly compared with f reconstructed vegetation patterns (also shown in Fig. 2)

The position of the tundra-forest boundary is relatively well placed in all of the 6 ka simulations (Fig. 7). About 75% of the sites north of 70°N are correctly predicted in the A 6 ka simulation. Vegetation feedback has no discernable impact on summer temperature at high latitudes. The incorporation of ocean feedback, which produces a slight summer warming at high northern latitudes, improves the simulation of high-latitude biomes. In both the OA 6 ka and OAV 6 ka simulations, 81% of the sites north of 70°N are correctly predicted (Table 4).
Table 4.

Comparison between observed and simulated 6 ka distributions of major biomes (as defined in Table 1). The number of observed points used in each comparison is indicated. When there are several observed sites in a single model grid cell, all of the sites are compared independently with the simulated vegetation pattern

Area used for comparison

A 6 ka

OA 6 ka

AV 6 ka

OAV 6 ka

No. of sites

     

All land N of 40°N

70.0

69.1

62.1

61.9

1005

     

N of 70°N

75.0

81.3

75.0

81.3

16

     

60–70°N

61.4

61.0

57.7

59.3

241

     

50–60°N

76.1

75.7

70.3

64.5

259

     

40–50°N

71.0

69.1

59.5

61.1

489

     

The simulated occurrence of temperate and cool-temperate forests north of their present position at 6 ka, particularly in Europe, is supported by the data. Vegetation feedback appears to play an important role in the correct simulation of these boundaries: the position of both the temperate deciduous broadleaf forest and the cool mixed forest in Europe is more correctly placed in the AV 6 ka and OAV 6 ka simulations than in the A 6 ka simulation, while the southward displacement and fragmentation of the temperate deciduous broadleaf forest shown in the OA 6 ka simulation is unrealistic.

The simulation of expanded grasslands in the mid-latitudes of Asia in response to 6 ka orbital forcing is not supported by the data, nor is the massive expansion of these grasslands engendered by ocean and vegetation feedbacks and the synergy between them. Only 53% of the 114 sites between 40–60°N in the Asian sector are correctly predicted in the A 6 ka simulation, and this number is reduced to 37% in the AV 6 ka, and 36% in the OAV 6 ka simulations. The simulation of expanded grasslands in mid-continental North America in response to orbital forcing is credible (Harrison et al. 2003). However, the expansion of grassland in eastern North America in the region north of the temperate forests, shown in both the AV 6 ka and the OAV 6 ka simulations, is unrealistic (Williams et al. 2000). As a result, the agreement between observed and simulated vegetation between 40–50°N in eastern North America is reduced from 85% in the A 6 ka simulation to 68% in the OAV 6 ka simulation.

4 Discussion

The mean annual, seasonal and all individual months in the OAV 6 ka experiment are warmer than today. Temperatures are at least +2.1 °C warmer in summer, and +0.6 °C warmer in winter. Winter minimum temperatures increase more than the maximum temperatures. Mean annual precipitation is increased, chiefly because of large increases in winter (+0.08 mm/day). These changes are a direct response to orbital changes in the mid-Holocene, and of ocean and vegetation feedbacks. The atmosphere response to insolation changes leads to warming in summer and cooling and drying in other seasons. The ocean feedback counteracts the orbitally induced cooling and drying in autumn and winter. Vegetation feedback induces significant warming in spring and enhances summer drought. Synergies between ocean and vegetation feedbacks increase temperature all through the year, though the impact is largest in winter, and increases precipitation mainly in winter and spring. As a result of these changes in climate, the simulated vegetation cover over more than 50% of the land area north of 40°N is different from today.

