 |
| Titel |
Quantifying uncertainty in multi-decadal climate projections of surface temperature change |
| VerfasserIn |
Dan Rowlands, Dave Frame, Tolu Aina, Milo Thurston, Myles Allen |
| Konferenz |
EGU General Assembly 2011
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 13 (2011) |
| Datensatznummer |
250057753
|
|
|
|
|
|
| Zusammenfassung |
| The much quoted IPCC AR4 ‘likely’ uncertainty estimate for 21st century surface temperature change, of -40% to +60% of the CMIP3 ensemble mean, is a clear indication that ensembles of opportunity do not span the full range of anticipated uncertainties. As a result, alternative approaches have been developed to test and challenge this estimate. To date these have been based on simplified low dimensional climate models and statistically generated ensembles, in the absence of large ensembles of GCM simulations of transient climate change.
Here we present results from the climateprediction.net BBC climate change experiment, the first multi-thousand-member GCM ensemble exploring uncertainties in transient climate change. The ensemble explores uncertainties arising from perturbations to atmosphere, ocean and sulphur-cycle parameters, natural forcing and initial conditions in the UK Met Office HadCM3L climate model. This provides a systematic exploration of uncertainty in the HadCM3L response to historical forcing up to 2000 and the SRES A1B scenario out into the 21st century.
One of the key questions we have considered is how observational constraints (also known as metrics or diagnostics) from the recent past can help us constrain the uncertainty in future projections of surface temperature change. Much research points toward simple measures based on time-averaged climatology providing a weak constraint, and so we have considered observational constraints based on the transient evolution of temperature over the past 50 years, similar to those used in detection and attribution studies. We argue that using a transient constraint is potentially more powerful since, especially for mitigation, it captures the very ability of a model to respond correctly to time changing forcing that we are interested in for the future.
Based on this observational constraint consisting of the regional temperature response over the period 1961-2010, we find a likely range of 1.1–3.1 K for global mean warming in 2050 relative to 1981-2000. This uncertainty estimate should be interpreted as a 66% confidence interval, reflecting the range of warming shown by ensemble members that provide an adequate simulation of the observations at this level of significance. We argue that this lower bound can be explained from simple energy balance arguments, while reducing the upper bound further requires the application of additional constraints based on climatology, which are contestable since they can project rather weakly onto the forecast warming that we are trying to constrain. In the light of this we consider how this uncertainty estimate may evolve into the future as we observe more of the transient response, and conclude with a set of checks that we recommend undertaking when attempting to constrain climate projections using observations in this fashion. |
|
|
|
|
|
|