BIOME4: the vegetation model
BIOME4 (Kaplan et al., 2003) was developed from the BIOME3 model of Haxeltine & Pren-tice (1996). BIOME4 is a 1-D coupled carbon and water flux model that predicts global steady state vegetation distribution, structure, and bio-geochemistry. The model is driven by long-term averages of monthly mean temperatures, daily minimum temperatures, sunshine and precipita-tion. CO2 concentration has to be prescribed.
BIOME4 is based on twelve plant functional types (PFTs) ranging from cushion forbs to trop-ical rain forest trees (Kaplan, 2001). Each PFT is assigned a small number of bioclimatic lim-its which determine whether it could be present in a given grid cell. The computational core of BIOME4 is a coupled carbon and water flux scheme, which determines the seasonal maxi-mum leaf area index that maximizes the net primary production for any given PFT. To identify the biome for a given grid cell, the model ranks the tree and non-tree PFTs that were calculated for that grid cell. The resulting ranked combi-nations of PFTs lead to 27 different biomes (Ta-ble 1.2).
High-latitude biomes are represented by com-binations of frost-tolerant PFTs. Three PFTs (cold shrub, cold graminoid or forb, and cush-ion forb) used to distinguish the tundra biomes have been newly defined for BIOME4. These three new tundra PFTs are shallow rooted, and are sensitive to water stress and fire. The non-tundra PFTs used by BIOME4 to simulate high-latitude vegetation types include cold and tem-perate broadleaved and needlers trees, xero-phytic shrubs, and temperate grasses.
GRISLI: the ice sheet and ice shelf model
In this work, I have not really used the GRISLI ice model although, during this PhD, I have im-plemented a module accounting for the monthly climatology instead of mean annual climatology as usually prescribed in ice sheet models to im-prove the ablation calculation using the positive degree day method.
However, this thesis is based on the Late Saalian Eurasian ice sheet that has been built using GRISLI by Peyaud (2006). Consequently, I dedicate this section to the description of the main features of this ice sheet model.
GRISLI is a 3-D thermodynamical ice model that simulates the dynamics of grounded ice as well as ice shelves and ice stream regions. In-land ice deforms according to the stress balance using the shallow ice approximation (Morland, 1984; Hutter, 1983). Ice shelves and dragging ice shelves (ice streams) are described follow-ing MacAyeal (2001). This model has been de-veloped and validated over Antarctica by Ritz et al. (2001) in which a comprehensive descrip-tion of the model is provided. Here we list some of the recent improvements presented in Peyaud (2006) and Peyaud et al. (2007):
1. The thermo-mechanical coupling is ex-tended to the ice shelves and ice streams. Ice viscosity depending on the temperature field is integrated over the thickness.
2. The basal drag τB under ice streams is re-lated to ice velocity (τB = βU, where U is the horizontal velocity). The factor β depends on the effective pressure N : β = –cf × N , where cf is a constant term.
3. Basal water drainage is computed using a Darcian flow into a sediment layer. The thickness of the sediment layer is set to the ad-hoc value of 50 m. This description is too simplistic to account for the real basal pro-cesses but constitutes a siple representation of the drainage patterns.
4. Location of the ice streams is determined by the basal water head. Ice stream re-gions correspond to areas where the sedi-ment layer is water saturated.
5. Ice shelf front positions are determined with a scheme in which two criteria must be fulfilled. To calv the ice from the front grid point, first, the ice thickness must de-crease below 150 m. This corresponds to a general value in line with what has been observed for several of the Antarctic ice shelves. Secondly for each grid point at the front, the ice coming from the up-stream points must fail to maintain a thick-ness above the threshold. This ability to maintain a sufficient thickness is estimated on the basis of a semi-Lagrangian scheme. Ice shelf front position changes at each timestep, and appropriate boundary con-ditions, adapted from Rommelaere & Ritz (1996) and Ritz et al. (2001), are applied for the different front configurations. At-mospheric conditions have an implicit con-trol on ice shelves. Indeed the surface mass balance prevents the ice shelves to form in warm regions (Mercer, 1978). Simulations of West Antarctic ice shelves give front posi-tions in agreement with observations.