According to our simulations, the atmospheric response to orbital forcing is the most important cause of summer warming in the northern latitudes. The combined impact of the feedbacks, however, causes an additional warming equivalent to ca 85% of the atmospheric response (Table 5). Ocean feedback by itself is less important, contributing only 14% of the additional warming due to feedback. Vegetation feedbacks, and the synergy between vegetation and ocean feedbacks, are responsible for 35% and 51% of the additional warming respectively. The feedbacks effectively reverse the impact of orbitally forced winter cooling in our simulations. Ocean feedback appears to play a more important role in winter warming than the vegetation feedback, although again the synergy between these feedbacks is more important than either alone.
Table 5.

Comparison of changes (temperature in °C; precipitation in mm/d) due to orbital forcing, ocean and vegetation feedbacks and the synergy between them, as simulated by the IPSL OAGCM asynchronously coupled to BIOME1 and by the CLIMBER model, for the land area N of 40°N. Results from the CLIMBER model are derived from the simulations made by Ganopolski et al. (1998)

IPSL-BIOME NHland (N of 40°N)

CLIMBER NHland (N of 40°N)

IPSL-BIOME NHland (N of 40°N)

CLIMBER NHland (N of 40°N)

        

Temperature summer (JJA)

Temperature winter (DJF)

Temperature summer (JJA)

Temperature winter (DJF)

Precipitation summer (JJA)

Precipitation winter (DJF)

Precipitation summer (JJA)

Precipitation winter (DJF)

        

A

1.2

–0.6

2.1

–0.6

–0.01

–0.01

0.18

–0.04

        

OA

1.3

–0.2

1.6

–0.3

–0.01

0.03

0.15

–0.03

        

AV

1.5

–0.3

2.7

–0.5

–0.04

–0.01

0.29

–0.03

        

OAV

2.1

0.6

3.2

0.8

–0.05

0.08

0.33

0.03

        

O feedback

0.1

0.4

–0.5

0.3

0

0.04

–0.03

0.01

        

V feedback

0.4

0.3

0.6

0.1

–0.03

0

0.11

0.01

        

S (synergy)

0.5

0.5

0.9

1.0

–0.01

0.06

0.07

0.05

        

There are differences in the modelling setup and experimental design that mean the two sets of simulations are not absolutely comparable. CLIMBER is a fully-coupled OAVGCM and the simulations are run to equilibrium. The IPSL-BIOME results are obtained with asynchronously coupling, and the A, and AV simulations incorporate some elements of the ocean feedback (unlike CLIMBER, see text). The CLIMBER simulations are run with CO2 at 280 ppm, whereas the IPSL-BIOME simulations were run with CO2 at 345 ppm

According to our simulations, vegetation feedbacks considerably amplify the orbitally induced reduction in summer precipitation. The magnitude of this effect is more than three times the initial atmospheric response (Table 5). Although ocean feedback alone appears to have no impact on summer precipitation, the synergy between ocean and vegetation feedbacks causes a reduction in precipitation of the same order of magnitude as the atmospheric response. Comparisons with palaeovegetation data suggest that the strength of the vegetation feedback is unrealistic: there is no evidence for massive aridity in mid-continental Eurasia as shown in the IPSL-BIOME simulations. Vegetation feedback plays no role in the simulated increase in winter precipitation, which is overwhelmingly driven by ocean feedback.

The role of ocean and vegetation feedbacks on Arctic climate has also been investigated by Ganopolski et al. (1998) using a series of experiments with the CLIMBER model. The two models are very different in structure: CLIMBER is an earth system model of intermediate complexity (EMIC) with a 2.5 dimensional dynamic statistical atmosphere model and no east-west structure within each ocean basin, whereas the IPSL model is a fully-coupled OAGCM. However, vegetation in CLIMBER is dynamically integrated with the other components, whereas we employ asynchronous coupling between the vegetation and the coupled OAGCM. The two simulations are also not precisely comparable: Ganopolski et al. (1998) used a pre-industrial CO2 level (280 ppm) in their experiments whereas we used a modern CO2 level (345 ppm). Nevertheless, comparison of the results from the two series of experiments does yield some insights.

