Paper: Evaluation of CMIP5 palaeo-simulations to improve climate projections

[donsutherland1]

A newly published paper evaluates the performance of the climate models using paleoclimate simulations.. In general, CMIP5 performed well on a global scale, but not very well on a regional basis. 

 

Excerpts from the paper's conclusion:

 

Evaluation of the CMIP5 palaeosimulations demonstrates the value of including past climate states as targets for model intercomparison. Systematic examination of features that are characteristic of future climate simulations in palaeoclimate experiments and palaeoclimate reconstructions provides an opportunity to determine whether these features are robust characteristics of the climate system, and whether they are features of the actual response of the climate system to changes in forcing rather than model artefacts^{7, 16, 18}. The broad-scale temperature and precipitation responses seen in future simulations are present in palaeosimulations and correctly represented in both LGM and historical simulations. This gives us confidence that the projected changes in land–sea temperature contrast, high-latitude amplification, temperature seasonality, the scaling of precipitation with temperature and the differential precipitation–temperature scaling over land and ocean are reliable. Similarly, the fact that models produce large-scale changes in climate consistent with palaeo-reconstructions for multiple different climate states enhances our confidence in the simulated changes shown in future projections. 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 reliable.

 

Nevertheless, the modest overall skill of the CMIP5 models for the mid-Holocene and LGM shows the limitations of the current generation of models. Specifically, the models are unable to reproduce the magnitude of changes in regional climates, even when taking into account the uncertainties inherent in the palaeo-reconstructions...

 

http://www.nature.com/nclimate/journal/v5/n8/full/nclimate2649.html

[nflwxman]

A newly published paper evaluates the performance of the climate models using paleoclimate simulations.. In general, CMIP5 performed well on a global scale, but not very well on a regional basis. 

 

Excerpts from the paper's conclusion:

 

Evaluation of the CMIP5 palaeosimulations demonstrates the value of including past climate states as targets for model intercomparison. Systematic examination of features that are characteristic of future climate simulations in palaeoclimate experiments and palaeoclimate reconstructions provides an opportunity to determine whether these features are robust characteristics of the climate system, and whether they are features of the actual response of the climate system to changes in forcing rather than model artefacts^{7, 16, 18}. The broad-scale temperature and precipitation responses seen in future simulations are present in palaeosimulations and correctly represented in both LGM and historical simulations. This gives us confidence that the projected changes in land–sea temperature contrast, high-latitude amplification, temperature seasonality, the scaling of precipitation with temperature and the differential precipitation–temperature scaling over land and ocean are reliable. Similarly, the fact that models produce large-scale changes in climate consistent with palaeo-reconstructions for multiple different climate states enhances our confidence in the simulated changes shown in future projections. 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 reliable.

 

Nevertheless, the modest overall skill of the CMIP5 models for the mid-Holocene and LGM shows the limitations of the current generation of models. Specifically, the models are unable to reproduce the magnitude of changes in regional climates, even when taking into account the uncertainties inherent in the palaeo-reconstructions...

 

http://www.nature.com/nclimate/journal/v5/n8/full/nclimate2649.html

 

Don, thanks for this link. From a holistic stand point this makes sense.  The climate system, while very complex regionally, is a closed a system and runs on relative simple thermodynamic principals.  This is based mostly on energy in versus energy out (in terms of temperature response).  The community has, in my mind, fallen short of producing regional phenomenon (including arctic warming) without tweaking ECS to an absurd degree.  Unfortunately with AGW, effects will be regionally felt dis proportionally.  The global model suite of CIMP5 has performed reasonable well on a decadal basis according to retroactive studies on the topic, but I'm sure folks in Bangladesh don't care about the global model performance..

[donsutherland1]

Don, thanks for this link. From a holistic stand point this makes sense.  The climate system, while very complex regionally, is a closed a system and runs on relative simple thermodynamic principals.  This is based mostly on energy in versus energy out (in terms of temperature response).  The community has, in my mind, fallen short of producing regional phenomenon (including arctic warming) without tweaking ECS to an absurd degree.  Unfortunately with AGW, effects will be regionally felt dis proportionally.  The global model suite of CIMP5 has performed reasonable well on a decadal basis according to retroactive studies on the topic, but I'm sure folks in Bangladesh don't care about the global model performance..

Thanks. I fully agree with you. These outcomes to date aren't very surprising. Still, it is encouraging to see that early evidence suggests that the latest climate models are doing quite well on a global basis. That's a good step forward and I'm hopeful that over time the regional forecasting capabilities will be improved, as well.