Abstract :
[en] A new three-dimensional Earth system model of intermediate complexity was developed. This model, named LOVECLIM, consists of five major components representing the atmosphere (ECBilt), the ocean and sea ice (CLIO), the terrestrial biosphere (VECODE), the oceanic carbon cycle (LOCH) and the Greenland and Antarctic ice sheets (AGISM). It also includes a global glacier-melt algorithm which is run in off-line mode. It is worth mentioning that there are very few models of this type worldwide. ECBilt is a quasi-geostrophic atmospheric model with 3 levels and a T21 horizontal resolution. It includes simple parameterisations of the diabatic heating processes and an explicit representation of the hydrological cycle. Cloud cover is prescribed according to present-day climatology. CLIO is a primitive-equation, free-surface ocean general circulation model coupled to a thermodynamic–dynamic sea-ice model. Its horizontal resolution is 3° × 3°, and there are 20 levels in the ocean. VECODE is a reduced-form model of vegetation dynamics and of the terrestrial carbon cycle. It simulates the dynamics of two main terrestrial plant functional types (trees and grassland) at the same resolution as that of ECBilt. LOCH is a comprehensive model of the oceanic carbon cycle that takes into account both the solubility and biological pumps. The version utilised here has the same resolution as the one of CLIO, which greatly facilitates the coupling between both models. Finally, AGISM is composed of a three-dimensional thermomechanical model of the ice sheet flow, a visco-elastic bedrock model and a model of the mass balance at the ice–atmosphere and ice–ocean interfaces. The Antarctic ice-sheet module also contains a model of the ice-shelf dynamics to enable interactions with the ocean and migration of the grounding line. For both ice sheets, calculations are made on a 10 km × 10 km resolution grid with 31 sigma levels.
The performance of LOVECLIM was assessed by conducting ensemble simulations over the last few centuries. Starting from different initial conditions, the model was integrated from year 1500 AD up to year 2000 AD with solar irradiance, volcanic activity, tropospheric ozone amount, greenhouse-gas (including CO2) concentrations and sulphate-aerosol load evolving with time according to reconstructions. Over the last 140 years, the model simulates a global surface warming ranging from 0.33°C to 0.43°C, with a mean value of 0.38°C. This value is
about 0.15°C lower than the observed one. A detailed analysis of the results has revealed the model behaves reasonably well at mid- and high latitudes. By contrast, at low latitudes, the agreement between the model results and observational estimates is less good, especially in the Southern Hemisphere. In those regions,
LOVECLIM significantly underestimates the warming and the climate variability observed during the last few decades. The coarse resolution of the model and the simplified representation of the atmospheric dynamical and physical processes seem to be the two major candidates responsible for this deficiency. Regarding the
Greenland ice sheet, we found a slightly increasing ice volume during the period 1700–2000 AD. This trend is largely explained as a residual response to the late Holocene forcing, in particular to the Little Ice Age cooling after year 1500 AD. The effect is not particularly large, however, amounting to only 1.2 cm of global sea-level
rise over the entire period. The growing trend stabilizes during the 20th century, with almost no net effect on ice volume. Only during the last decades of the 20th century, the ice volume begins to decrease in response to the imposed warming. We also found the Antarctic ice sheet to be retreating slowly at a rate equivalent to a global
sea-level rise of about 1.7 cm during the 20th century. This evolution is mostly due to a long-term background trend of +2.6 cm, mitigated by about 0.9 cm from slightly rising accumulation rates over the same period. The ongoing dominance of past climatic changes on the contemporary ice-sheet evolution is a fine illustration of the
inertia encountered when studying the response of large continental ice sheets. In this case, it mainly results from an ongoing grounding-line retreat in West Antarctica following rising sea levels since the Last Glacial Maximum. As far as mountain glaciers and small ice caps are concerned, their area and volume are found to reach a maximum in the late 19th century corresponding to the Little Ice Age, but this maximum and the ensuing 20th century glacier retreat are not very pronounced. Over the last hundred years, the model simulates an ice loss equivalent to only 0.89 cm of sea-level rise. This value is at the lower end compared to other assessments. One reason is the low total ice volume assumed by the global glacier-melt algorithm (about 20 cm of total sea-level rise, a factor 2.5 less than previous estimates). A second reason is the prescribed global ice mass balance for the 1961–1990 reference period, which is also