The CLIMBER simulation, in agreement with our results, shows that ocean- and vegetation-feedbacks modify the orbitally induced pattern of summer warming and winter cooling (compared to present-day) and results in year-round warming in the high-latitudes. The magnitude of the warming is larger in the CLIMBER experiment, although the relative magnitude of the summer (+3.2 °C) and winter (+0.8 °C) warming is comparable to the summer/winter warming obtained in our OAV 6 ka simulation (+2.1 °C in summer and +0.6 °C in winter). However, in the CLIMBER simulation ocean feedback appears to have a negative impact on summer temperature and the positive impact of vegetation feedback is almost double that shown in our simulations (Table 5). The impact of the ocean feedback on winter warming in the CLIMBER simulations is of a comparable magnitude to its impact in our simulations. Vegetation feedback appears to be a less important cause of winter warming in the CLIMBER simulation than in our simulations. That vegetation feedback has a smaller impact on winter temperatures than ocean feedback is in basic agreement with our findings, but the magnitude of the vegetation feedback is smaller in the CLIMBER model (0.1 °C) than in IPSL-BIOME (0.3 °C).

The CLIMBER simulation shows an increase in summer precipitation in the high northern latitudes in response to orbital forcing (Table 5). This increase is reduced somewhat by ocean feedback, but amplified by vegetation feedback and synergy between the vegetation and ocean feedbacks. The simulated changes are thus completely different from those obtained with IPSL-BIOME, which shows summer drying resulting from the atmospheric response to orbital forcing and further amplified by vegetation feedback and synergy. The increased precipitation in the CLIMBER simulations is concentrated in the mid- to high-latitudes of Eurasia (see Fig. 2C, D in Ganopolski et al. 1998), and appears to be associated with the simulated expansion of the Afro–Asian monsoon. Thus, the difference between the CLIMBER and the IPSL-BIOME results may reflect the differences of resolution or in the representation of the hydrological cycle.

Both sets of experiments show an increase in high latitude winter precipitation, although the increase is larger in the IPSL-BIOME simulation (Table 5). The simulations disagree about the cause of this increase. In the IPSL-BIOME simulations, the increase in rainfall is driven by ocean feedback and ocean-vegetation synergy. Vegetation feedback has no impact on winter precipitation. In the CLIMBER simulations, both ocean and vegetation feedbacks have a small impact on winter precipitation, but the positive anomaly reflects the synergy between ocean and vegetation feedbacks.

Ocean and vegetation feedbacks are clearly important in modulating the Northern Hemisphere climate response to orbital forcing. Differences between the results of our simulations and those made with the CLIMBER model suggest that the role of these feedbacks is not yet completely understood. Differences in the impact of ocean feedbacks on monsoon climates between simulations run with different models have already been identified (see e.g. Hewitt and Mitchell 1998; Braconnot et al. 2000a; Liu et al. in press). These discrepancies between the impacts of feedbacks in different models suggest that the treatment of these feedbacks in climate models needs careful attention. Intercomparison of the regional impact of ocean and vegetation feedbacks as simulated by several models is a necessary first step towards the required diagnosis and is one of the explicit goals of the second phase of the Palaeoclimate Modelling Intercomparison Project (Harrison et al. 2002).

Acknowledgements.

We would like to thank Martin Claussen and Viktor Brovkin for stimulating our interest in examining the role of feedbacks on Arctic climates. We thank Kerstin Sickel for assistance in running the diagnostic simulations with BIOME4, Gerhard Bönisch for his help with accessing data from the PAIN and BIOME 6000 databases, Silvana Schott for the final figure layout and Claudia Kubatzki for providing us with data from the CLIMBER model. We thank Colin Prentice and Claudia Kubatzki for reviews of an earlier version of the manuscript. This is a contribution to the Palaeoclimate Modelling Intercomparison Project (PMIP), to the TEMPO (Testing Earth-system Models with Palaeoenvironmental Observations) project, and to the BMBF-sponsored project “Past Climate Sensitivity and Variability”.

Copyright information

© Springer-Verlag 2004