Did you ever wonder why organisations such as NASA, NOAA, Berkeley, CRU and the UKMO use anomalies rather than actual values when describing global temperatures? And why is it they only calculate monthly or annual anomalies and not daily ones? And why do they all seem to calculate their anomalies using different periods for their long term averages? Did you ever wonder just how accurate the estimates of global are? Well I do, and that’s why I decided to do create my own do-it-yourself global temperature series to answer some of these questions, and here is how I did it:-
- I downloaded gridded 0.995 sigma surface air temperatures from 1948 from the ESRL website via FTP. These values are on a 2.5° x 2.5° grid for the whole globe so the resolution is quite low. There are four temperature values for each day for 00, 06, 12 & 18 UTC. This makes even the compressed size of these files large (71 x 21 MB) for each year of data they contain, in total 7.79 GB.
- Once I had downloaded the last 70 years or more of files I had to convert them from the NetCDF format into something readable. So I download a free conversion utility program called NCDUMP to do this with, and converted the raw NetCDF file into a CDL text file. I then parsed the CDL text file that contained the six hourly data for the whole year into individual six hourly global temperature files (71 x 365 x 6 files of 53 KB)
- Because this is gridded data rather than observational data there is no requirement to derive a temperature over the sea from sea surface temperatures, neither are adjustments required for urbanisation, but because it’s simpler to work with a single daily global temperature file, I created a single mean daily temperature file for each day of the year since 1948.
- I now calculated a global and zonal values for each day, and whilst I was at it I decided to slice the globe into five zones, Arctic, northern mid-latitude, equatorial, southern mid-latitude and Antarctic, and work out mean temperature for these areas as well. I did that using the latitudes of the Arctic and Antarctic circles and those for the tropics of cancer and capricorn as a guide. I did have to employ a weighting factor to calculate a more meaningful value because the area of each grid gets smaller as you move from the equator and to the poles. So nothing fancy, I just multiplied the latitude of each grid point by the cosine of its latitude. I did find in a bit of research that the area within the arctic circle makes up only 4% of the total surface area of the globe. So even though temperatures are rising in the Arctic faster than anywhere else it’s still a relatively small area globally.
- I require long term averages to calculate anomalies, but because computers are so fast these days, I decided to do this on the fly and let the user (me) select whatever anomaly period he wanted to use, beit 1950-1981, 1981-2010, or even for the whole record 1948-2018 which I prefer.
- Finally I set to work to add some extra functionality to my Reanalysis Temperature application so that I could visualise the global temperature data that I had created, in both tables and graphs. And that’s all there was to it!
How do DIY values compare with GISTEMP?
Well the overall shape of the bar charts are vaguely similar all things considered. My estimates seem to have captured the recent warm years quite accurately, with 2016 the warmest, followed by 2017 and then 2015 just as in the GISTEMP series. The biggest difference is that my estimated anomalies are much lower in amplitude than those in the GISTEMP series. The linear trend for the GISTEMP series (since 1948) shows a warming trend of +0.146°C per decade compared to just +0.083°C per decade in my DIY calculations, which is over 40% lower an increase in global temperatures. These findings might find favour from any climate change deniers, but they do look low, and why this is is rather puzzling to me. Who knows, maybe NOAA apply an additional weighting factor or do a bit of jiggery pokery when they calculate their GISTEMP series?
So what is the global mean temperature at the moment?
Well the average DIY global mean surface temperature is 9.521°C to be exact. Figure 1 shows the latest daily global temperatures, and as you can see the mean daily value varies from around 8.4°C to just over 11.1°C through the year and traces a curve that you would normally expect to find from a place in the northern, rather than the southern hemisphere. I imagine global temperatures follow this kind of trace because the bulk of the world’s land mass lies north of the equator.
As you can see the great thing about my DIY series is that you can watch values change on a daily basis (fig 2). I now don’t have to wait three weeks for the Americans to release their latest monthly anomalies, or the even longer six or more weeks for the UKMO to do the same thing. Reanalysis data is updated at least a couple of times a week, and although changes are slow its good to have the immediacy that daily values offers. Bar charts of monthly, seasonal and annual values or anomalies are all very well, but personally I prefer to see an annual 365 day moving average to watch new trends emerge or old trends fade away. Here for example is a 365 day moving average line chart for the series since 1948 (fig 3).
Here’s a closer look at the global 365 day moving average since 2000 (fig 4). You’ll notice that the linear trend since 2000 is now much steeper at +0.171°C per decade, and closer to the linear trend from the GISTEMP and CRUTEM series. You can also see how global temperatures have recovered since mid 2018 after around two years of global cooling since the record warm spurt of 2016. I personally think that my DIY temperature series is just as accurate and sensitive to global fluctuations than both of these temperature series. I think I can safely say that because it’s impossible to verify who is correct, because let’s face it what an impossible task it would be to accurately calculate a single temperature for the whole globe.
Finally, here’s a table of means and anomalies since 1985 (fig 5). It’s noticeable that in those 34 years just how few negative anomalies feature in any of the five zones.
Some of my readers may remember that I did some similar research into global temperatures from reanalysis data around six or seven years ago which I wrote up into an article. I’d just like to say that this time I started right from scratch and have taken as much care as I can to make sure that the DIY series is as accurate as I can make it. The logic is simple, but I must be missing something because the size of the anomalies does not match the results from other estimated global temperature series. The one weakness of my DIY temperature series is that it does depend on the gridded reanalysis surface temperature data from NASA being accurate as possible, and of course it may well not be.