To reconstruct the climate at the surface of the ice sheet, the AGCM air temperature and pre-cipitation (ice equivalent) are adapted follow-ing Charbit et al. (2002). Temperature is cor-rected for altitude changes iteratively calculated by the ice sheet model. Vertical temperature gradients are based on classical values ranging from 5◦C.km−1 to 8◦C.km−1 (Krinner & Gen-thon, 1999; Abe-Ouchi et al., 2007). Accumula-tion is the solid fraction of the total precipitation and ablation is calculated according to a positive degree day (PDD) method (Reeh, 1991). The precipitation P0 (in ice equivalent) at the sur-face of reference is assumed to depend on the annual temperature T with an exponential law reflecting the saturation water pressure (Charbit et al., 2002):
P0 = P × e[0.05×(T0−T )] (1.10)
This parametrization is corrected from alti-tude variations. A fraction of the melting is likely to refreeze. As in the AGCM (Krinner et al., 2007) this fraction increases as the amount of melting compared to snow fall decreases (Sec-tion 1.3.1).
On the global ice volume
Global sea level variations result from geoid2 and solid Earth crustal height variations (Figure 1.9). Sea-level changes over long time scales are generally caused by the modification of the shape and/or the depth of the ocean floor as well as changes of the ice volume stored on the continents in the form of ice sheets and glaciers. These variations are called tectono-eustatism and glacio-eustatism respectively.
During the glaciations, the growth and de-cay of the ice sheets over the continents affect the solid Earth. The ice load creates large de-flections, affecting both the lithosphere and the mantle, that relax during and after the ice melt-ing until reaching the isostatic equilibrium. The isostatic depression from the ice load or the fol-lowing post-glacial rebound may have a direct impact on the near-field areas, i.e. areas located near or underneath the ice sheets, and conse-quently it may have an impact on the regional sea level in case these areas interact with the ocean. On the contrary, far-field areas, i.e. ar-eas not directly affected by the glacio-isostatism, are assumed to only record the eustatic sea level.
Eustatic sea level estimates
For the LGM, far-field areas are determined us-ing post-glacial rebound modelling and the eu-static sea level is determined in such areas us-ing coral measurements. Sub-surface corals rep-resent good indicators of sea level since their growth is sensitive to water depth, and thus, to fluctuations of the surface of the ocean. But since they are attached to the continent, sea level estimates have to be corrected for local tectonics in order to extract the pure eustatic sea level (Milne & Mitrovica, 2008). In their study, Milne & Mitrovica (2008) show that Bar-bados, Sunda, the Bonaparte Gulf , which are usually considered as reference far-field sites 3, hold an uncertainty on the eustatic sea level estimates of ≈10 m (Figure 1.10). They sug-gest that sites such as Seychelles (Indian Ocean) can provide improved estimates of past eustatic sea level since they are only slightly affected by glacio-isostatic adjustment. (Figure 1.10). Other sites such as Bengazi (Mediterranean) provide well constrained glacio-isostatic adjustment sig-nal for which, RSL can be easily corrected for to extract the eustatic sea-level.
For previous glacial periods, RSL coming from the sites described above time series are not available and a linear relationship between the δ18O record derived from measurements on ben-thic foraminifera has been used at it is assumed to record the global continental ice volume (Shackleton, 1987; Bintanja et al., 2005). Elabo-rate relationships using benthic foraminifera are also used as first approximation, but hold lots of uncertainties (Waelbroeck et al., 2002) (Fig-ure 1.11). Direct geological evidence of sea level positions through time, e.g. marine notches, ter-races, archaeological data, beach rocks, peats or coral reef, can be collected on continental mar-gins to calibrate the curves of local RSL. How-ever, their spatial and temporal resolution is most often low. Indeed, sea level records from continental margins are subjected to the Earth local subsidence or uplift. Therefore, derived RSL records need to be corrected for the effects of tectonic sediment loading and compaction, glacio- and/or hydro-isostasy and gravitational potential, in order to obtain data effectively interpreted as eustatic sea level (Rabineau et al., 2006).
