Interactive comment on “ A model-data assessment of the role of Southern Ocean processes in the last glacial termination ” by Roland

Eichinger et al. present results of simulations performed with DCESS model of intermediate complexity. The main goal of the study is to understand the changes in atmospheric D14C between 17.5 and 14.5 ka B.P. The impact of changes in permafrost, deep ocean ventilation, dust and a higher PO4 content are briefly looked into over the whole deglaciation. My main criticism is that in some ways the authors try to do too much without really looking at each of the processes. So, at the end little new information is coming out of the manuscript or some important information is missing to really understand the implications. Please find some specific comments below.


Introduction
This supplement provides additional material to the article "A model-data assessment of the role of Southern Ocean processes in the last glacial termination".This includes the description of the profile function of the ocean vertical diffusion in the high latitude model ocean applied for generating Last Glacial Maximum (LGM) conditions in the DCESS model.Furthermore, the newly developed scheme for the representation of the terrestrial biosphere in the DCESS model is presented in detail.
The formulation of the vegetation zones and their impact on the atmosphere-biosphere exchange fluxes as well as an evaluation of the module is given.We also describe our treatment of dust radiative forcing and high latitude ocean iron-limitation and provide additional information about the various 14 C production rate time series that were applied in the model simulations.

Ocean vertical diffusion profile
In the article, we present a profile for the high latitude ocean vertical diffusion with depth, which is applied to generate an analogy to isolated Southern Ocean (SO) deep and intermediate waters (see e.g. Watson and Naveira Garabato, 2006).In this section, we present details about this profile and explain how it was established.The mathematical formulations of the profile and of the resumption to full vertical diffusion in the transient simulations are given.

Description of the profile
For pre-industrial (PI) climate conditions, the ocean vertical diffusion parameter is set to D P I = 2.3 • 10 −3 m 2 s −1 evenly throughout the high latitude DCESS model ocean.To generate isolated deep water and thereby LGM climate conditions, we impose the function on this parameter.This signifies a sharp reduction of the high latitude ocean vertical exchange at an ocean depth depending on p a (see below).Here, d denotes the ocean depth (in m) and the factor 0.3 was chosen as a best-guess to set the steepness of the transition.
and For our efforts to generate an LGM climate state, which constrains all the mentioned variables to proxy data records concurrently, we used the parameter p a to vary the transition depth and thereby the oceanic volume of isolated deep water.This yielded p a = 5 for the presented representation of LGM climate, which implies a transition at around 1800 m depth.

Resumption of deep ocean mixing
In order to simulate the resumption of deep ocean mixing during the Mystery Interval (MI), we impose a steady deepening of the transition depth of the imposed vertical diffusion profile.Starting at year 17.5 kaBP we modify p a in Eq. 1 by where n denotes the time step after the start of the resumption.This steadily decreases the transition depth of the vertical diffusion ocean profile over time.Effectively, it leads to the resumption of the high latitude ocean to PI mixing within around the 3 ka of the MI.In other words, this simulates a steady, time-dependent upwelling of the deep ocean waters that had been isolated through the imposed vertical mixing profile.Thus, it describes an analogy to the sudden event of deep ocean mixing during the MI, that has been hypothesised in a number of studies (e.g.Burke and Robinson, 2012).
3 The new terrestrial biosphere scheme Here, we provide a detailed description of the newly developed terrestrial biosphere scheme of the DCESS model.First, a presentation of the characteristics of the chosen vegetation zones and their latitudinally variable borders is given.Then, the new calculations of the biosphere-atmosphere exchange fluxes of CO 2 and CH 4 for 12 C as well as for the rare carbon isotopes 13 C and 14 C are shown.
Moreover, in this section, we provide a brief evaluation of the new module, in order to ascertain that the general purpose of this biosphere scheme, the representation of land-atmosphere carbon fluxes across the last glacial termination, is reasonable with this approach.

