The ultimate objectives of this programme were to improve the representation of natural phenomena in integrated models assessing climate-related policies and to pinpoint the methodological problems linked to the treatment of any damage. This investigation also induced considerations on the modelling architectures available for coupling climatic and economic models.
The first component of the programme is based on an optimal control model that symmetrically processes activities pertaining to the reduction of greenhouse gas emissions, the prevention of deforestation and the sequestration of carbon in biomass. Two new instruments emerged from our investigations:

a) A carbon cycle model in which the terrestrial element explicitly includes the variations in surface areas and the natural dynamics of the considered biomes for the four regions of the SRES scenarios. The terrestrial part of the cycle comprises two biomes – forests and agricultural lands – in their climacic state, and disturbed areas in transition from one to the other due to changes in land allocation, i.e. afforestation or deforestation. In each region, biomes are classified as temperate, tropical or boreal according to their climatic and biological type.

b) A global temperature evolution model to study the short-term response of the climate (2020-2050). Two variables are selected: the mean temperature of the continents and the mean temperature of the ocean surface. The development of this temperature module was needed to examine optimal policies regarding climate, with constraints both on the acceptable elevation in temperature and on the rate of this elevation, rather than on the concentration level only. While keeping to a cost-effectiveness approach, taking into account these two constraints has the advantage of implicitly evaluating damage in terms of climate corridor through linking with available impact studies, which use the global mean temperature, while adding the effect of the rate of climate change, which the standard damage functions are unable to do.
Initial results stress the importance of taking into account feedback mechanisms between CO2 emissions, land-use changes and the carbon cycle, since this led to revise upwards the concentrations generated by the SRES scenarios by 20 to 80 ppm, thus raising by a similar amount the effort required to keep within a given concentration level and placing more emphasis on the need to slow deforestation rates and to speed up sequestration in the biomass.

Other results underline the weight of the uncertainty regarding the sensitivity of climate. We show that, as long as this parameter remains of unknown value, taking it into account amounts to considering ambitious objectives in terms of emission reductions since this particular uncertainty dominates all other uncertainties. This uncertainty increases the value of information on climate dynamics because, symmetrically, it will induce considerable economic regret if its value eventually turns out to be low. We note moreover that the value of this information explodes as from 2040.
Our investigations also focused on the assessment of damage directly rather than through global constraints. We specifically worked on the non-linearities of climate change: the non-linearity of certain types of damage (disruption of the thermohaline circulation) is represented in DIAM (Dynamics of Inertia and Adaptability Model) by means of a threshold damage function, the uncertainty pertaining here both to the level of damage and to the value of the threshold.

A first set of results stresses the importance of the introduced thresholds and of the uncertainty that affects them: the perspective of a relatively slight decrease in consumption (4%) sufficiently soon and concentrated over time may justify substantial reduction efforts in the short term – whereas such efforts do not appear justified when regular functions are used. It is, moreover, important to note that the optimal strategy is more sensitive to the position of the threshold than to the amplitude of the surge in the damage function.

Since non-linearities in damage may occur at the regional level (such as those revealed when using the IMAGE model to study the sensitivity of regional agricultural production to varying degrees of climate change) as well as at the global level, we focused on the problems linked to the aggregation of the regionalized assessments of impacts. We showed that this method completely occults potential shocks to the most vulnerable regions, which are often among the poorest and therefore of little weight in terms of woldwide income. Moreover, the aggregation procedure currently in use implies the existence of credible compensation mechanisms between regions and minimizes thereby the amount of required transfers. For example, highly local disruptions may, if they threaten the basic needs of the populations, build up considerably and trigger a propagation process to the developed economies, relatively spared by the direct effects of climate change.
This is why the uncertainty margins associated with the regionalized assessments of damage must be highlighted and integrated in the line of reasoning in order to avoid the temptation to analyse in terms of winner and loser regions – as seen in certain circles – which would be an open door to endless controversies about the estimation of damage, controversies that could paralyse all attempts at international coordination.
The authors therefore first defined geographic divisions of the globe so as to satisfy the separate requirements of economics and of the universe sciences. They then proposed a quantification of the uncertainty pertaining to the regionalized assessments of damage.

The underlying hypothesis of this exploratory investigation is that the areas where the responses of the climatic models diverge the most are also those where the uncertainties are the greatest.
The identification of such areas on the basis of a statistical study of recognized climatic models (experiment CMIP1) made it possible to propose a simplified regionalization. Using the variability of the regional climatic response as a measure of the uncertainty, we calculated, for each region, the risk premiums associated with uncertain regional damage. We showed that, given the uncertainty attached to regionalized estimates of damage, the existence of risk aversion in the populations exposed to climate change could incite them to express a high risk premium, in particular when they are less developed and very vulnerable (such as Africa and South-East Asia). At the global level, on the other hand, taking into account uncertainty is much less important than the equity considerations that govern the aggregation of regional damage.
The nature of this work is clearly exploratory. Reflection on the climatic parameters to be selected for the construction of regionalized damage functions (mean temperature, extreme temperatures, temperature range, rate of change, precipitation, etc.) is just beginning.

The third component of the programme is more methodological. It allowed us to lay the groundwork for incorporating the information from climatic models and impact studies in an integrated framework, consistent with the economic theory of decision-making in conditions of uncertainty: expertise and operation of IMAGE, use of the CMIP base, carbon cycle model, investigation on the uncertainty attached to the sensitivity of climate. These will prove all the more useful since part of this methodological programme was devoted to testing a model-coupling technique (TEF-ZOOM) – adapted to the analysis of couplings and feedbacks within complex systems – in the perspective of developing simplified models to study the links between climatic processes and economic responses.

This research programme centred on integrated modelling facilitated organized dialogue between scientific teams from different disciplines – a significant feat and an asset in the perspective of the current constitution of European networks of excellence in integrated modelling.