Coupled Model Errors in the Tropical Atlantic in CMIP5 and their impact on the reliability of climate projections.
Lead Research Organisation:
University of Reading
Department Name: Meteorology
Abstract
Many of the climate models that contributed to the third coupled model intercomparison project (CMIP3) which were extensively used in the IPCC 4th Assessment Report share common biases in the tropical Atlantic. These biases are characterised by a warm sea-surface temperature (SST) bias in the Eastern Tropical Atlantic and cold SST biases in the western tropical Atlantic and either side of the equator. These SST biases are associated with too strong north-easterly trade winds and too weak easterly flow on the equator. SSTs play a large role in determining regional weather systems including the monsoons, but are also partly determined by the regional circulation. Through the large-scale circulation of the atmosphere, sea surface temperatures in one region of the globe can also influence the weather remotely.
The relationship between the SST biases and the wind biases is consistent with their development through a coupling between biases in the atmospheric and oceanic components of the model which reinforce each other, and as such it is difficult to determine the original source of the bias.
In this project we propose to assess the quality of the models contributing to the latest model intercomparison project (CMIP5) which will be the main source of information to inform the 5th Assessment Report of the IPCC. We will use the contributions to the CMIP5 archive which have been initialized from the observed state of the atmosphere and ocean and run for 10 years to examine how these errors develop and identify those processes which contribute to the development of the biases to inform future model development. We will also identify the impact these biases have on our confidence in near term (next 10 -20 years) prediction of the climate, in particular in the regions around the tropical Atlantic (e.g. Africa and the Amazon) but also globally.
The relationship between the SST biases and the wind biases is consistent with their development through a coupling between biases in the atmospheric and oceanic components of the model which reinforce each other, and as such it is difficult to determine the original source of the bias.
In this project we propose to assess the quality of the models contributing to the latest model intercomparison project (CMIP5) which will be the main source of information to inform the 5th Assessment Report of the IPCC. We will use the contributions to the CMIP5 archive which have been initialized from the observed state of the atmosphere and ocean and run for 10 years to examine how these errors develop and identify those processes which contribute to the development of the biases to inform future model development. We will also identify the impact these biases have on our confidence in near term (next 10 -20 years) prediction of the climate, in particular in the regions around the tropical Atlantic (e.g. Africa and the Amazon) but also globally.
Planned Impact
The two potential impacts of this work relate to improvement of climate models used for climate projections and an assessment of the confidence in our climate projections, particular for the regions around the tropical Atlantic.
The most likely economic and societal benefits of this work relate to better informed decisions on the mitigation and adaptation strategies in response to climate change. Firstly through improved confidence in our climate projections as a result of improved models, this impact is likely to take some time to realize as it relies on a second step of improving the models based on the new understanding developed. The second is a potentially more immediate impact, through an improved understanding by policy makers of the levels of confidence in climate projections, and hence making decisions based on climate information with a realistic assessment in the uncertainty of that information. The obvious route to generate this impact is through an effective engagement with the IPCC AR5 processes. We are fortunate that there are lead authors in both of the most relevant chapters within NCAS Climate and we are therefore in a good position to ensure that they are aware of our results and to get early feedback on them in the assessment process.
Outside of the immediate impact of this work the techniques used will be widely applicable to the assessment of reliability climate model projections of the future climate and the involvement of the researchers in an international project and the personal skills involved in liaising with a large group of people and delivering research to a tight deadline are widely applicable across a range of professional sectors.
The most likely economic and societal benefits of this work relate to better informed decisions on the mitigation and adaptation strategies in response to climate change. Firstly through improved confidence in our climate projections as a result of improved models, this impact is likely to take some time to realize as it relies on a second step of improving the models based on the new understanding developed. The second is a potentially more immediate impact, through an improved understanding by policy makers of the levels of confidence in climate projections, and hence making decisions based on climate information with a realistic assessment in the uncertainty of that information. The obvious route to generate this impact is through an effective engagement with the IPCC AR5 processes. We are fortunate that there are lead authors in both of the most relevant chapters within NCAS Climate and we are therefore in a good position to ensure that they are aware of our results and to get early feedback on them in the assessment process.
Outside of the immediate impact of this work the techniques used will be widely applicable to the assessment of reliability climate model projections of the future climate and the involvement of the researchers in an international project and the personal skills involved in liaising with a large group of people and delivering research to a tight deadline are widely applicable across a range of professional sectors.
Organisations
Publications
Toniazzo T
(2013)
Development of warm SST errors in the southern tropical Atlantic in CMIP5 decadal hindcasts
in Climate Dynamics
| Description | Historical simulations from 25 CMIP5 models were examined in terms of the mean biases in the Tropical Atlantic (and the seasonal cycle of these biases). There were qualitatively consistent biases all models, with positive SST biases in the Southern Eeastern Tropical Atlantic (SETA) and Eastern Equatorial Atlantic, and concomitant precipitation and wind biases with a southwards shift of the ITCZ, drying over South America and increased precipitation over the SETA and Africa. These errors are very similar to those in the CMIP3 historical integrations, which indicates that little or no progress has been made in addressing these biases in the intervening period. Decadal hindcasts from 3 modelling centres contributing to the CMIP5 intercomparison project were used to analyze the time evolution of these biases and the mechanisms responsible for them. These comprise 4 hindcast sets (2 with Canadian Centre for Climate Modelling and Analysis's CANCM4, and one each with National Centre for Environmental Predictions's CFSv2-2011 and with the Met Office's HadCM3), In all 3 models the signal of the climatological biases is well established within the 1st year and by the second year it is fully developed. The error evolution in the CANCM4 model is very consistent between hindcasts. In CFSv2, whilst there is considerable interannual variability in the early stages of the error evolution, no systematic relationship between error growth and the initial state of the coupled system with regards to large-scale modes of variability was identified. Despite the broadly similar climatological biases different mechanisms appear important in each model. In CFSv2 there is rapid initial surface warming as a result of excessive surface shortwave fluxes, which is subsequently reinforced by ocean dynamics, such that the resulting bias is greater than would be generated by the flux error alone. In CANCM4 and HadCM3 the errors grow due to propagation of anomalous sub-surface warming in response to tropical wind errors. However the timescales for development of these errors in the two models are very different with tropical wind errors in CANCM4 present in the first few days of integration, but a progressive growth HadCM3 coupled with the SST biases, through what appears as a positive, non-local feedback loop mediated by ocean dynamics. In spite of some similarities in the error growth, especially the coupled nature of the full error growth and the link between the Equatorial Atlantic and SETA, the diversity of initial error growth mechanisms indicates a more complex picture than the role of local errors in the wind and buoyancy forcing often blamed for the errors. |
| Exploitation Route | These results could be used by the modelling centres concerned to inform the development of their individual models. The techniques used to identify the mechanisms for error evolution could be more widely applied by other modelling centres and for other regions of the globe. |
| Sectors | Environment |