A comparison between global surface temperature and satellite anomaly datasets

UPDATE: This post was properly criticized for being an incomplete analysis, see this update and note to readers below about why it was incomplete.

Guest Essay By David Dohbro.

Comparing five monthly datasets since 1979. Three land-based data sets consistently report monthly higher values. The land-based data sets report in all most all cases monthly GSTA that are higher than the satellite based GSTAs.

Several global surface temperature anomaly datasets are publicly and freely available. These can be divided into two categories: surface-based and remote (satellite). The first category includes NASA’s GISS (1), NOAA/NCDC’s GSTA –here called NCDC- (2), UK’s Meteorological Office Hadley Centre’s HadCrut 4 (3), and several others. Satellite based GSTA are produced by NSSTC’s –here called UAH-(4) and RSS (5). Each of these produces GSTAs on a monthly basis. One can write a long essay about all the important difference between how each calculates the monthly GSTA, and for sure that is important, but here I simply and only want to compare each data set and see how well (or bad) they match each other. E.g. is one data set consistently reporting higher or lower than the others, are these differences increasing or decreasing over time or not? Etc. I am not assigning any subjective value to these possible differences; I just want to see if there are any differences and if there is a trend in these differences.

I used UAH’s data set as reference. I could have used any other data set as a reference, but it doesn’t matter which data set is compared to which since difference is relative. The satellite based datasets start in the year 1979, whereas the land-based datasets in some cases go all the way back to the year 1850. Hence, only the data from 1979 onward can be compared. That’s still over 35 years worth of data (n>420) and a large enough sample size to say something meaningful about the possible differences between reported GSTA for each dataset. I then simply subtracted the UAH monthly GSTA from the corresponding monthly GSTA of the other data sets (in this case, GISS, Hadcrut 4, RSS and NCDC; Data from January 1979 through March 2014). I then plotted these differences for each corresponding month and performed linear regression through each set of differences (Figure 1). A value of 0 means that the UAH data and the other dataset are similar, a value >0 means the other dataset reports a higher monthly GSTA compared to UAH and vice versa.

Figure 1.

As you can see, the three land-based data sets consistently report monthly GSTA higher than that of UAH. With NCDC > GISS > Hadcrut4 > RSS. NCDC’s data set reports on average a monthly GSTA 0.41°C higher than that of UAH. This value is almost the same as the average monthly GSTA reported by NCDC since 1979 (Table 1).

Table 1: summary statistics of several GSTA data sets and the difference between each of these data-sets with UAH’s

The other satellite based data-set, RSS, reports values rather similar to UAH (average difference of 0.058°C). In addition to the summary statistics, liner trend analyzes (assuming normal distributed data), shows that the difference between GISS and UAH is getting less over time (negative slope), while that for the three land-based data sets is increasing over time. The increase in difference over time between NCDC’ and UAH is the smallest (slope almost 0), followed by Hadcrut 4 and GISS (Figure 1). In fact, the difference between UAH and GISS has increased from 0.34°C/month in 1979 to 0.40°C/month by 2014, which is an 18% increase in difference in total or +0.5%/year. If this trend continues it would mean that GISS will not only diverge more from UAH report monthly GSTAs, but also diverge more from the other data sets. In addition, GISS will also surpass NCDC’ difference with UAH’ data which currently is the largest difference, as the current difference between GISS and NCDC is now only less than 0.05°C.

In summary, all five GSTA datasets analyzed here show an average GSTA over the past 35 years of between 0.01 to 0.42°C above their respective baseline period that varies between each data set. The land-based data sets report in all most all cases monthly GSTA that are higher than the satellite based GSTAs. In addition, there is a general trend towards larger differences between the former and later data-sets over time (since 1979). The GISS data-set has the strongest trend in difference over time and will soon report the largest difference with UAH if this trend continuous, as well as diverge more from the other land-based data-sets. The continuing divergence to the point where the difference is larger than the long term averages between satellite-based and land-based reported GSTAs warrants more in-depth analyses and attention.

References

(1) http://data.giss.nasa.gov/gistemp/tabledata_v3/GLB.Ts+dSST.txt

(2) http://www.ncdc.noaa.gov/monitoring-references/faq/anomalies.php

(3) http://www.metoffice.gov.uk/hadobs/hadcrut4/

(4) http://vortex.nsstc.uah.edu/public/msu/t2lt/tltglhmam_5.6

(5) ftp://ftp.remss.com/msu/monthly_time_series/rss_monthly_msu_amsu_channel_tlt_anomalies_land_and_ocean_v03_3.txt

========================================================

NOTE to readers: see this update from Paul Homewood.

This post was properly criticized for being an incomplete analysis, on a related note, I’m often criticized for being in the employ of “big oil” and flush with cash.

If that were so, I’d be able to hire assistant editors, and simple mistakes like this one wouldn’t have seen the light of day. Such is the issue of a being lone editor on a very demanding blog.

I actually hadn’t intended to publish this post, and had planned to contact Mr. Dohbro today for his reworking of it. I had saved this post to drafts, and as Mosher noted, noticed he didn’t do a baseline alingnment for anomalies. The post was originally a draft,  joining dozens of other posts that I put into the system, but have not published for various reasons; it wasn’t set to publish. So that I can do my work during the day, since I can’t rely on those “big oil” checks, I often schedule posts to auto-publish in advance. Posts loaded into drafts have a collaboration tool that I planned to give Mr. Dohbro access to so that they can be updated and corrected as needed.

Last night I rescheduled and shuffled a bunch of new posts around after the changes I made yesterday to the WUWT format change and the news about UoQ and Shollenberger’s letter challenge required changing the schedule for Monday morning. Somehow, I set this post to auto-publish. I may have simply hit the wrong button, as Steve McIntyre recently did on Mann Misrepresents the EPA – Part 1, or I may have loaded the wrong story and got distracted and simply scheduled the wrong story. I just don’t know. I do know that whatever happened is entirely my fault.

I noted the post was published early today, and considered taking it down then, but thought better of that idea and decided I’d attend to it tonight and set things straight. Paul Homewood beat me to it and I thank him for doing so, and I’m updating this note from work.

My apologies to David Dohbro for publishing his essay without notice (he had no idea it had been published), and an apology to readers for publishing an essay that obviously needed some additional work to be fully accurate. – Anthony