The maximum eustatic sea level drop esti-mated for the Late Saalian glacial maximum ranges from about 92 m (Rabineau et al., 2006) to 150 m below present sea level (Waelbroeck et al., 2002). The estimate by Rabineau et al. (2006) based on measurements carried out in the Gulf of Lyon, should probably be considered on the smaller end for the Late Saalian period as their value derived from the same site for the LGM is too small due to a biased correction of the isostatic contribution. The Gulf of Lyon is located in the vicinity of the Alps that hosted large glaciers during glacial periods. Spada et al. (2009) have shown that the Alps have a sig-nificant influence on Mediterranean sea level records from the LGM until today. I have not find any additional data that shed new lights on the maximum estimate of -150 m by Waelbroeck et al. (2002). However, the Late Saalian RSL was recorded to range between 90 m to more than 150 m below present level in Eurasia (see Lam-beck et al. (2006) for the review of available RSL data). These values are however not corrected for the isostatic component and do not represent eustatic values. Therefore in all our Late Saalian experiments we set the sea level to -110 m (As-takhov, 2004) and to -130 m (Peltier, 2004) for our LGM simulations.
From sea level to ice volume
From the previous section, it is evident that global estimates of the Late Quaternary sea level fluctuations are still far from well constrained. One could argue that all the LGM (MIS 2) values converge towards -130 m ±10 m and that Late Saalian (MIS 6) oxygen isotopes fluctuations ex-hibit similar values to that of LGM. But the prob-lem is more complex because the global equiva-lent sea level ice volume has to be spatially dis-tributed between the individual ice sheets that developed during the various ice ages. If we con-sider that each eustatic value derived from corals or oxygen isotopes hold an error of about 10%, the maximum estimate for the Late Saalian has an uncertainty of 15m while the minimum has 9m. Assuming this uncertainty, it is evident that the uncertainty amounts to at least half of the equivalent sea level (ESL) volume attributed to the Eurasian ice sheet in ICE-5G during the LGM (≈17 m).
sheet has increased by 20 m compared to the previous versions, the Eurasian ice sheet volume has decreased by 7 m ESL and that of Antarc-tica has decreased by 10 m (Table 1.3). The ESL contribution of the latter seems still to be overes-timated and recently Ivins & James (2005) pro-posed new estimates for the Antarctic ice sheet’s volume variations since LGM, based on a syn-thesis of the current constraints on past ice his-tory and present-day mass balance, which sug-gest an ESL contribution well below what have been suggested by other publicly available ice models (Table 1.3).
Distributing ice volume between the various ice sheets is not easy. Two approaches are clas-sically used: the first one consists of using a glacio-isostatic adjustment model in which ini-tial ice ESL is prescribed through the imposed ice growth and melting chronology, as well as mantle viscosity and lithospheric elastic thick-ness, and then adjusting the ice volume combin- the second one involves using a dynamic ice ESL values attributed to the various ice sheets model coupled to a simplified isostatic module, have changed substantially since the first ice forced with a prescribed climatology, and let the model ICE-1 of Andrews (1976). In the model model run until the ice sheet reaches equilib-version ICE-5G (Peltier, 2004), the ESL at- rium (≈200 kyrs, e.g., Siegert, 2001; S.J. et al., tributed to the volume of the Laurentide ice 2000).
