The automatic adjustment procedure is almost guaranteed to produce spurious, artificial warming, and here’s why.
Guest essay by Bob Dedekind
Auckland, NZ, June 2014
In a recent comment on Lucia’s blog The Blackboard, Zeke Hausfather had this to say about the NCDC temperature adjustments:
“The reason why station values in the distant past end up getting adjusted is due to a choice by NCDC to assume that current values are the “true” values. Each month, as new station data come in, NCDC runs their pairwise homogenization algorithm which looks for non-climatic breakpoints by comparing each station to its surrounding stations. When these breakpoints are detected, they are removed. If a small step change is detected in a 100-year station record in the year 2006, for example, removing that step change will move all the values for that station prior to 2006 up or down by the amount of the breakpoint removed. As long as new data leads to new breakpoint detection, the past station temperatures will be raised or lowered by the size of the breakpoint.”
In other words, an automatic computer algorithm searches for breakpoints, and then automatically adjusts the whole prior record up or down by the amount of the breakpoint.
This is not something new; it’s been around for ages, but something has always troubled me about it. It’s something that should also bother NCDC, but I suspect confirmation bias has prevented them from even looking for errors.
You see, the automatic adjustment procedure is almost guaranteed to produce spurious, artificial warming, and here’s why.
Sheltering
Sheltering occurs at many weather stations around the world. It happens when something (anything) stops or hinders airflow around a recording site. The most common causes are vegetation growth and human-built obstructions, such as buildings. A prime example of this is the Albert Park site in Auckland, New Zealand. Photographs taken in 1905 show a grassy, bare hilltop surrounded by newly-planted flower beds, and at the very top of the hill lies the weather station.
If you take a wander today through Albert Park, you will encounter a completely different vista. The Park itself is covered in large mature trees, and the city of Auckland towers above it on every side. We know from the scientific literature that the wind run measurements here dropped by 50% between 1915 and 1970 (Hessell, 1980). The station history for Albert Park mentions the sheltering problem from 1930 onwards. The site was closed permanently for temperature measurements in 1989.
So what effect does the sheltering have on temperature? According to McAneney et al. (1990), each 1m of shelter growth increases the maximum air temperature by 0.1°C. So for trees 10m high, we can expect a full 1°C increase in maximum air temperature. See Fig 5 from McAneney reproduced below:
It’s interesting to note that the trees in the McAneney study grow to 10m in only 6 years. For this reason weather stations will periodically have vegetation cleared from around them. An example is Kelburn in Wellington, where cut-backs occurred in 1949, 1959 and 1969. What this means is that some sites (not all) will exhibit a saw-tooth temperature history, where temperatures increase slowly due to shelter growth, then drop suddenly when the vegetation is cleared.
So what happens now when the automatic computer algorithm finds the breakpoints at year 10 and 20? It automatically reduces them as follows.
So what have we done? We have introduced a warming trend for this station where none existed.
Now, not every station is going to have sheltering problems, but there will be enough of them to introduce a certain amount of warming. The important point is that there is no countering mechanism – there is no process that will produce slow cooling, followed by sudden warming. Therefore the adjustments will always be only one way – towards more warming.
UHI (Urban Heat Island)
The UHI problem is similar (Zhang et al. 2014). A diagram from Hansen (2001) illustrates this quite well.
In this case the station has moved away from the city centre, out towards a more rural setting. Once again, an automatic algorithm will most likely pick up the breakpoint, and perform the adjustment. There is also no countering mechanism that produces a long-term cooling trend. If even a relatively few stations are affected in this way (say 10%) it will be enough to skew the trend.
References
1. Hansen, J., Ruedy, R., Sato, M., Imhoff, M, Lawrence, W., Easterling, D., Peterson, T. and Karl, T. (2001) A closer look at United States and global surface temperature change. Journal of Geophysical Research, 106, 23 947–23 963.