Description of the vegetation zones
Based on the modelling study by Gerber et al. (2004), we defined a latitudinal distinction of three vegetation zones, tropical forests (TF), grasslands, savanna and deserts (GSD) and extratropical forests (EF).Tab. 1 shows the characteristic values of biomass reservoirs and net primary production (NPP) of those vegetation zones at PI climate conditions for one hemisphere (Saugier et al., 2001;Gower et al., 1999;Sterner and Elser, 2002;Zheng et al., 2003;Chapin et al., 2011).The values in Tab. 1 have been constrained such that the sum over the three vegetation zones adds up to PI values of the original biosphere model (Shaffer et al., 2008).The latitudinal limits of these vegetation zones are dynamically defined.In general, the extent of certain vegetation zones depends mainly on temperatures and precipitation.However, the limitations of the DCESS model require a somewhat more general approach.We therefore determine the division of the three vegetation zones solely by the deviation of the global mean atmosphere temperature from its PI value.For this purpose, we derived two polynomial functions from a study by Gerber et al. (2004, in particular from the results in their Fig. 4).In that study, the complex LPJ vegetation model (Lund-Potsdam-Jena Dynamic Global Vegetation Model) was applied to distinguish between a number of vegetation zones based on several variables.The two latitudinal limitations of the vegetation zones are described by the two 5th order polynomials and which depend only on the deviation of the global mean atmosphere temperature δT glob .L T F −GSD denotes the latitude of the border between the TF and the GSD zones and L GSD−EF the latitude between GSD and EF.These two 5th order polynomials are illustrated in Fig. 1.The EF vegetation zone additionally is limited by either the snowline (defined as the latitude of 0 • C global mean temperature) or the line of the terrestrial ice sheet extent, depending on which one of the two lines expands the farthest from the pole at the current time step (see next section).
Based on these latitudinal limits, the total CO 2 and CH 4 fluxes between the terrestrial biosphere and the atmosphere are now determined by the sum of the three vegetation zones and dependent on the respectively covered vegetation areas as well as their current values of NPP and stored biomass.

Extension of the carbon flux equations
The equation from Shaffer et al. (2008) for the NPP is now subdivided into the three equations and for the different vegetation zones, respectively.The factors A T F , A GSD and A EF are calculated by and and scale the contributions of the respective NPP by the current area of the individual vegetation zone.The index P I stands for reference PI conditions and f CO2 for the CO 2 fertilisation factor.
In the original configuration, this factor was set to 0.65, which was in good agreement with results by Friedlingstein et al. (2006).However, a revision of this value in a model intercomparison study yielded a lower value of 0.37 to be a more suitable value for the terrestrial biosphere (Zickfeld et al., 2013) and thus has also been used in the present study.Now, the four conservation equations per carbon isotope ( 12,13,14 C) (see Shaffer et al., 2008) have to be calculated for each vegetation zone separately.The losses for reservoir size of litter and soil were dependent on the mean global 100 atmosphere temperature in Shaffer et al. (2008) for the uniform vegetation.Now, we approximate a mean atmosphere temperature for each vegetation zone separately by and in order to achieve a more realistic dependency of this process.Here, T atm,LL denotes the mean atmosphere temperature in the low latitude sector and T atm,HL in the high latitude sector.L snow/ice stands for the minimum of the latitude of the snow and the ice sheet line.The fluxes between the terrestrial biosphere and the atmosphere are determined by Due to global temperature changes on glacial-interglacial time scales, L snow/ice advances and retreats on large spatial scales and thus organic carbon is buried/released below/from permafrost areas.
Therefore, additional land-atmosphere carbon flux variations due to the changes of permafrost area have to be considered.For this, we add the permafrost flux term 12,13,14 F CO2,P F to Eqs. 18 to 21, which is calculated by 12,13,14 F CO2,P F = ∆A snow/ice • 12,13,14 C P F .
Since the contribution of carbon release from permafrost to methane is very small and moreover, we prescribe the methane radiative forcing in our simulations, we neglect the separation of these effects here.∆A snow/ice denotes the change in snow or ice covered area and is computed by For this, the respective minimum of the snow-or the iceline (L snow/ice ) of the previous (t − 1) and the current (t) time step is taken.R denotes the Earth radius and the factor (270/360) takes account for the land fraction in the model geometry.
12 C P F , the amount of carbon being stored in permafrost, was approximated to 30 kg • m −2 by Schuur et al. (2015).For the stable 13 C P F isotope, carbon is buried and released through permafrost with the same isotope ratio.We hence set the value to 0.33 kg•m −2 , to yield a typical isotope ratio for EF soil of δ 13 C = −24‰.For 14 C P F , however, radioactive decay (T 1/2 ( 14 C) = 5730 a) across the entire last glacial period, when large parts of the high latitudes were covered by terrestrial ice sheets, has to be considered.We therefore assume carbon to be released from permafrost radiocarbon dead (∆ 14 C = −1000‰) during the last deglaciation (see also Zech, 2012), while being buried with the current isotope ratio of soil.As is, the land area of the model simply covers 1/3 of the globe across all latitudes.In our configuration for the last glacial termination, the permafrost affects latitudes between 47 • and around 54 • (see Fig. 3 in the manuscript).Across these latitudes, the land fraction averaged over both hemispheres is around 30% (see e.g.Matney, 2012).Thus, a further adjustment for the permafrost effect due to land fraction was not considered to be necessary.

