Monday, August 01, 2011

Centennial scale warming over Japan: are the rural stations really rural?

This is the title of a paper we recently published here. It has been widely observed that Japanese met stations show abnormally large warming over the 20th century - there are about 60 stations that have 100y time series (there's an open access paper here) - but significant uncertainty about the attribution of this to the possible causes. Obviously there's been a lot of urbanisation over this period, which must have affected many of the sites. The Japan Meteorological Agency has produced a subset of 17 stations which it claims are basically rural, but the reasons for their choices don't seem to be very clearly documented, and neither is it clear that these "rural" stations actually are unaffected by urbanisation. For one thing, even these stations still show somewhat greater warming than is seen over most land areas globally (at least when we take the lowish latitude and maritime influence into account, some northern and continental areas have warmed more).

Our idea was that we could investigate this using observations of the surrounding ocean, for which there is also a long-term observational analysis. While we wouldn't expect the ocean and land to warm exactly the same, there should be quite a strong link given the physical proximity. Furthermore, this land/ocean relationship can be investigated with the IPCC AR4 models. Globally, these models give quite a wide range of land/ocean warming ratios (eg) but over a local region temperature anomalies are generally well correlated.



This pic is the main summary of the analysis. The blue dots show the results of the IPCC AR4 models, ocean (x-axis) versus land (y) warming over the 20th century, just considering the region around Japan. There's clearly a strong linear relationship with a slight amplification of 13% over land. The observed warming over the ocean is 0.8C as indicated by the vertical line, and the various coloured dots on that vertical line represent different subsets of the land stations - red is the largest cities which have warmed (and grown!) massively, and the three green dots are different attempts at selecting rural stations (including the JMA set). These all lie pretty close to the regression line, suggesting that the observed warming at those stations is not really out of line of what would be expected from the surrounding ocean warming and thus there is no evidence of substantial urbanisation in these stations. It's also clear that the real ocean has warmed more than in almost all of the models, but we didn't look into the reasons for this, and it could just be due to natural variability - the obs aren't even outside the range of the models, so there is not really any cause for concern here.

7 comments:

SteveF said...

I can forsee a blog post by Roger Sr coming up, moaning about you not citing his work (i.e. like all of his blog posts these days). Even if his work is barely relevant, he'll manage to find some way to complain.

James Annan said...

He's still going, is he? I gave up on him a long time ago.

EliRabett said...

You might want to send a copy to Chiefio, the flavor of the month for the rejectionists about a year and a bit ago, who claimed that the observed warming was because all stations above 300 meters were closed.

This is a clever method. Can it be bootstrapped inland??

James Annan said...

I guess it will work inland reasonably well to a distance of a few hundred km or so, which obviously covers Japan but will leave big holes in the middle of the main continents. But this is a bit of a rough guess. GISTEMP has a smoothing radius of 1200km, but this is the range at which anomaly correlations are pretty small.

I suppose that could still be interesting for many cities, which tend to be coastal.

Anonymous said...

Results from IPCC AR4 models. That's really going to convince the skeptics.

James Annan said...

Surprising as it may seem, convincing the sceptics is not high on my list of priorities.

EliRabett said...

JA and A. Turn it around, this is a pretty good test of the GCMs, or at least the consistency between sst and surface temperature.