Up down all around it ends where it begins

A Mobius strip

Most people take for granted that New Zealand’s surface air temperatures (SAT) have gone up over the last century. But is that true?

NIWA’s official graph of the national temperatures is well known. You can get it from their web site. By the way, if you go there, let me know if you think they’ve recently stretched the image laterally; it is definitely wider than before — even the text is obviously stretched sideways. Could they be trying to make the vertical change less prominent and thus reduce the apparent slope of the trend line? What goes through their minds at NIWA? Anyway, as copied a few months ago, and so a little different, it looks like this:

NIWA's official NZ temperature graph

You should be aware that the Climate Conversation Group and the NZCSC have delivered a rip-snorting criticism of this graph, so we don’t agree with it even for a moment. Now I will raise further objections to NIWA’s graph because it contains features inconsistent even with NIWA’s own conclusions.

The first thing to check is the country’s average temperature in the 1860s. There is good material on this in the National Library. It turns out that in 1867 the New Zealand Institute recorded what was happening with the temperature. How convenient.

The new Dominion had already been recording temperatures in the main centres for about 15 years. So by 1867 the Inspector of Meteorological Stations could report that the average mean temperature to 1867 was 55.6°F — or 13.1°C. That was measured over about 15 years — giving a fair picture of the climate.

So how does it compare with today’s temperature? We shouldn’t take just a single year, for by carefully choosing the start and end points you will “discover” warming or cooling at will; it’s called cherry-picking and scientists frown on it but more importantly, you don’t discover the truth. Still, let’s take a brief look at recent annual temperatures for the sake of a human comparison: what did those pioneers actually experience?

NIWA’s web site has Climate Summaries for recent years giving various climatic information. For 2005, the national average temperature was 13.1°C. Last year, 2009, it was 12.3°C, nearly a degree lower than 2005. So in the 1860s it was 13.1°C, then, in 2005, after 138 years, the temperature was just the same. Last year it was slightly below 1867, at 12.3°C.

So our forefathers experienced the same temperatures we’ve been having recently!

Now compare the recent 30-year average of 12.5°C with the average from the 1860s of 13.1°C. It’s not very different: the temperature has moved down by only 0.6°C. That’s outside the likely margin of error (probably around 2°C) which means there’s no real change.

So those official records, pioneering and modern, show no warming, but actually a slight cooling.

But wait, there’s more. Look at NIWA’s official graph. Observe the decline into the first decade of the 20th Century, followed by a rise to about 1950. Finally, observe there is no significant trend since 1950. This graph highlights the period I’m talking about:

NIWA's official NZ temperature graph

For the last 60 years the temperature shows no trend, so all of the rise — not just some of it, but all of it — claimed by NIWA, about 1°C, occurred before 1950.

It doesn’t look as though carbon dioxide dominates the temperature of the world’s climate, since while most of the CO2 was being emitted the temperatures didn’t go up. Because New Zealand sits in the world’s greatest ocean it’s a good proxy for global warming; the fact our temperatures have been pretty stable for 60 years says a lot about the globe’s temperatures.

Remember, I disagree with the facts this graph purports to show, but it’s still the official NIWA graph. Assuming it’s correct, it contradicts what NIWA themselves conclude. They say New Zealand (with the world) has been warming under the influence of mankind’s emissions of carbon dioxide. NIWA manufactured uses this graph in support of this contention — in fact, it’s the only evidence they offer.

Yet it shows clearly that there has been no warming in New Zealand for the last 60 years.

This is laughable. What more evidence is needed to demonstrate NIWA’s incompetence?

It’s a disgrace.

Leave a Reply

10 Comment threads
23 Thread replies
Most reacted comment
Hottest comment thread
8 Comment authors
Notify of
Mike J

“What more evidence is needed to demonstrate NIWA’s incompetence?”

It would be more convincing with empirical data supporting your claim of no warming for 60 years.

Mike, I’ve used NIWA’s own graph, from their web site. For the actual data, go to NIWA’s web site and download the Excel spreadsheet linked there.

Mike J

You want me to download data to strengthen your point? My comment was that if you show data which (rather than a fuzzy line on a small graph) in your article, it would make your case more convincing (imho).

Ah, I misunderstood, sorry. I wasn’t intending to strengthen my point, but to provide you with the empirical data you asked for.

But the graph is compelling by itself. Click on it to see a larger version, by the way.

I don’t know what “fuzzy line” you’re referring to. The area of interest is highlighted in yellow; the individual data points show no marked trend. The curving solid black line is NIWA’s “smoothed time series”. It shows an insignificant rise between about 1950 and 2008.

If you don’t believe the graph or find it hard to understand, complain to NIWA. They produced it; they concocted the figures and they drew the graph.

Mike J

I find it difficult to gauge a trendline in a bar graph with such wide variation, and with trendlines and smoothed curves superimposed. You are asking the reader to believe your interpretation of this graph without providing the data and maths to prove your assertion.

On your advice, therefore, I downloaded data from the NIWA site and selected 2 random stations: Eyrewell Forest and Reefton
1961 10.7 11
1962 11.3 11.7
1963 10.4
1964 11.2 10.6
1965 10.6 10.3
1966 10.9 10.9
1967 10.8
1968 11 10.6
1969 11.2 11
1970 11.8 11.7
1971 11.9 12.1
1972 11.2
1973 11.8 11.5
1974 11.1 12
1975 11.2 11.4
1976 9.8 10.6
1977 10.3 10.4
1978 11.3 11.5
1979 11
1980 10.7
1981 11.5
1982 11.1 10.3
1983 11 10.6
1985 11.3 11.6
1986 11
1988 11.8
2000 11.6
2001 11.6
2002 11.4
2003 11.2
2004 10.6
2005 12.1
2006 11.3
2007 11.7
2008 11.4
2009 11.1

Which produced a graph showing trends of 0.2 and 0.4 degrees centigrade increase from 1961 to 2010. The job of harvesting the datasets for the entire country is one I can’t spare the time for just now. The graph I ended up with can be viewed here:


That’s industrious, and admirable of you to take such an interest!

However, the article is about NIWA’s graph and their contention that it shows warming. You must examine the same data or you’re missing the point. If I was more skilful, I would do it myself, but what is needed is to combine the data from the seven stations into a single graph, which is what NIWA have done. So we have their graph, but not the combined data.

Also, the station notes must be examined to determine whether adjustments are required. Which is the whole point of our battle with NIWA, trying to get them to tell us what adjustments they’ve made and why, exactly, they made them!

But your interest is invaluable. I hope it spreads.

You’ll see, if you take your enquiry just a little further, that the raw data for each of the seven stations present a very different picture from the one NIWA paints in its iconic graph!

Mike J

ahh – found the Composite data. Still shows over a 0.3 deg. C increase. http://a-plus.co.nz/pub/2010-04-26_NIWA-7stn-comp.jpg

Mike J

“You must examine the same data or you’re missing the point. If I was more skilful, I would do it myself ”

Have created a bar-graph similar to the NIWA style using the empirical data as supplied by them, just to make comparison easier. http://a-plus.co.nz/pub/2010-04-26_NIWA-&stn-comp-anom.jpg

So, there is in fact an increase in their graph over the last 60 years, of around 0.3 degrees.

Could you let us have a link to that composite data, please?

Thanks for the graph! So the warming would be the equivalent of about 0.6°C per 100 years. That’s not unprecedented, is it?


Aha! So it stopped warming when greenhouse gases started to build up in the 1960s. So, if we can keep increasing emissions, we should be able to move to a cooling phase and offset that pre-1950 warming.

But wait a minute – how can the temperature start from 13.1°C, then move up 1°C, but still be only 13.1°C? Has NIWA explained this? Can anybody explain this?

Well, to a degree, it’s explainable. (Couldn’t resist.)

Firstly, note that the one-degree rise was not measured from the 1860s, but from 1905. By then the temperature had gone down, as you can see in the graph, over half a degree, perhaps even 0.7°C. So some warming took place early in the century.

But we’d like to hear NIWA’s explanation and we think it’s important for the country to hear it, too.

See, we think the graph is the product of shonky science. NIWA seem to agree, because they’ve junked it and they’re starting again. For $70,000 and about seven months’ work, they’re constructing a brand-new temperature history. It’ll be peer-reviewed and everything! We can’t wait. This demonstrates NIWA’s loss of faith, because if they believed in Salinger’s national temperature graph they wouldn’t reconstruct it. Update it, yes, but they’re doing much more than that.

If you’ve read our paper, Are We Feeling Warmer Yet? you’ll know that the early temperatures were adjusted downwards to give the impression of warming but the raw temperature readings have no trend.

In other words, the rise claimed by NIWA is false and cannot be explained by events in the real world. And that’s our explanation. To a degree.

Mike J

If NIWA used a polynomial trend rather than linear their graph would appear much more natural and less alarming. I wonder why they insist on a linear regression with an R-squared value of only 0.0545 (ideally the R-squared value is 1.0 – the closer it is to this, the better fit the trend line).
The polynomial in this link has an R-squared value of 0.3171:

You obviously know your way around graphs, Mike.

The polynomial version is a nice graph. I think it demonstrates little warming over the last 60 years, though I should talk to a scientist friend before I comment.

I mean, if the polynomial was applied only to the last 60 years, the amount of warming would be smaller than it appears here. I think.

But we’ve got a problem. Couple of problems. First, the trend you report is much smaller than what David claims. Second, NIWA claim warming “over 1909 to 2008 [of] 0.92°C/100 years”.

I’ve asked David what data he used. Are you using the “adjusted” data from NIWA’s “seven stations”?

Be aware that NIWA’s adjustments are unknown. They have declared the magnitude of the adjustments, but, apart from Hokitika, still refuse to disclose why they were made, so it is impossible to assess their quality. Peer review is impossible.

The raw readings only show warming after the adjustments. Did you know that?

I haven’t heard enough here yet to convince me that the national temperature graph should stay on NIWA’s web site. It shoud be withdrawn until we’ve got a proper scientific one.

But, go ahead and impress us!

Mike J

My source from NIWA:


“This spreadsheet contains the adjusted data used to compile the ‘seven-station’ series. In this file, you will find:
Annual average temperature for each location, and for the 7 sites combined
Annual anomaly (difference from the 1971-2000 average) for each location, and for the 7 sites combined”

I am using the adjusted data because that is the basis of the graph to which your article refers. You contend, “For the last 60 years the temperature shows no trend, so all of the rise — not just some of it, but all of it — claimed by NIWA, about 1°C, occurred before 1950.” The graph in fact shows a trend of 0.3 degrees warming over the 60 years.

I think that if you are to make such a strong claim (that NIWA is incompetent) and not appear to be incompetent yourself, you should examine the facts more closely and provide empirical data to support your claims (which was my opening comment on this weblog).

You guys just get funnier and funnier. Taking the NIWA data from 1949-2008 (your 60 year period) the trend is 0.07C per decade. That trend is significantly different than 0 (p=0.01).

The point of the linear regression isn’t to smooth the data (that have a smoothed line after all) it’s to see if there is an underlying trend in the noisy data. You can always getting a higher r2 by using a polynomial fit, since it’s adjusting a linear model to fit the data points, but if just wanted a better fit then other methods (loess, wavelets…) that would a better job.

Oh, I should have said it’s the cranks that get funnier and funnier,

Mike’s questions are, by comparison, perfectly reasonable.

I consider the use of “cranks” insulting. Don’t spoil the conversation, David. We’re trying to discover the truth here.

Or please describe what you find humorous.

I’m surprised that someone who could be part of something as breathlessly abusive as the “are we feeling warmer yet” press release could blanch and something as mild as being called a crank.

A crank is someone who’s position doesn’t change after they’ve been shown to be wrong. So here’s your chance to not be a crank – a link to a spreasheet with the NIWA combined data from their website clearly showing a warming trend over the time you claimed there was not warming.

OK, nice job, David. I said: “Yet it shows clearly that there has been no warming in New Zealand for the last 60 years.”

Sorry, I should have qualified that statement to read something like “no significant warming”. No matter. What’s the margin of error in these data, you reckon? Some people estimate about 2°C. Which means the warming trend you report is insignificant. It could even be cooling..

How much warming do you reckon occurs in NIWA’s official graph between 1909 and 1949? How does that balance with the warming you report from 1949 to 2008?

Sorry, I should have qualified that statement to read something like “no significant warming”. No matter. What’s the margin of error in these data, you reckon?

Google docs doesn’t do it, but R does, the trend is 0.007 per year, the std. error is 0.003, which is, as I said earlier significantly different than no trend (p = 0.011)

How much warming do you reckon occurs in NIWA’s official graph between 1909 and 1949? How does that balance with the warming you report from 1949 to 2008?

I don’t know. I don’t understand the second bit.

Funnier? How?

By “NIWA data” what do you mean? Do you have a link? The trend you obtain is significantly higher than the 0.3°C reported by Mike over the same period.

Mike J

@ DavidW Firstly, your use of the word ‘cranks’ is uncalled for and typical of AGW alarmists. Secondly, your definition of the word, once called on it by Richard, is quite erroneous and supports the suspicion that you are simply stirring a pot empty of substance. When you define cranks as “someone who’s position doesn’t change”, this is in fact a definition given by a crank which, according to Microsoft Word Thesaurus is “oddball, non-conformist and eccentric”. So please desist from ad hominem jibes in the interests of civil debate. Back to the empty pot. You state, “The point of the linear regression isn’t to smooth the data…”. I never said it was. You also state, “but if just wanted a better fit then other methods (loess, wavelets…) that would a better job.” Yes – a better fit would be nice. Precisely my point. How about you produce a graph using your mentioned techniques and make a point of your own, rather than stirring that empty pot? My point, in experimenting with a polynomial, was to explore something other than a linear trend which in my opinion erroneously reinforcing a notion that we… Read more »


You might be amazed to learn that Microsoft Word is not the sole aribiter of a word’s definition.

The rest of you post is a bit confused. The linear trend in the NIWA graph isn’t aimed at getting a best fit and it certainly isn’t about projecting future change (that’s what climate models are for). It’s just _testing_ if their is an underlying signal in the noisey data. As you fit more complex models to the data you are guaranteed to explain more of the variance in the data, but you also run the risk of confusing noise with signal (in fact, if you wanted the “best fit” you’d just join up all the dots rsq = 1!)

Mike J

Sorry David – your strawman rubbish is not even worthy of a reply. Thanks for coming, though.


Name calling is usually a sign of desperation. Richard is right, unless we know why adjustments are (were) made it is impossible to verify anything, comparing raw data from a station that was moved 3 km up the road with the raw data prior to the shift is also inaccurate. Here is my experience. Up to 4 years ago we lived in Torbay, Auckland. The winters were very mild and the summers never too hot. Richard will know what I mean as he lives in a similar position nearby. Then we shifted about 15 km as the crow flies north west inland. We went from an elevation of about 20 mtr and 300 mtr from the sea to an elevation of 94 mtr and about 8 km from the sea. The effect on temperature? Much colder winters and warmer, drier summers. So if one wanted to compare seasons, with the two data sets, in the new location we would have to adjust up in winter and down in summer. Not just up or just down. Then you have the problem of how much do you adjust, we clearly can’t use the seemingly accepted… Read more »

Mike J

If a weather station has to be moved then an entirely new record should be started for the new location. Otherwise the new location has to be ‘adjusted’ forever which is absolutely ridiculous, and gives rise to a never ending debate on methods and maths used.


But this is what appears to be happening. The station is moved and a calculation is made what the possible temp difference can/should be taking a number of factors into account and this is then applied to compare with the previous station data. Seemingly little effort is made to verify the adjustment as applicable for the situation, which you can’t unless you run the stations at the same time for a lengthy period.

You should be aware that when adjustments are called for, it’s the “closed” or “to be closed” station that gets adjusted. The “open” or current station is always considered to give an accurate reading at the moment. It’s the history that gets adjusted.

The margins for error in these temperature datasets are quite large, about 2 degrees. It’s not possible to measure accurately to less than a degree, especially for decades at a time. So trends of fractions of a degree are immeasurable, undetectable by our senses and give an entirely false impression of reality. We can’t even know their sign. It is meaningless to reach conclusions based on them.

So I guess that my fundamental point, that there’s been no (significant) warming since about 1950, stands. What do you think?



This is what I’m talking about with you inability to update your ideas in the face of evidence. You’ve been shown that there is significant warming since 1950, but instead of accepting this and moving on you introduce an unsourced claim, that temperature can’t be to be less than a degree. This will come as new to Watts, whose report on the US temp record reports accuracy down to about 0.06 C.

David, You say This is what I’m talking about with you inability to update your ideas in the face of evidence. You’ve been shown that there is significant warming since 1950, You seem to be claiming 0.42°C for the 60-year period from 1949. Have I got that right? What is your source for that? Incredibly, you seem to want to disbar me from questioning what you say: but instead of accepting this and moving on you introduce an unsourced claim, that temperature can’t be to be less than a degree. The last phrase is incomprehensible; you might wish to restate it. But you do admonish me for not citing a source: my source is several practising climate scientists in private communications. I’m sorry I don’t have a reference but I will request one. Then you vaguely cite Anthony Watts (I presume): This will come as new to Watts, whose report on the US temp record reports accuracy down to about 0.06 C. If you’re trying to imply that Anthony is generally satisfied with the accuracy of temperature records, you are poorly informed. He knows, possibly better than anyone on the planet, the inaccuracies… Read more »


Ultimately it does not matter which station gets adjusted as long as it is constant and relevant. That is where the problem sits: it is very hard to determine the constant adjustment and you certainly add a lot more guesswork and margin of error by just closing one and opening at another site, then to reach for the calculator and say “it should be colder on the new site so lets deduct 0.5 degrees of all readings of the old station based on ?????? (some theory that generally means an increase or decrease for certain conditions but unproven for that site)” to get it “in line” with the supposedly cooler new site (or warmer whatever the case may be). You are right, a margin of error of 2 degr is large where we talk about (depending who we talk with) a supposed increase of 0.5 to 0.7 degr over the last 60 years. And with the sites being monitored largely affected by the urban heat island scenario and therefore skewing the total averages upwards it should not be unreasonable to think that there has been no change whatsoever other then annual weather fluctuations.… Read more »


Now – Im just an average Joe Public with only a rudimentary understanding of graphs, But graphs are made to show relationships and trends (any of you plotted the All Blacks trendline for world cup results ? base on that, they may as well not turn up at the next cup round…) Anyway – ASSUMING the NIWA graph is based on correct figures (and as we all know, many graphs – eg hockey stick – are based on adjusted and/or biased figures), then its pretty obvious that since about 1850 , NZ cooled thru to about 1900; then from 1900 thru to about 1955 it warmed (by about as much as it cooled from 1850 thru to 1900) ; and since 1955 things have pretty much flatlined (in fact probaly cooled). Now one really has to put aside all the upmarket methods of calculating figures and constructing graphs and trend lines- they’re not the important aspect. If the NIWA graph started in 1955. it would show no changes (certainly no changes beyond margin of error. By the way – I read a study about how accurately people read mercury thermometers – bloody hopeless… Read more »

Mike J

Good logic Barry, and well-said.

Post Navigation