Evaluation of the new biosphere scheme
In order to demonstrate that the newly developed land biosphere scheme in the framework of the DCESS model adequately represents the different vegetation zones for global climate changes and thus meets the needs for the simulation of the last glacial termination, we here present a brief evaluation of the new module.For this purpose, we show the reaction of the model vegetation zones and the different vegetation reservoirs on a reduction and an increase of the atmospheric temperatures and pCO 2 .Moreover, we compare the results of the cooling experiment with complex vegetation models as well as with data reconstructions.
The maximum rise in atmospheric temperatures in future climate change simulations with the DCESS model is around 5.2 • C at year 2100 (in the A2 emission scenario, see Shaffer et al., 2009).
For the evaluation, we carried out two warming simulations (with and without the new biosphere scheme) with a gradual increase of pCO 2 to 1000 ppm within around 300 a and fix the global mean atmosphere temperature (T glob ) once it has risen by 5.2 • C. The left panel of Fig. 2 shows the land biomass carbon (LB) for the three individual vegetation zones and their sum for the two warming simulations and the right panel shows the global sum of LB of the vegetation reservoirs above ground (leaves + wood) and below ground (litter + soil).Since the system seems to be in equilibrium after around 1 ka, we integrated these simulations over 2 ka.The warming experiment shows an overall increase in land biomass in both simulations, with more change in the simulation with the new biosphere scheme.The land biomass change happens slower in the simulation with the new biosphere scheme, because the land biomass changes in the EF zone mainly depends on variations in soil, which has a slow response time and is the largest biomass reservoir.The TF zone adapts much quicker to the new climate conditions, because in this vegetation type the biomass is dominated by leaves and wood.
To evaluate the vegetation scheme for LGM conditions, we carried out two cooling simulations.
For these, we gradually decreased the temperature until T glob =11.5 • C was reached and prescribed the atmospheric pCO 2 to 190 ppm.Fig. 3 shows the results of this cooling experiment, depicted as in the warming experiment above.
Here, basically the same effects as in the warming experiment can be observed.The overall change is larger in the simulation with the new biosphere scheme.The GSD vegetation zone shows the smallest change in biomass in both experiments, because the area of this vegetation zone changes only little, it rather just shifts latitudinally.
We calibrated the latitudinal dependency of the vegetation zone borders to match the LPJ model results.However, the calculation of carbon stored in the terrestrial biosphere at different climate conditions also depends on other parameters.Hence, we first evaluate the performance of the new Overall, it can be stated that the new biosphere scheme with the three vegetation zones constitutes a significant improvement for the representation of the terrestrial biomass, and thereby also the land-atmosphere carbon exchange rates in the DCESS model.This new implementation allows more nuanced investigations concerning the complex interactions between the terrestrial and the atmospheric carbon exchange that is required to obtain a better understanding of the processes that determine the climate change from the LGM to the Holocene.

