Economic consequences of CMIP diversity and targets for CMIP7

The CMIP ensembles represent a partial exploration of our uncertainty in the future physical climate under various scenarios for future greenhouse gas emissions. They are thus valuable tools for exploring the potential consequences of climate change for society. One aspect of this is the impact on global and national economies. In the economics literature this is often addressed through a “damage function” which relates economic damages to national, regional or global changes in temperature.

Here we will present an assessment of the economic damages implied by the CMIP6 ensemble for various nations/regions, different Shared Socio-Economic pathways, and, crucially, a variety of different damage functions. A number of important factors will be highlighted including:

• The uncertainty in economic damages which arises from the chaotic nature of the climate system, characterised by those CMIP6 models with relatively large initial condition ensembles.
• The relative consequences for economic assessments of uncertainty in the damage function, the choice of CMIP6 model (model uncertainty), chaotic uncertainty (initial condition uncertainty), and scenario uncertainty.
• How these factors vary by country and region.

 

CMIP6 only represents a limited exploration of uncertainty in the physical climate response and there is also considerable uncertainty in the damage functions beyond that currently explored in the literature. These represent deep uncertainty. We will present plans for future work to embed more thorough explorations of epistemic uncertainty into future analyses, including the consequences of crossing tipping points. These considerations are valuable when considering the design and implementation of the CMIP7 project. What would be the most useful design characteristics if the target were economic assessments? We will address this question in terms of both the size of initial condition sub-ensembles, the diversity of models included, and the value of a mixture of higher and lower resolution model implementations.