2. Hessell, J. W. D. (1980) Apparent trends of mean temperature in New Zealand since 1930. New Zealand Journal of Science, 23, 1-9.
3. McAneney K.J., Salinger M.J., Porteus A.S., and Barber R.F. (1990) Modification of an orchard climate with increasing shelter-belt height. Agricultural and Forest Meteorology, 49, 177-189.
4. Lei Zhang, Guo-Yu Ren, Yu-Yu Ren, Ai-Ying Zhang, Zi-Ying Chu, Ya-Qing Zhou (2014) Effect of data homogenization on estimate of temperature trend: a case of Huairou station in Beijing Municipality. Theoretical and Applied Climatology February 2014, Volume 115, Issue 3-4, 365-373
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
Nick Stokes says: June 10, 2014 at 6:04 am
“And as for Auckland, it’s a composite record between Albert Park and the airport at Mangere, which opened in 1966. I don’t know when the record switched, but there is a break at 1966. Before that there is 100 years of Albert Park, with no adjustment at all except right at the beginning, around 1860.”
The break at Albert Park is a downward adjustment of about 0.6°C. (At least it was in v2, I’ll have to check now with v3, unless you can tell me.)
It’s a perfect example of the second (Hansen) type of adjustment error. We have a station with known long-term sheltering problems that were never resolved (no clearing of vegetation) which drove up the temperatures. Even NIWA acknowledged the sheltering problem.
Then in 1966 the whole previous record is adjusted down because of the 0.6°C difference with Mangere. Textbook case. Not only was no trend reduction performed on the station, they made it considerably worse by an incorrect adjustment!
These are the sanity checks that seem never to be performed after automatic adjustments.
Bob Dedekind says: “If you’re suggesting that Hansen-like problems don’t occur, then Williams (2012) disagrees with you, since they postulate exactly that mechanism for why there is a bias:”
Why would I ask you to check whether a saw tooth is a serious problem if I thought that a saw tooth never occurs?
The problem Hansen worried about was homogenization using only information from the station history. Information that leads to jumps (relocations, new instruments, shelters) is better documented as changes that lead to gradual inhomogeneities (less wind or more shade due to growing tress, urbanization or irrigation).
That is why you should not only correct jumps known in metadata, but also perform statistical homogenization to remove the unknown jumps and gradual inhomogeneities. Other fields of science often use absolute homogenization methods (finance and biology), with which you can only remove jumps. In climatology relative homogenization methods are used that also remove trends if the local trend in one station does not fit to the trends in the region. Evan Jones may be able to tell you more and is seen here as a more reliable source and not moderated.
P.S. To all the people that are shocked that the raw data is changed before computing a trend: that is called data processing. Not much science and engineering is done without.
Victor Venema says: June 10, 2014 at 2:36 pm
None of what you say addresses the problem. Whether the jump is known from metadata or found via automatic processing is irrelevant. The problem is that an incorrect adjustment is made that increases the trend artificially. That is what Hansen shows.
And that is what GHCN does to at least one site I know of: Albert Park in Auckland. There will be many others around the globe, and it doesn’t take too many errors like this to skew the trend.
Didn’t it ever worry you that the adjustments always had a net warming effect, across all stations and all long-term timeframes? What if the effect had been a net cooling, would it have gone unchecked?
So when can we expect the ‘sudden change’ shown in Mann’s hockey stick graph to be corrected?
Like I wrote, relative statistical homogenization methods used in climate do not only correct for jumps, but also for gradual non-climatic changes. I hope Evan Jones can convince you.
Didn’t it ever worry you that the adjustments always had a net warming effect, across all stations and all long-term timeframes? What if the effect had been a net cooling, would it have gone unchecked?
No, that is not a reason to automatically worry. If it were, we would worry. It is somewhat strange to assume that scientists are stupid.
There are many reason why there might be a net cooling in the raw data. Many stations started in cities to be used for meteorology, for which less accuracy is sufficient. Now that climatology has become more important, stations are relocated to positions outside of cities. There have also often been relocated from warm cities to relatively cool airports outside the urban heat island in the 1940s. We have more irrigation nowadays. Old measurements were often not as well protected against radiation errors (solar and heat radiation). In the 19th century, the measurements were often performed on a North wall of a non-heated room or a free standing stand in the garden. Quite often the sun did get on the instrument or warm the wall underneath it. Thermometers are nowadays very often mechanically ventilated, in the past they were not.
These cooling effects are mostly understudied, unfortunately, so I cannot quantify them or tell you which effects are most important. Most studies have been on non-climatic effects that would produce a net warming effect. We wanted to be sure that these effects are smaller than the observed temperature increases. These studies are important for the research on the detection of climate change.
Bob Dedekind says: June 10, 2014 at 3:00 pm
“And that is what GHCN does to at least one site I know of: Albert Park in Auckland. There will be many others around the globe, and it doesn’t take too many errors like this to skew the trend.
Didn’t it ever worry you that the adjustments always had a net warming effect, across all stations and all long-term timeframes?”
You have no quantification of the effect of sheltering at Albert Park. And you don’t know that the adjustment for the move was incorrect.
It certainly isn’t true that adjustments always have a warming effect. Someone mentioned Barrow AK at Lucia’s. It turned out that there adjustment greatly reduced the trend.
After reading Mr. Dedekind’s post, as well as NCDC’s paper on the subject, here are a couple of additional thoughts to consider:
1) The NCDC automated homogenization algorithm (AHA) is trying to accomplish a task for which it is fundamentally unsuited. No matter how carefully crafted the algorithm, it is, at it’s most basic level, trying to get the effect of adding information to the raw data record (in order to make it usable), without actually adding the required information. It’s interesting to note that the authors of the paper acknowledge the fact that the “correction” of the raw data used to be a manual process by which the specific circumstances of any given site were taken into account (information added to the record). Since this became too cumbersome, the AHA was created to automate this process. Unfortunately, I believe the creators of the AHA forgot that the whole point of the exercise was to add in any necessary information required to render the raw data record usable, and instead, shifted their focus to statistical analysis, WHICH PRESUPPOSES THAT ALL REQUIRED INFORMATION ALREADY EXISTS WITHIN THE DATA RECORD.
2) On the surface, the claim that “sheltering” has caused a statistically significant number of breakpoints seems anecdotal to me. Is there some data out there to suggest that it’s the prevalent mechanism causing breakpoints? I ask because it seems like there could be any number of external factors that might cause said breakpoints. Then again, maybe it’s irrelevant. If the primary behavior of the breakpoints is “slow up, fast down” and the AHA always addresses this by (in math terms) shifting the Y-Intercept of the older data, rather than adjusting its slope, then it would seem like a pretty easy case to make that the algorithm is inappropriately making a judgement call in the absence of the actual information by which to make it.
3) At a philosophical level, it seems like this incessant effort to “correct” the historical data record is based on the assumption that’s it’s necessary, NOW, to know what the climate trend is so “we can do something.” If you subtract out the belief that “we must do something now” then there’s no reason we can’t just wait for the trending data to be established by these new temperature measuring systems. (Didn’t we just read about this great US system, like, yesterday?)
4) Back to the AHA for moment, notwithstanding my observation above, it’s really difficult not to respect the effort that went into creating the algorithm. Reading through the paper, to me, was a testament to the thoughtfulness of the authors who were diligently trying to make their own proverbial “chicken salad”. Still, the mere fact that such manipulation is required tells me that the whole date set should have a big fat “FOR INFORMATION ONLY” stamped all across it. Meaning, you can review it for curiosity’s-sake, but it’s not valid to use as a basis for engineering purposes (my world) or policy decisions.
Anyway, these are just thoughts…maybe valid, maybe not.
rip
It doesn’t work as science, but it achieves its objective.
June 10, 2014 at 12:32 pm | Willis Eschenbach says:
Here neither … Win 7, Chrome
Nick Stokes says: June 10, 2014 at 3:33 pm
“You have no quantification of the effect of sheltering at Albert Park. And you don’t know that the adjustment for the move was incorrect.”
Actually I do. NIWA compared the Albert Park record to Te Aroha, a suitable rural site, and found a differential of 0.09°C/decade. We know that the wind run decreased from 1915 to at least 1976 (Hessell, 1980). That’s about 0.5°C over sixty years, right there.
We checked this and arrived at the same result. When we performed our Auckland analysis we therefore reduced the Albert Park slope a la Aguilar (2003) and only then did we check the offset wrt Mangere. Needless to say, this accurate manual approach produced a trend way lower than the incorrect GHCN adjustment, which not only didn’t detect and correct the inflated Albert Park trend it actually made the Auckland combined record much worse by introducing an erroneous 0.6°C downwards adjustment on top!
“It certainly isn’t true that adjustments always have a warming effect. Someone mentioned Barrow AK at Lucia’s. It turned out that there adjustment greatly reduced the trend.”Classic strawman. Where did I say that adjustments “always have a warming trend.”? I said there is a net warming trend after adjustments.
A good test would be what happens to the global trend
Before and after breakpoints.
I wonder what you all suppose.
Willis?
Another good test is what happens with synthetic data
Do the changes move you toward the truth or away?
Willis?
More later.
But ask yourself this. If your theory is that breakpoints
Move the global average in a significant way.. what
Will you say if the evidence shows otherwise?
If your theory is that changes or adjustments move
Your estimate away from the truth.. what will say
When tests double blind tests show the opposite.
Hypothetical questions .. perhaps those tests have been
Done.. perhaps not.
Hmm what would feynman say if the tests contradicted a theory of how adjusting works?
More later. Cant text and drive..
Jonathan Abbott says:
June 10, 2014 at 5:15 am
Could anyone post up explicit examples of these types of adjustments in any of the various temperature series?
Right from the birth of GW in the 1986 papers of Jones et al – this html version shows 4 diagrams with steps corrected as discussed.
Full TR027 Southern Hemisphere Book html version
http://www.warwickhughes.com/cru86/tr027/index.htm
scroll down to “STATION HOMOGENEITY ASSESSMENT”
If you go back to –
http://www.warwickhughes.com/cru86/
there are pdf of TR022 for the northern hemisphere
Zeke Hausfather says; June 10, 2014 at 10:39 am
“Its really a question of scale. Changes in climate over time tend to be pretty highly spatially correlated.”
Please provide data, a data analysis, or a creditable research paper that confirms your assertion. And does the correlaton truly depend on data or are they corrupted by assumptions, or a’priori reltionships that guarantee the outcome.
Thanks
Dan
Mosher:
“But ask yourself this. If your theory is that breakpoints
Move the global average in a significant way.. what
Will you say if the evidence shows otherwise?”
I actually don’t care. All I ask is that someone understands and communicates exactly why it happens. This of course also implies that they check the actual stations (or a reasonable subset) before and after adjustments to ensure that sanity prevails.
However, right now it’s patently obvious that a station like Albert Park has been royally screwed up by the adjustments, and nobody noticed. But as usual we are treated to various arguments about why we, and not the automatic adjustment system, are wrong.
The fact is that there are no checks in place to prevent this sort of error occurring all around the world, otherwise Albert Park wouldn’t have happened. It could well be one of the contributors to the +0.3°C/century trend increase after adjustments, and I suggest it is.
Bob Dedekind says: June 10, 2014 at 3:54 pm
“Classic strawman. Where did I say that adjustments “always have a warming trend.”? I said there is a net warming trend after adjustments.”
