I’m down in Canberra at Senate Estimates this week. Over many years now I have consistently grilled the Bureau of Meteorology at Estimates over their methodology in ‘homogenising’ or changing raw temperature data. These changes include adjustments to make the recorded temperatures colder in earlier years and warmer in more recent years, making the supposed warming trend seem worse.
Transcript
Thanks Senator Green, Senator Roberts.
[Malcolm Roberts]
Thank you Chair. Thank you all for attending today. You got my questions in writing, Dr. Johnson?
[Johnson]
I did Sir.
[Malcolm Roberts]
The first question was please confirm whether you agree that any data adjustments need to be rigorous, independently replicable, and accurately supported with metadata.
[Johnson]
Senator, I might just ask my colleague Dr. Stone to take those questions. Thank you.
[Malcolm Roberts]
Sure.
[Dr Stone]
Yeah, thanks Senator.
[Malcolm Roberts]
Nice to see you again.
[Dr Stone]
Yeah, likewise The Bureau does agree that homogenization adjustments need to be rigorous and homogenization needs to be independently replicable using agreed peer reviewed methods, but it doesn’t actually need, it’s not a requirement that it’s supported by metadata as you’ve suggested. The purpose of the homogenization is to adjust for discontinuities where they’re detected, as I was describing earlier, in comparison with nearest neighbours. Metadata such as documentary evidence of a site move tells you that you might like to check for discontinuity, but it doesn’t tell you to make the adjustments themselves. So adjustment occurs only where a discontinuity in observations crosses what I referred to earlier as a threshold of significance. So it’s possible for homogenization to occur without metadata. And it’s also possible that metadata can describe a situation where there has been a change in observation practise, but homogenization isn’t required because that change doesn’t result in a discontinuity in observation. So it’s actually the discontinuity in observation that’s critical in the determinant of whether or not a homogenization occurs and the scale and direction of the change.
[Malcolm Roberts]
Okay, thank you. Moving on to some we’ve got data on many more which are similar, but the Townsville Weather Station according to the BOM’s metadata said it’s had one move. Whereas in fact, it’s had eight. The Rockhampton Weather Station has had according to BOM, one move. Whereas in fact, it’s had four. The Cairns Weather Station has had according to BOM. two moves the fact that it’s moved six times and the Chaliver Weather Station, BOM says it’s moved twice and it’s at four moves. Why did BOM and the various peer reviewers fail to detect and discuss these inaccuracies?
[Dr Stone]
Sorry, I missed the last part.
[Malcolm Roberts]
Why did the Bureau of Meteorology and the various peer reviewers fail to detect and discuss these inaccuracies in the metadata?
[Dr Stone]
Yeah, no, thank you. They’re not inaccuracies in metadata. So the metadata that either exists or it doesn’t, and in the cases you’ve described, there is instances where a shift in the med station has occurred and there’s not metadata that describes that. So it’s not actually in an inaccuracy in the metadata. And third of what I was saying too earlier, whether or not there’s metadata doesn’t impact on the integrity of their marginalisation process because it’s actually looking for that discontinuity in observations that determines whether or not there is a marginalisation that occurs.
[Malcolm Roberts]
That surely if there’s data about the movement of stations and that data is inaccurate, then the metadata is wrong.
[Dr Stone]
Now, in this case that the metadata is not present. It would be wrong if it said that there was a shift and there wasn’t one. What you’re describing is where a change hasn’t been recorded. So there’s not metadata that relates to it.
[Malcolm Roberts]
Right.
[Dr Stone]
There’s a difference.
[Malcolm Roberts]
So the BOM’s claim that has moved once in Townsville where the station moved once is not accurate because there’s no metadata on them?
[Dr Stone]
Sorry, we don’t claim it’s been moved once but we have metadata that shows that it was moved once.
Yeah. So we wouldn’t claim that there have I haven’t been shifted. We don’t have that data.
[Malcolm Roberts]
So you’ve got metadata for only one move. Whereas in fact, we know it’s been eight moves.
[Dr Stone]
We have metadata for Townsville. I’m sorry, I can’t tell you how many of those that we have metadata for but the principle remains the same. There are instances in the historical record where there’ve been changes made and they weren’t recorded at the time.
[Malcolm Roberts]
So what are the consequences of these areas specifically in terms of recording weather, data such as temperature?
[Dr Stone]
Yeah. I know there is. I just want to be clear about that. So the presence or absence of metadata, doesn’t imply an error.
[Malcolm Roberts]
If the station’s been moved and it hasn’t been noted in the metadata, then it’s not even recognised.
[Dr Stone]
Correct. But if we’re talking about the impact of that on the homogenization process, it doesn’t result in an error because the homogenization only occurs where there’s a statistical discontinuity in the data detected. So you can have moves that don’t result in homogenization being triggered, whether or not there’s metadata and vice versa. So it doesn’t be, I can be crystal clear. It doesn’t result in inaccuracies in the estimation of climate trends. If there’s metadata or not.
[Malcolm Roberts]
Has or not. Has BOM done any analysis to quantify the effects of the station moves especially the ones that it didn’t know about?
[Dr Stone]
No, absolutely. So, as I say, the process of homogenization actually looks back through the records for a given station, looks for discontinuity and measurements compared with nearest neighbours. So it steps through. And does that, so a high proportion of the homogenization changes that are made aren’t triggered by metadata they’re triggered by, as I mentioned, a discontinuity in the observations. And that’s determined by comparing with a large number of nearest neighbours, which we can do with temperature because temperature is reasonably conservative across geographic space. And it’s why, for example, you can’t really homogenize for rainfall because it’s much more spatially viable.
[Chair]
So do you have anymore questions Senator Roberts.
[Malcolm Roberts]
Yes. I’ve got a couple of more questions, Chair. You said you’ve been able to analyse these past records. Could you please provide for each of the four sites that I’ve mentioned that’s Townsville, Rockhampton, Cairns, and Charleville the quantified analysis that Bureau of Meteorology has done and document the independent peer review process used just on notice, please.
[Dr Stone]
That’s all on the website. Yep, no problem.
[Malcolm Roberts]
Thank you. Last pair of questions, Chair. What are the specific quantified consequences of BOM’s inaccuracies on CSIROs use of BOM data? I’m particularly interested because CSIRO has admitted to me that it does no due diligence of its own on temperature data that it merely accepts from the Bureau of Meteorology.
[Dr Stone]
Yeah. Thanks for the question. As I described, the presence or absence of metadata doesn’t result in inaccuracies in the homogenization process. So inaccuracies have not been passed on to CSIRO or any other user because of concerns about metadata. It’s fundamentally a statistical process.
[Malcolm Roberts]
What are the consequences on the government policy and the general assumption that Australia temperatures are increasing?
[Dr Stone]
Yeah. As I said, so if the question is about the accuracy or otherwise of the estimates, presence or absence of metadata isn’t material. And, you know, I can confirm the global trend for warming is around 1.1 degrees Celsius since the pre-industrial period.
[Malcolm Roberts]
And when you say pre-industrial, what year?
[Dr Stone]
1850.
[Malcolm Roberts]
Thank you.
[Dr Stone]
And–
[Malcolm Roberts]
It’s just at the end of the little ice age.
[Dr Stone]
What I’m seeking to do is describe the difference between the global trend, the homogenized trend and the raw observation trends. So the global trend is around 1.1, the unadjusted trend is 0.95 Degrees Celsius plus or minus 0.24 over the same period. And the homogenized trend is 1.44 plus or minus .24 So neither the raw nor the homogenized trend differ from the, significantly from the estimate of the global trend.
[Malcolm Roberts]
Okay, thank you. Are you aware just by coincidence that CSIRO has admitted to me that today’s temperatures are not unprecedented? And then after it admitted that it said that what is unprecedented is the, they claimed is the rate of recent rates of temperature rise. Yet the papers they gave us, not one of them shows that. And two of the papers show that past temperature rise, rate of past temperature rise has been warmer than the recent temperature rise which ended about 1995.
[Dr Stone]
Yeah. Thank you. I haven’t seen the CSRO papers or–
[Malcolm Roberts]
We’ll have to have a chat.
Thank you.
