Steve Milloy
Recall that my 13 million-view tweet sent Big Climate into a paroxysm of desperate but fake ‘fact checks.’ Is NOAA now trying to slowly warm the current pause with an eye toward eliminating it? Or is this just innocent data correction?
So here is the graph that the 13 million-view tweet was based on. The image was taken on January 12, 2023.
Now here is the image taken today.
They are very similar. But note the trend. In January, the 2015-2022 cooling trend was -0.11°C/decade. Today, the trend has been knocked back to -0.07°C/decade.
Is this a mere data correction/adjustment by NOAA or the beginning of something more sinister? Stay tuned.
We now know the explanation for the change. NOAA upgraded NOAAGlobalTemp from version 5.0 to 5.1. 5.1 contains more data and is now performing a full sphere measurement.
https://www.ncei.noaa.gov/products/land-based-station/noaa-global-temp
https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2020GL090873
The implementation of full spatial coverage increased the trend from 1979/01 to 2022/12 from +0.172 C/decade to +0.180 C/decade.
From the abstract of that paper:
“The National Oceanic and Atmospheric Administration (NOAA) maintains an operational analysis for monitoring trends in global surface temperature. Because of limited polar coverage, the analysis does not fully capture the rapid warming in the Arctic over recent decades. Given the impact of coverage biases on trend assessments, we introduce a new analysis that is spatially complete for 1850–2018. The new analysis uses air temperature data in the Arctic Ocean and applies climate reanalysis fields in spatial interpolation. Both the operational analysis and the new analysis show statistically significant warming across the globe and the Arctic for all periods examined. The analyses have comparable global trends, but the new analysis exhibits significantly more warming in the Arctic since 1980 (0.598°C dec−1 vs. 0.478°C dec−1), and its trend falls outside the 95% confidence interval of its operational counterpart. Trend differences primarily result from coverage gaps in the operational analysis.”
v5.0 omitted parts of the Arctic. That has the effect of treating that region as if it behaved as the average of the remaining regions. But in fact the Arctic is warming faster, and if you take account of that (v5.1) global warming trends up a little. It had the same effect as going from HAD4 to HAD5.
Exactly. To provide a global average temperature from a grid mesh with unfilled cells you have to infill no matter what. The no-effort naïve method is to assume they behave like the remaining cells like what 5.0 effectively does. That’s obviously going to have more error than local strategies like what you did with TempLS or what NOAAGlobalTemp v5.1 now does.
BTW…as best I can tell this upgrade happened very recently. In fact, the 5.1 files are not yet available. I had to manually download the data in csv format from the website.
You might also find this interesting. It is a version of NOAAGlobalTemp that uses artificial neural networks to do the infilling.
https://journals.ametsoc.org/view/journals/aies/1/4/AIES-D-22-0032.1.xml
So we now have an explanation for the changes. Perhaps NOAA should have drawn attention to the fact that the basis of its chart had changed. Of course, this article simply noted the changes to recent years rather than the whole record. Doubtless now we have an explanation we can take time to analyse the new supposedly complete record back to 1850, and discuss to what extent it is a reliable improvement, and to what extent it suffers from shortcomings.
Also up for discussion will be kriging methods to cover the Arctic historically, and how they compare with the measurements now being incorporated. To simplify the issue, imagine that there were adequate readings previously at 75 N, well into the Arctic, but leaving everything from there to 90N to be kriged. You might expect that at 75N there would be a strong signal of Arctic warming over time. Is the pole itself suddenly found to have been warming even faster? The quality of the ICOADS and IBAP data now freshly incorporated, particularly historically also merits evaluation.
It is important to note that since NOAA now claim that the record is spacially and temporally complete there are few excuses left for further large adjustments for missing records – the excuse trotted out for much of this thread until persistent questioning forced research into the real explanation for the changes.
“persistent questioning forced research into the real explanation for the changes”
So why is it always bdgwx who has to do the research? Might we not expect the author to make the basic check of whether the two versions are the same? Or at least some other commenter.
In fact the new version both has better treatment of the Arctic and new stations.
So why is it that I had to ask the question so many times, and point out the inconsistencies in his answers with the evidence? Why didn’t NOAA annotate their chart to show that they were using a new dataset? I did the research that showed what the changes were. You didn’t, and neither did he. He did portray himself as the expert on NOAA data upload procedures. I am certainly not that. But I do know how to do some basic analysis of the difference between the datasets and to recognise that the stories we were being told didn’t ring true.
I’m not an expert. And it’s a stretch to even call me an amateur.
Taking the fact on the fly they keep adjusting exist reading of stations because it can’t possible be that cold there without understanding why it so cold there. Topography defies hominization, just because nearby stations are not as cold does not mean the measurement is wrong. Especially when the trees drop their leave two weeks earlier and leaf out two weeks later than said nearby stations. Add in the high ground where those station are on the foot of can get snow when everywhere else get rain.
It’s why Hubbard and Lin found 20 years ago that regional adjustments to station readings simply don’t work. Micro-climate differences are just too significant. Adjustments to station readings must be done on a station-by-station basis. If there is no station you can’t just assume you can plug in an average of nearby stations, i.e. homogenization or in-filling.
No such thing!
Especially as the various governmental agencies perform the corrections.
They keep changing past temperature data. When are they going to stop? Ever?
Here’s my problem. How can I trust the current record if I know it will change tomorrow? Ever-changing data is meaningless. And when something stops being meaningful, I throw it out. What do you do?
Louis Hunt: “When are they going to stop?”
The reason for the change here is because they incorporated more observations and started including the polar areas. The changes are unlikely to stop anytime soon since observations are still being digitized and uploaded into the various repositories and better methods are developed to incorporate those observations.
Louis Hunt: “What do you do?”
When new data points become available I incorporate them into my analysis.
When better analysis techniques (enhancements, bug fixes, etc.) are developed I use them.
What do you do?
In other words, when they stopped infilling and used actual data, the output changed.
Are you still claiming that infilling actually improves the data?
Also, by your admission, the act of infilling increased the slope of the trend.
Do you really expect to get an answer to this?
MarkW, I think there is still confusion. Read [Vose et al. 2021] and familiarize yourself with what changed. I’m happy to discuss it with you, but that is difficult to do when your posts contain statements that are patently false.
the important issue for me is that the raw data is not available & the adjustments are often hidden, disguised or repeated – and again are not publicly available
free the data! we have had significant resistance in Australia against free access to publicly funded data
The raw data is available here and here.
The source code for the adjustments is here.
That’s just the old data.
It’s certainly not realtime, but I wouldn’t call it old data either.