at the lower end of other simulations. For the 20th century, LOVECLIM explains about 7.6 cm of sea-level rise. The bulk of that value, about 4.7 cm, comes from thermal expansion of the World Ocean. The Antarctic and Greenland ice sheets combined lead to a sea-level rise of 2 cm, and glaciers and ice caps are responsible for about 0.9 cm of sea-level rise. These numbers are similar to those that have been derived for the IPCC Third Assessment Report (TAR) for the same components except for the lower glacier contribution as found here. Over the industrial era, the net uptake of carbon by the ocean simulated by LOVECLIM is within the range of current estimates, although at the lower end of this range. It should be noted that a detailed evaluation of the performance of the terrestrial carbon-cycle module was impossible to perform given the very wide range of available data. Experiments with interactive atmospheric CO2 concentration were also carried out with LOVECLIM forced by CO2 emissions from fossil fuel burning and land-use change. Interestingly enough, the atmospheric CO2 level computed by the model in year 2000 AD compares relatively well with the observed one.
A series of climate-change projections were then conducted over the 21st century. In these experiments, LOVECLIM was driven by changes in greenhouse-gas (including CO2), tropospheric ozone and sulphate-aerosol concentrations following the IPCC SRES scenarios B1, A1B and A2. In year 2100 AD, the model predicts a
globally averaged, annual mean surface warming of 1°C, 1.4°C and 1.8°C for scenarios B1, A1B and A2, respectively, and an associated increase in precipitation of 3.6%, 5.1% and 6.6%, respectively. In agreement with studies performed with climate general circulation models (CGCMs), a weakening of the Atlantic meridional
overturning circulation (MOC) is noticed in all runs. At the end of the 21st century, the decrease in the maximum value of the annual mean meridional overturning streamfunction below the surface layer in the Atlantic basin, which is an index of the MOC intensity, reaches 19% for scenario B1, 21% for scenario A1B and 27% for
scenario A2. In our model, as in the majority of CGCMs, this decrease is caused more by changes in surface heat flux than by changes in surface freshwater flux. Under the forcing scenario A1B, LOVECLIM simulates a global sea-level rise of 31.3 cm in year 2100 AD. As for the 20th century, the most important contributor is the
oceanic thermal expansion (+18.8 cm), followed by the contributions from the Greenland ice sheet (+5.2 cm), glaciers and ice caps (+3.8 cm) and the Antarctic ice sheet (+3.5 cm). The total rise is equivalent to a quadrupling of the sea-level rise simulated for the 20th century. Our sea-level value is somewhat lower than the central estimate for the same four components of about 40 cm in the IPCC TAR predictions. This can be explained by the low climate sensitivity of LOVECLIM, and hence the lower global temperature rise, which mostly affects the largest contribution of thermal expansion of the World Ocean. Another difference with the IPCC TAR predictions is the positive contribution from Antarctica of several cm of sea-level rise. That is in contrast to most other simulations showing a growing ice sheet and a negative contribution to global sea level of typically between -5 and -20 cm. The IPCC TAR also found a generally larger contribution from mountain glaciers and small ice caps. Our glacier-volume loss is smaller because of the lower initial glacier volume assumed by the glacier-melt algorithm. The total projected sea-level rise for the 21st century is only slightly affected by the scenario itself. For the range of SRES scenarios used by LOVECLIM, the total sea-level rise is found to vary between +22 and +35 cm by year 2100 AD. The much larger range of between +9 and +88 cm obtained for the IPCC TAR arose mainly from the inclusion of model uncertainties, and not from the greenhouse-gas-forcing scenarios employed. As expected, climate change impacts the air–sea CO2 exchange in the model by lowering the solubility and hence the net uptake of carbon by the ocean. The effect is however rather modest at the century time-scale given the moderate increase in sea-surface temperature simulated by LOVECLIM. In addition, we do not observe any significant change in the oceanic biology at the global scale during the 21st century. The picture is a bit different regarding the terrestrial biosphere. Both the climate and fertilization effects strongly increase the carbon uptake in VECODE. A number of experiments with interactive atmospheric CO2 concentration were also carried out over the 21st century. Contrary to other modelling studies, LOVECLIM predicts lower atmospheric CO2 levels at the end of the 21st century when the effect of climate change on the carbon cycle is accounted for in the model. The warming enhances the net uptake of carbon by the terrestrial biosphere which more than offsets the reduction in oceanic uptake resulting from the solubility decrease.
