What is IsItHotOrNot?


This is the landing page for the Twitter account IsItHotOrNot. I’ll probably move it to a proper home at some point.

IsItHotOrNot displays the global temperature anomaly, month by month. Currently the data is sourced from GISTEMP, specifically the “Global-mean monthly land-ocean temperature index”. The file, GLB.Ts+dSST.txt, also has annual and seasonal values, but I don’t currently use those.

What does the number mean in “2009 Nov +0.680, hottest November on record”? It’s November’s temperature anomaly (in degrees C). The anomaly for November is the difference between this November and the average November temperature between 1951 and 1980, the base or reference period. This year’s November was 0.68 degrees C warmer than the average between 1951 and 1980. Obviously the anomalies for other months are calculated in a similar way. Note that this method takes account of the fact that Novembers are, on the whole, colder than Junes, by using the average November temperature between 1951 and 1980, not the average temperature. In other words it does seasonal correction.

It would be nice to say “about 1.0 degrees C” warmer than pre-industrial levels, like the EEA do, and I can do that by simply adding 0.3 to all the values, but since the errors for the 1850 to 1899 (which is what is usually used when you see the phrase “pre-industrial”) average are larger (on account of having fewer records), I decided not to do that. You can wing the error and add 0.3 to all the values if you like.

There are other temperature indexes available: NCDC do one; there’s the HadCRUT3 diagnostics; the Japanese Metereological Agency have their global average surface temperature anomalies. The different indexes use different reference periods (NCDC use 1900 to 1999, HadCRUT3 use 1961 to 1990, JMA use 1971 to 2000), the only effect of which is to move the graphs up and down the page (it does mean that if you want to compare the time series from two different providers, you need to rebase them to give them a common reference period).

Where does the data come from? Well, from GISTEMP, but the land and ocean index is a data product synthesized from other sources:GHCN (land), USHCN (USA land), HadSST2 (ocean), Reynolds SST (ocean), SCAR READER (antarctic land). The other indexes use more or less the same data, the differences are mostly in how it is processed.

Why GISTEMP? All the indexes are basically the same, so it doesn’t really matter which one I pick. Because of my work on Clear Climate Code I happened to be familiar with their datasets, so that was simpler from a programming perspective. The differences between the indexes are typically very small, but it does mean that the statements like: “second hottest September on record” are very sensitive to the exact dataset used. The JMA had September 2009 tying for first place, whereas GISTEMP have it second. I regard the “hottest November on record” stuff as nothing more than a popular discussion point, not an absolute truth.

7 Responses to “What is IsItHotOrNot?”

  1. Gareth Rees Says:

    Typo: “month’s” for “months” [ed: fixed, ta]

    Also, is the third decimal place in “+0.680” really justified? The GISTEMP anomalies are given as multiples of 0.01 °C so I think you should give only two decimal places.

    Might be a good idea to put the units in the tweets too?

  2. drj11 Says:

    The third place is not really justified (I was thinking that just this morning). I _think_ it’s there because I noticed that HadCRUT3 reports 3 decimal places, and I thought it might be nice to use that (instead of/as well as GISTEMP at some point). But there’s obviously no reason why I can’t report GISTEMP with 2 d.p. and HadCRUT3 with 3 d.p.. The second decimal place is slightly bogus anyway.

    I think putting the units in the tweets is a good idea.

  3. Gareth Rees Says:

    I don’t think the second decimal place is bogus at all. The 95% confidence interval (I think this is for the error due to lack of station coverage) is around ±0.05 °C, so one decimal place would be too little information.

    I suppose you could put the confidence interval in the tweet: “2009 Nov +0.68±0.05 °C, one of the hottest six Novembers on record, with 95% confidence”

    Hmmm … but when I put it like that, it’s not so twitteriffic. Keep it simple!

  4. drj11 Says:

    You say “not bogus at all” but the error exceeds half of one decimal place. That’s why I said slightly bogus.

  5. Lisa Evans Says:

    David – where do you get your weather station raw data from to test the results of your code against that of the Fortran GISTEMP? Does someone collate the raw weather station data? Lisa

  6. mjb67 Says:

    Has it stopped working?

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: