Defining a rapidly changing climate

If the climate is changing, what is it that’s changing? Climate, they say, is what you expect, while weather is what you get. But if climate is changing, how do we know what to expect?

That may sound flippant but it has become quite a serious question in climate science circles of late, for two reasons: communicating climate science to the public, and dealing with the statistical problem of defining a moving target.

On the first question, for instance, we have a typical member of the public asking on the RealClimate open thread“One thing that has confused me is how long it takes for weather to become climate.”

One of the regulars replied: “WMO (World Meteorological Organization) states that climate is 30 or more years of weather data.” Another backed him up, saying, “The traditional answer is 30 years or thereabouts.” ‘SecularAnimist’, another regular, answered at greater length:

I think this question is increasingly irrelevant, and the “traditional” answer is becoming obsolete.

The question was relevant when we were asking whether the various atmospheric conditions, processes, events and patterns of events that comprise “climate” are in fact changing, and wanted to know over what length of time we’d need to observe those phenomena as ever-changing, short term “weather” to be able to conclude that the changes are sufficiently long-term to be considered “climate” change. But we already know that the climate is changing, and will continue to change, as a result of our CO2 emissions. We don’t need 30 more years of observations to tell us that, now.

And the “traditional” answer is obsolete because it presumes that the Earth’s climate is sufficiently stable, and changing so slowly, that it really does take 30 or more years of observations to detect any long-term, large-scale change. That’s no longer the case, because the climate system is being driven to change more rapidly and extremely than it has ever done in human history. It’s unlikely to take another 30 years for the American midwest to become desert. It’s unlikely to take another 30 years for the Arctic sea ice to disappear completely during summer — with all of the prodigious effects that implies.

There is every reason to expect that permanent, large-scale, dramatic changes, which cannot reasonably be called anything but “climate change”, may now occur on time scales of a few years, rather than a few decades, as would have been expected in the pre-AGW world.

The more technical problem with the definition of climate is that weather needs to be averaged over enough annual cycles to iron out the bumps caused by specific weather events (that’s where the ’30 years’ comes from) but if, for instance, the averages of the 1980s, 1990s, and 2000s are successively quite different from each other then a ‘climate’ defined by the 1980-2010 average becomes meaningless.

Should we just average over a shorter period, then? No, because known multi-year cycles such as El Nino will distort the result too much. As far as I know, the experts are still working out what to do about that. Meanwhile they tend to define our baseline climate by the period from 1951-1980. James Hansen, whom I have mentioned before, presents the reasons very sensibly in a very recent discussion of his recent ‘climate dice’ paper:

Studies of climate change generally use some base period to define an average climate and calculate “anomalies” relative to that average, i.e., climate anomalies are the deviations from that average climate. In our papers we used 1951-1980 as the base period.

Global temperature change over the past century (Fig. 1) helps us discuss possible effects of the choice of base period. Our choice of 1951-1980 as a base period has several merits:

1) The period 1951-1980 is prior to the large warming of the past few decades. If we wish to examine the effect of that global warming on climate, we must compare with the climate that existed prior to that warming.

2) The 1951-1980 period has the best global data coverage and can best characterize climate variability. Spatial coverage of data was poorer at earlier times.

3) 1951-1980 was the base period used by the National Weather Service and other researchers when we made our first analyses of observations and climate simulations. For comparison with these early analyses and climate simulations we should use the same base period.

4) Many of today’s adults, baby-boomers, grew up during 1951-1980, so it is recent enough for many people to remember what the climate was like.

The whole discussion is well worth reading, since it presents the key findings of the longer and more technical paper in a very approachable way. Download it (pdf) from here.

Leave a Reply