Finally, we have thoroughly analysed the model response to a range of stabilized anthropogenic forcings over the next millennia. For the variety of forcing scenarios considered, LOVECLIM simulates a globally averaged, annual mean surface warming ranging between 0.55°C and 3.75°C and an associated decrease in Arctic and Antarctic sea-ice extent. However, no simulation predicts an entirely ice-free Arctic Ocean during summertime at the millennium time-scale. In the most pessimistic case, a small ice pack of about 0.5×106 km2 persists. Our results also suggest that it is very likely that the volume of the Greenland ice sheet will largely decrease in the future. After 1000 years of model integration, the ice volume is reduced by more than 20% when the radiative forcing is higher than 6.5 W m-2. Moreover, for a radiative forcing greater than 7.5 W m-2, the ice sheet melts away in less than 3000 years. Note that the ice-sheet disintegration might be even more rapid if processes responsible for the widespread glacier acceleration currently observed in Greenland were taken into consideration in the model. We also found that the freshwater flux from the melting Greenland ice sheet into the neighbouring oceans, which peaks in the most extreme scenario tested at 0.11 Sv (1 Sv = 10^6 m3 s-1) and remains above 0.1 Sv during three centuries, is not large enough to trigger a shutdown of the Atlantic MOC in our model, in contrast to some other models. Those models are however more responsive to freshwater perturbations than ours. Besides, we showed that climate feedbacks play a crucial role in the ice-sheet evolution and that the Greenland deglaciation considerably enhances the greenhouse-gas-induced warming over Greenland and the central Arctic. This stresses the importance of incorporating the two-way interactions between the Greenland ice sheet and climate in climate- and sea-level-change projections at the millennial time-scale. For the Antarctic ice sheet, the response is much less drastic than for the Greenland ice sheet. For instance, after 3000 years of 4×CO2 forcing (∼7.7 W m-2), the Antarctic grounded ice volume and area are reduced in our model by only 8% and 4%, respectively. For a sustained radiative forcing of 8.5 W m-2 (the highest forcing scenario considered in our study), LOVECLIM predicts a global sea-level rise of 7.15 m by year 3000 AD. Most of it is due to melting of the Greenland ice sheet (+4.25 m), followed by melting of the Antarctic ice sheet (+1.42 m), thermal expansion (+1.29 m) and the contribution from mountain glaciers and small ice caps (+0.19 m). Our results show that it will be very difficult to limit the eventual sea-level rise to less than 1 m after 1000 years, unless the atmospheric CO2 concentration can be stabilized to less than twice its pre-industrial level. Such a goal can only be reached by emission reductions far larger than any policy currently pursued. Concerning the carbon cycle, the experiments carried out with LOVECLIM highlight the opposite responses of the terrestrial and oceanic carbon reservoirs to climate change. We also found that, when anthropogenic CO2 emissions cease, the terrestrial biosphere becomes a weak carbon source, while the ocean continues to be a sink. It should be mentioned that no dramatic change in the global marine productivity is observed in our simulations. This arises from the fact that the modifications of the oceanic properties that affect this productivity (stratification, meridional overturning, …) are rather moderate. The effects of climate change are however not negligible. In particular, the decrease in sea-ice extent predicted by the model results in a longer growing season and a larger nutrient uptake (especially silica) in polar regions. As a result, by the end of the 23rd century, silica concentrations in the upper 100 m of the Southern Ocean drop by as much of 30% for the most extreme forcing scenarios.
