Rethinking the Default Construction of Multimodel Climate Ensembles

Abstract

We discuss the current code of practice in the climate sciences to routinely create climate model ensembles as ensembles of opportunity from the newest phase of the Coupled Model Intercomparison Project (CMIP). We give a two-step argument to rethink this process. First, the differences between generations of ensembles corresponding to different CMIP phases in key climate quantities are not large enough to warrant an automatic separation into generational ensembles for CMIP3 and CMIP5. Second, we suggest that climate model ensembles cannot continue to be mere ensembles of opportunity but should always be based on a transparent scientific decision process. If ensembles can be constrained by observation, then they should be constructed as target ensembles that are specifically tailored to a physical question. If model ensembles cannot be constrained by observation, then they should be constructed as cross-generational ensembles, including all available model data to enhance structural model diversity and to better sample the underlying uncertainties. To facilitate this, CMIP should guide the necessarily ongoing process of updating experimental protocols for the evaluation and documentation of coupled models. With an emphasis on easy access to model data and facilitating the filtering of climate model data across all CMIP generations and experiments, our community could return to the underlying idea of using model data ensembles to improve uncertainty quantification, evaluation, and cross-institutional exchange.

Is Institutional Democracy a Good Proxy for Model Independence?

“…Another interesting result is that of the Hadley Centre models, which despite their important differences-they belong to different generations-nevertheless show striking agreements in their projections. This is consistent with Rauser et al (2015), who showed that successive generations of climate models can overlap significantly in terms of performance. While the nature and meaning of similarities between Hadley Centre models are beyond the scope of this study, several questions are nonetheless prompted: from the role played by the shared radiative transfer parameterization package (Edwards and Slingo 1996) to the impact of developing models under the same ''culture,'' including here validation and tuning methods.…”

Evaluation of CMIP5 palaeo-simulations to improve climate projections

“…4). There is growing feeling that future analyses of climate change and its impacts should be based on cross-generational ensembles of model output constrained by observations 68,69 , and palaeo-evaluations support this approach, not only because there seems to be little improvement between different generations of models, but also because some models are better than others at reproducing the magnitude and patterns of large climate changes.…”

“…We have re-examined this finding by combining the CMIP5 and PMIP2 ensembles (following the approach suggested in ref. 69), taking the mean of the outputs where more than one integration was carried out by closely related models. This gives a total of 11 simulations and a weak correlation between tropical temperature during the LGM and equilibrium climate sensitivity, which is barely significant at the 90% level.…”

“…It is of concern that the current generation of climate models does not perform better overall than previous generations of models, in terms of either modern climate or palaeoclimate changes 13,69 . On the positive side, this opens up the possibility of using cross-generational ensembles for projections of climate and climate impacts, which would provide a larger ensemble and more robust measurements of uncertainties.…”

What have we learnt from palaeoclimate simulations?

“…The palaeo‐record has the ability to discriminate between models where they show differences in the response to forcing, and again this provides a way of determining which models are more or less reliable. Efforts to improve the skill of climate models based on evaluation using modern climate states are having a declining impact (Knutti, ; Rauser et al ., ), pointing to a need for innovation (Stevens and Bony, ; Palmer, ). We are therefore at a key moment for the climate modelling enterprise to benefit from insights gained from the study of past climates.…”