A thermometer too far

This article is very heavily based on the orginal work done by E.M.Smith which can be found here.

If we want to understand how the global temperature changes over time, we need a temperature record over time, and one that covers the globe.

Lets look at how GIStemp create their record. The system and data GIStemp use is more or less the same as HADCRUT3 so the comments below apply to both the world’s most important global temperature records and to pretty much to all the other less important surface temperature records.

GIStemp attempts to create a temperature history for the whole globe with full coverage over a long period of time which means it should be based on measurements taken over time in unchanging circumstances in as many different spots around the globe.

Unfortunately the data starts with just one thermometer in Germany in Berlin in 1701.

Over time more thermometers were added to this record and most of these early thermometers are located in either the USA or Europe. As time passes and as new thermometers are added to the record the area with some consistent thermometer readings slowly migrates south and to both the new, and old, worlds. Eventually, about 1900 A.D., there are sufficient thermometers on the globe to get a partial idea what is happening. But climate is subject to cyclical changes. Some, like the Pacific Decadal Oscillation, have about a 40 to 60 year full cycle length. Others, like solar cycles that run 178 years, and Bond Events – a 1500 year cycle, are a bit longer. A 100 year record is inadequate to allow for these events.

Here is an animated film that shows the distribution of surface measuring stations each year from 1880 to 2007, red means there is an interruption in the data for that year.

You may have noticed that towards the end of the animation above the number of dots, each representing a temperature measurement station, appeared to decline sharply. That is not an error. They really did decline and this strange phenomena is described here.

At its core, GIStemp tries to bridge this gap, both in time and in space, between the one thermometer and the globe, and between the 100 years and the 1500. Under ordinary circumstances of scientific inquiry this would be a worthwhile and useful thing to do, to try to assemble a very rough outline of temperature changes over the last 150 years or so no matter how incomplete or imperfect. But these are not ordinary times and now this temperature record has taken on huge political significance and could be instrumental in convincing the worlds leaders to make fundamental changes to the way the global economy is managed, changes which could adversely impact on the lives of billions. So we need to understand the limits of this record temperature. How good is it? What are its weaknesses?

The temperature record is unevenly distributed and changes over time. And yet we have a process that says it can stitch all this together and come up with a simple set of statements about something called the average global temperature over a period of time. How does GIStemp do this?

First GIStemp deletes everything older than 1880 because it says data from before then is unreliable and too patchy. This sounds reasonable and is done for good reasons but it does have the effect of making the temperature record start at the end of the Little Ice Age. Starting at the end of a cooling period does mean that the likely direction is going to be up.

In many cases, our thermometer record is made of fragments. A thermometer may appear in the record for a decade (sometimes less) then disappear just as quickly. The Second World War caused a great ‘drop out’ of Pacific Island thermometers, for example. The growth of jet travel by tourists added thermometers at vacation spots around the globe many at tropical vacation destinations, but not all these kept going, some came and went. Thermometers recording the temperature in “the same place” might actually move around a fair bit over time. So we have a thermometer here, and it moves there. Two records from different places. One over grass near the woods, the next over tarmac at the airport. GIStemp tries to stitch these patchworks together into a smooth quilt of coverage. Some thermometers get stretched this way or that (over time and over space). Some get their temperatures adjusted higher or lower (via a thing called “The Reference Station Method”) to better join with their neighbours. Where needed, missing data is often estimated (i.e.made up with an educated guess) to try and glue the bits together. If a piece of data, even after such a stretch, is shorter than 20 years, it gets thrown away.

In the end, we are still left with gaps. The entire southern hemisphere ocean band has less than 1% of the thermometers, and those are at the airport on a few specific islands for the most part. So we have a patchwork quilt, but with some rather large holes, and some pieces are stretched out of all recognition. A thermometer may be stretched to 1000 km away. Rather like saying that the temperature in London is a good proxy for how hot the beach is in the south of France.

Some places have changed over time. Cities grow, and get hotter, as they fill with cars, tarmac, heaters or air-conditioning vents, airports and jet engines, and coal or nuclear power plants.These sorts of developments make the local area hotter so any thermometer located in such a developing area would give a false warming signal in the data. To adjust for this, GIStemp looks at “nearby” stations up to 1000 km away and guesses who is rural and who is urban and “adjusts” for it. Unfortunately, like all guesses, this sometimes does not work well. Large airports are often marked as “rural” since they have few residents living there even though all that tarmac and jet exhaust will up push the local temperature. The largest US Marine Air Station, Quantico Virginia, is classed as rural, for example. The readings from Pisa in Italy are altered with reference to Hohenpeissenburg on the German approach to the Alps as a ‘nearby’ rural station and Pisa promptly has it’s past made colder which has the effect of making its present-day look even warmer in comparison.

So we’ve ironed out our quilt, even if some bits stuck to the iron and got scorched a bit and others were melted and smeared. At this point, the globe is divided into a “grid” of “boxes”. But still we have “holes” in time and in space. Lots of the boxes are empty – there is just no data at all for those many empty parts of the global grid.

The data that we do have (after the stitching and stretching and ironing) are now assumed to be pristine and pure and suitable for telling us reliable things about all the places where we have no data. A station with a record (a record that as we have seen might already have been adjusted and occasionally guessed) may now be used to fill in a set of the empty boxes on the grid up to 1200 km away. This means, for example, that the airport on Diego Garcia can “fill in” the ocean covering an area roughly the size of Western Europe.

So we finally have our global grid of temperatures.

In the final steps, the grid of boxes is compared to the past for those grids of boxes (said past having been dutifully made up, adjusted, estimated etc if need be) and an “anomaly map” is made which would then shows that the ocean 1200 km out to sea (but actually based on the thermometer reading on the tarmac at the new military jet airport in Diego Garcia) is now warmer than when a passing ship dunked a bucket in it during a passage of the 1950’s.

This is not bad science.

This is probably about the best we can get under the circumstances.

We can’t go back in time and add more thermometers.

But can this system give precise measurements of changes in the global temperature down to a few decimal points with a high degree of certainty? Can it say things like the global temperature has gone up by 0.72 of a degree in the last 100 years? I think the best it can do is say things like “its got a bit warmer in places – probably by less than one degree but its all a bit uncertain”.

And that is a very different message than we get about climate science in the public debate.

Here we are told that the science is done, that there is no uncertainty, that all the facts are known and that all thats left to be done is turn out the lights.