Transition function for dust concentration
As described in the manuscript, the dustier atmosphere during LGM climate conditions (e.g.Maher et al., 2010) leads to the assumptions of an additional radiative forcing (A Dust ) of −1 W m 2 (see Mahowald et al., 2006) and an increased iron limitation factor (f F e−lim ) in the high latitude ocean (Lambert et al., 2013(Lambert et al., , 2015)).For the transition of these parameters between LGM and PI conditions in the transient simulations, we applied functions derived from the correlation between dust flux and global mean atmosphere temperatures in proxy data records.Fig. 4 shows proxy data records from Antarctica of the dust flux and the temperature (Jouzel et al., 2007;Lambert et al., 2012).As stated in Lambert et al. (2012), the major changes in dust flux occurred before 15 kaBP.An analysis of the correlation of the two quantities provides an exponential dependency with an exponent of −0.4.The same climate-dust exponential relationship also holds true for a remote site from the central Pacific (Winckler et al., 2008).
This relationship provides a relatively simple approach for the temperature-dependent transition functions of A Dust and f F e−lim from LGM to PI conditions.For this, we use the condi-    2004) determined the 14 C production rate as a function of past changes in geomagnetic field intensity using the conversion of Masarik and Beer (1999).Hain et al. (2014) have further processed these data with a Monte-Carlo simulation with 1000 members of randomised magnetic field strength records within the GLOPIS-75 uncertainty envelope and a more recent 14 C production rate model by Kovaltsov et al. (2012) with updated cosmic ray spectra.Muscheler et al. (2004) have used a different approach to infer the 14 C production rate across the last 25 ka based on high resolution 10 Be measurements from GRIP and GISP2 ice cores.
Assuming that the atmospheric processes of 10 Be (attachment on aerosols and deposition after a mean residence time of 1−2 years in the atmosphere) are understood, the 14 C production rate can be estimated rather directly, because 10 Be and 14 C are produced by similar processes in the atmosphere (Lal and Peters, 1967;Masarik and Beer, 1999).This reconstruction takes into account all probable causes of production rate variations, such as variations in geomagnetic field, solar activity and/or in the interstellar galactic cosmic rays flux (Muscheler et al., 2004).
Generally, the three individual 14 C production rates show similar patterns across the last 25 kaBP, although with a couple of differences.The data by Hain et al. (2014) and by Laj et al. (2004) are derived from the same approach and the general features and local extrema therefore show the same timing, although the time series by Hain et al. (2014) seems somewhat smoother with a smaller general trend.The data by Muscheler et al. (2004) shows far stronger and more high-frequent fluctuations than the other two data sets, the general trend, however, is similar.Only between 20 and 23 kaBP do the time series of the different approaches show an opposing behaviour.While the data set by Muscheler et al. (2004) shows a minimum during that period, the other two time series possess maxima here.Between the beginning and the end of the MI, all data sets are similar.A reduction of 0.11 atmos/cm 2 s can be seen in the data by Hain et al. (2014), 0.19 atmos/cm 2 s in the data by Laj et al. (2004) and even a slight increase of 0.02 atmos/cm 2 s in the data by Muscheler et al. (2004).with the 'Pee-Dee Belemnite' standard R 13 pdb = 0.0112372 (Zhang et al., 1990) for the 13 C/ 12 C ratio and by with the 'Oxalic Acid Standard' R 14 oas = 1.176 • 10 −12 for the 14 C/ 12 C ratio (Karlen et al., 1964).
3) are used to calculate the maximum and the minimum of the diffusion profiles in relation to the standard value for PI conditions (D P I = 2.3 • 10 −3 m 2 s −1 ).Here, p min represents the relative minimum of the diffusion, which is set to 0.03, and R C determines the change of diffusivity above the transition.However, the minimum of the vertical diffusion by depth is determined by the low latitude ocean value D p,min = 2 • 10 −5 • 1 + 5.5 • 1 − exp d 40 (4) and imposed by computing D p,tot = max(D p , D p,min ).We used R C = 1.0 to keep the parameter at standard PI conditions in the shallow ocean for simplicity.

Figure 1 .
Figure 1.Polynomial functions describing the dynamic latitudes of the borders between the three vegetation zones dependent on the deviation of the global mean atmosphere temperature δT glob .Red: Border between the TF and the GSD zone (LT F −GSD ).Blue: Border between the GSD and the EF zone (LGSD−EF ).The dots mark the conditions at PI (δT glob,P I = 0 • C; LT F −GSD,P I = 11.28 • , LGSD−EF,P I = 37.77 • ) and LGM (δT glob,LGM = −3.5 • C; LT F −GSD,P I = 7.17 • ; LGSD−EF,P I = 33.92• ) climate state.

Figure 2 .
Figure 2. Warming simulation (see text) for the model version with the new (solid) and the old (dashed) vegetation scheme.Left panel: Total land biomass carbon (in black) separated into the three vegetation zones (TF: green, GSD: red, EF: blue).Right panel: Land biomass carbon separated into the vegetation reservoirs above ground (leaves + wood) in red and below ground (litter + soil) in black.
tions f F e−lim = 0.36 for PI (T glob = 15 • C) and f F e−lim = 0.5 (see manuscript) for LGM (T glob = 11.5 • C) and A Dust = 0 for PI and A Dust = −1 for LGM.With the function a • e −0.4•T glob + b for the correlation between T glob and dust flux during this period derived from proxy data, the transition functions f F e−lim = 0.14 • e −0.4•T glob + 0.36 • e −4.6 − 0.5 • e −6 e −4.6 − e −6 (24)

Figure 5 .
Figure 5. Transition functions for iron-limitation factor f F e−lim (left panel) and for additional radiative forcing due to dust ADust (right panel) variations between LGM and PI.
as well asLaj et al. (2004) only provide normalised (around 1) values for the 14 C production rate, because the determination of the absolute values is very uncertain.Only the more recent study byHain et al. (2014) provides these absolute values (in atoms/cm −2 s).For this reason, we scaled the two other data sets by factors that yield 14 C production rates throughout the last 25 kaBP that are close to the data byHain et al. (2014) and moreover, lead to atmospheric ∆ 14 C results close to the data based reconstructions at the beginning of the MI in the DCESS simulations.An evaluation of good fits for these scaling factors yielded 1.5• 10 4 atoms • cm 2 s −1 for the data byMuscheler et al. (2004) and 1.35•10 4 atoms • cm 2 s −1 for the data