The Late Saalian Eurasian ice sheet
For the Late Saalian glacial maximum, two re-cent model reconstructions of the Eurasian ice sheet have been performed: the isostatic recon-struction of Lambeck et al. (2006) and the dy-namical ice sheet of Peyaud (2006). The recon-struction by Lambeck et al. (2006) is based on the following various assumptions and settings:
1- The lithospheric thickness and viscosity are set to 80 km and 1025 Pa.s (infinite) respec-tively
2- The Earth’s mantle is divided into two lay-ers with a viscosity 3×1020 Pa•s for the up-per mantle and 5×1021 Pa•s for the lower mantle
3- The elastic moduli and density of the Earth’s layers are taken from the PREM (Dziewon-ski & Anderson, 1981) and mantle rheology is described by a simplified Maxwell visco-elasticity
4- the initial geometry of the ice sheet is fixed, a priori, with two main domes: one over Scandinavia and one over the Kara coast-lines
5- Late Saalian eustatic sea level is taken from Lambeck & Chappell (2001) and equals to 140 m
6- Finally, the ice mechanics equations of Pa-terson (1994) and the isostatic radial re-bound calculation are used to estimates the ice thickness of the initial Late Saalian to-pography
The final reconstruction yields a mean ice ele-vation of approximately 3000 m (≈4000 m ice thickness) with a local deflection of ≈ 1000 m (Figure 1.12). The total ice volume is about 60 m ESL. This reconstruction links the British Isles ice cap to the huge Eurasian ice sheet. The QUEEN Late Saalian ice sheet is not extended over the southern part of the British Isles as the field data from this area are not conclusive (Svendsen et al., 2004).
The reconstruction of Peyaud (2006) has been computed using the GRISLI ice model based on the shallow ice approximation (Section 1.3.4). Temperature gradient and precipitation are pre-scribed so that the ice sheet morphology grows to fit the QUEEN limits (Figure 1.1). In the ex-periment, eustatic sea level was set to -110 m be-low present (Astakhov, 2004) and GRISLI ran for 200 kyrs in order to reach the ice sheet equilib-rium. The GRISLI model Late Saalian ice sheet has a shape rather similar to that of Lambeck et al. (2006). The mean elevation is also about 3000 m, domes are located almost at the same place and ice volume is ≈ 61 m ESL. The litho-spheric deflection allows proglacial lakes to de-velop in both Lambeck et al. (2006) and Peyaud (2006) scenarii, especially in the Siberian Plains.
In all the contributing peer-reviewed arti-cles presented below, we used Peyaud (2006) Eurasian ice topography. The Laurentide, the Greenland and the Antarctic ice sheets were pre-scribed according to the LGM ICE-5G ice topog-raphy since there are no direct evidence to con-train their topography during the Late Saalian.
Table of contents :
1.1 Background and Scientific Motivations
1.2 The Late Eurasian Saalian period (160 ka – 130 ka)
1.2.1 The Late Saalian Northern Hemisphere topography
1.2.2 Orbital parameters and Greenhouse Gases (GHG)
1.3 Numerical models
1.3.1 LMDZ4: the atmospheric general circulation model
1.3.2 Planet Simulator: the AGCM mixed-layer ocean model
1.3.3 BIOME4: the vegetation model
1.3.4 GRISLI: the ice sheet and ice shelf model
1.4 Boundary conditions
1.4.1 On the global ice volume
1.4.2 Ice-dammed lakes
1.4.3 Dust sources
1.4.4 LGM vegetation cover
1.4.5 Sea surface conditions
1.4.6 The Arctic Ocean ice shelf
1.5 Summary of the peer reviewed articles
Manuscript 1: Late Saalian climatic impact of regional factors
Manuscript 2: the Late Saalian vegetation cover
Manuscript 3: the Late Saalian surface ocean
Manuscript 4: the MIS 6 Canada Basin ice shelf
Manuscript 5: Synthesis of the Late Saalian climate (160 – 140 ka)
A Eurasian topography and Arctic IBCAO bathymetry
B Post-glacial rebound and sea level variations
B.1 Bounds on the Time-history and Holocene Mass Budget of Antarctica from Sea-level Records in SE Tunisia (in press)
B.2 Glacio-isostatic adjustment in the Po plain and in the northern Adriatic region (in press)
B.3 Post glacial readjustment sea level variations subsidence and erosion along the Italian coasts (in press)