Regional climate predictions, projections or possibilities?

In a recent edition of EOS (newspaper of the American Geophysical Union), a provocatively titled piece by Roger Pielke Sr. and Robert Wilby, “Regional climate downscaling: what is the point?”, questions the information contained in regional climate model projections for use by the impacts community. Following a description of the various types of downscaling approaches, the authors list a number of caveats related to type 4 downscaling, in which the output of Earth system models driven by radiative forcing changes are downscaled for multidecadal projections. The authors conclude:

“It is therefore inappropriate to present type 4 results to the impacts community as reflecting more than a subset of possible future climate risks.”

The article refers to a recent study by Pielke et al. (2012) which states:

“Neither dynamic downscaling or statistical downscaling from multi-decadal global model projections add proven value to spatial or temporal accuracy that can assist the impacts community in ways beyond what is already available from historical, paleo- or analogue records”. [emphasis in original]

The latter quote poses a direct challenge to the regional climate modelling community, some of whom reside here in CSAG. I’d certainly be interested in hearing the responses of those actively involved in downscaling activities. However, at risk of offending and/or appearing to pander to my colleagues, in this blog I address the former quote; a cowardly but, I hope you agree, sensible decision.

If climate scientists are only able to offer “a subset of possible future climate risks” then modelling centres really can’t purport to be offering climate predictions, on multi-decadal or longer time scales. This isn’t a wholly controversial point however. Indeed the IPCC and most modelling centres disseminate information about future climate in the form of projections, rather than predictions or forecasts. Nevertheless, do users fully appreciate the nuanced language of climate scientists? Moreover, do climate scientists fully appreciate the often far-from-nuanced interpretation of climate model projections?

The Ensembles project website offers a nice explanation of the difference between the terms prediction and projection. However, climate scientists aren’t always that careful when it comes to the separation of these terms. Even the UK Met Office is guilty. In their guide to climate change on their website, they carefully explain that climate models are used to make climate change projections. So far so good. Yet in the following section, they include animations showing “predicted” temperature rise up to 2100. Now this is either sloppiness on the part of the website editors or they really do predict temperatures will evolve according to the animations shown; of course the animations are emissions scenario dependent so that gives them a good get out if emissions follow a different path.

I am not convinced distinguishing between the terms projections and predictions really communicates the full story. Perhaps instead of using the word projection, consistent with the concluding statement of Pielke Sr. and Wilby (2012), we ought to refer to the output of regional climate models as climate possibilities. This seems, at least to me, a better way to articulate what we actually mean. The word projection implies some kind of expectation that the climate system will act in a similar way to the model whilst the word possibility implies that there is a chance the climate system will behave in a similar way to the model but equally that there is a likelihood the system may do something entirely different.

In the scientific discourse on climate modelling I have no problem with the use of the word projection. It is an accurate term to describe what a model is doing: projecting the dynamic behaviour of the climate system within a model domain. However, when we want to express how model projections relate to the real climate system, extending the use of the word projection may be misleading, particularly when such projections are being used to guide real-world policy decisions. Climate models are mathematical approximations to the climate system, they are not (and never will be) isomorphic to the real climate system. What is possible in “model world” may not be possible in the real system and vice versa. Therefore we must be wary about overstating our confidence regarding the behaviour of the real climate system based on model predictions, projections and even possibilities. I feel impelled to qualify that statement by saying that models are still valuable tools and essential for guiding climate change adaptation and mitigation policy. We would be foolish to dismiss model output because of model inadequacies and therefore ought to view the output as possible until proven otherwise.

In any case, I suspect climate modelling centres won’t start issuing “climate possibilities”. It certainly doesn’t have the same pizzazz as “climate projections” and given ever increasing pressure to provide policy-relevant climate science, simply saying “the models indicate a range of future climate possibilities” sounds like an admission of ignorance and creates an impression that climate scientists really don’t have a clue what to expect. Using the language of possibility also raises the troubling prospect of having to consider the inverse. Are we confident enough to profess that certain future climate states are, in fact, impossible?