You said:
“Didn’t it ever worry you that the adjustments always had a net warming effect, across all stations and all long-term timeframes?”
Maybe the comma has a subtle effect.
Well I am with Bob Dedekind on this one, I know its hard work but adjustments to be accurate and useful have to be manual with research. Otherwise your putting some kind of bias in the works and essentially making the entire data series useless for understanding Climate. We really have no clue if there is a warming or cooling trend over who knows how much of the record. When you throw in the error bars it gets worse. Even picking station moves and Doing auto adjusts for height over a land surface using the free air average lapse rate is going to introduce bias. in the wellington case the bias is warm and its documented. From NIWA’s web site
“The offset of 0.8°C used in the webpage illustration agrees with what we would expect for the altitude difference between the sites, based on the free atmosphere lapse rate of 0.65°C per 100 metres. In practice, the adjustment is calculated by comparing overlapping temperature records for other sites across the 1927/28 site change. The altitude effect is an a priori reason for expecting Kelburn to be colder than Thorndon, but there is no straightforward theoretical way of calculating the actual difference along a sloping land surface in the same way as there is for the free atmosphere. In fact, over a much broader spatial scale, the lapse rate along a land surface in New Zealand tends to average around 0.5°C/100m (Norton, 1985). This would equate to Kelburn being colder than Thorndon by 0.6°C; the larger calculated difference of 0.8°C being used in the “seven-station” series therefore suggests other local influences such as exposure or aspect may also be affecting the sites in question.”
By the way the overlapping site change was for 31 days or so, was acutally 1C, but they couldn’t choke that so they went with the free atmosphere lapse rate anyhow. Silliness really, there is a bias there, pretty significant warm or cold it doesn’t matter, Because the purpose of doing this averaging business is to understand what is happening globally and the data is too trashed to do that.
v/r,
David Riser
One more point Mosher. If someone fixes the system, I’ll go first to the adjusted Albert Park to see if it matches common sense, viz: trend reduced, correct offset applied relative to Mangere.
Until then, argue away, you won’t convince me.
Nick,
No, it was you leaving out the vital word “net” that did it, and it wasn’t all that subtle either.
Bob Dedekind says:
June 10, 2014 at 2:06 pm
Thanks, Bob. Indeed. I’ve never seen any computer based decision-making system that doesn’t at times make ridiculously bad decisions when faced with the eternal variability of nature. Which is OK if you check through them afterwords and find the bad ones and figure out how they happened and change your algorithm.
Regards,
w.
David Riser says: June 10, 2014 at 4:30 pm
Yes, and there are plenty of examples where lower stations have colder temperatures than higher ones. Hokitika comes to mind – and inversion effect operated on the town-based site that didn’t affect the airport site.
Willis Eschenbach says: June 10, 2014 at 4:38 pm
“Which is OK if you check through them afterwords and find the bad ones and figure out how they happened and change your algorithm.”
Exactly.
Steven Mosher says:
June 10, 2014 at 4:02 pm
As Holmes remarked, it is a grave error to theorize in advance of the facts …
And in any case, the question is not answerable as stated, because it depends entirely on the level at which you set the thresholds for intervention into the temperatures.
w.
Why are there so many more high max temperatures during the 1930’s (you know, the dust bowl years) than the last 20-30 years? Or have they been adjusted away or deleted?
Is it assumed weather station observers were too stupid to read thermometers?
Bob, what do you think of the claim made by NIWA on its
website that the Seven Station Series is “representative of
New Zealand” … a load of codswallop if you ask me, given
that six of the 7 stations are perimetral. Good for measuring
the effects of sea breezes !
Rex says: June 10, 2014 at 7:15 pm
According to Jim Salinger, there are six climatic regions in NZ, and the seven stations are representative, with one repeat obviously. I don’t know if that’s valid one way or the other, but I know that there are few long-term stations (well, pretty much seven), and I suppose that’s all that really matters in the end.