NZT7 uses defective raw dataRichard Treadgold | April 18, 2011
The official New Zealand Temperature Record is made up of historical temperature readings (raw data) and NIWA’s adjustments. Both of those components are unreliable.
The 169-page Report on the Review of NIWA’s “Seven-Station” Temperature Series, or the Review Report (RR), published by NIWA in December 2010 devotes very little space to that bane of climatologists — the urban heat island (UHI) effect. It has been long recognised that air temperature readings taken in towns and cities are affected by the heat absorption of concrete and tarseal surfaces; by exhausts of vehicles, aeroplanes and air-conditioners; and by structures which deflect wind and confine humidity.
Because a “heat island” is not representative of the wider region or country, most climatologists try to give them a wide berth. Wikipedia says that “the temperature difference between urban areas and the surrounding suburban or rural areas can be as much as 10°F”.
A similar enemy of the climate archivist is “shelter” — trees or structures which interfere with the thermometer’s normal exposure to wind or sun, and thereby cause distortions.
The mean temperature impacts of both UHI and shelter are typically gradual, but non-linear. They are hard to detect and almost impossible to correct. Most climate archivists simply omit any sites suspected of being contaminated by UHI/shelter.
Albert Park in Auckland is a prime suspect for contamination by both urban-driven heat and shelter. A glance at NIWA’s table of adjusted warming trends (RR, p 5) shows Auckland at a whopping 1.53°C/century, which is almost double the 0.83°C/century averaged by the other six sites.
The site was officially investigated in 1980 by J.W.D. Hessell, a senior meteorologist with the New Zealand Met Service, and was ‘found guilty’. His peer-reviewed paper, published in a leading journal, noted that:
Visitors to this central park today cannot fail to be impressed by the many large exotic trees, most of which were planted about the turn of the century and some of which, more especially those planted later, are still growing. The instrument enclosure is surrounded on all sides by trees and buildings which shelter the site to a great degree. Sheltering was much less in the years preceding 1940. This fact has been commented on by Finkelstein — who pointed out that in no year since 1930 has the annual wind achieved the mean of those before.
The author went on to consider “urbanisation” as a separate issue:
The urban area population of Auckland increased by 60% between 1936 and 1966, this percentage giving an index of urbanisation there. Quantitative assessments of sheltering and of urban “heat island” effects cannot be satisfactorily resolved unless either or both can be shown to be negligible. The Albert Park data have been treated as comprising two separate periods, 1910–40 and 1941–67 which are compared. The indications are that sheltering and urbanisation have contributed to the apparent temperature increase.
When Salinger was preparing the 7SS (1981), he did not turn his mind to UHI, perhaps because it was a new phenomenon in the 1970s. When he conducted NIWA’s adoption of the 7SS in 1992, he again failed to take account of UHI. He was also a co-author of the NIWA report on climate station histories, Fuohy et al (1992), which extended over 250 pages but nowhere discussed the likelihood of UHI at any of the urban sites.
Hessell’s verdict was only partly accepted by the Review Report, in which NIWA classifies Auckland as “significantly influenced” by urban heating effects:
There is evidence that Albert Park warmed relative to other locations in the area between about 1928 and 1960 due to tree growth in the park and increased urbanisation around the park. The warming due to these effects over the three decades is estimated in the Auckland document at about 0.3°C. Such a warming would increase the apparent 1909–2009 trend at Auckland by 0.38°C.
In comparing the Albert Park record with that of Te Aroha, a hand-picked comparator, NIWA admits to finding evidence of a sheltering (not UHI) effect. There is no contrary evidence.
They obviously considered discarding Auckland, and explain that this would reduce the overall NZ Temperature Record to 0.81°C.
Rhoades & Salinger (1993) say:
Where long-term homogeneous series are required, for example, for studies of climate change, it is best to choose stations that are unlikely to have been affected by gradual changes in shading or urbanisation. This is no easy task. Karl et al (1988) have concluded that urban effects on temperature are detectable even for small towns with a population under 10,000.
Hessell points out that it is worse if both flaws appear together:
In the case of Auckland City the decrease in minimum temperatures due to sheltering (wind reduction) appears to have been masked by the urbanisation (building heat emittance) effect.
But there might be a remedy in some cases:
Climatic temperature trends can only be assessed from rural sites which have suffered no major transformation changes in shelter or urbanisation or from sites for which records have been made homogeneous.
NIWA’s scientists accept that Auckland’s eccentric warming trend alone raises UHI suspicions, but says (RR, p 35) “it is not easy to determine” just how much the urban growth affected the temperature series.
They toyed with the idea of making Albert Park’s records homogeneous by “successively reducing the annual temperatures by 0.01°C more each year than the previous year” (RR, p 6, footnote 1), but did not pursue the idea.
In the end, NIWA elected to take no action at all — and retained the Auckland record, warts and all, subject only to their own adjustments. They state lamely that (RR, p 37) “further research is required to provide more confident bounds on the correction of the early Auckland record for non-climatic warming.”
The sole reason NIWA offer for excluding the period before 1928 is that their chosen comparator station, Te Aroha, has unreliable records up to 1927. This leaves them with the choice of finding another comparator or making the blind assumption that sheltering didn’t begin until Te Aroha was reliable. They opted for the latter course.
In his thesis, Salinger states the view that the Albert Park trees “reached their maximum height in 1930 and it is not expected that they will further affect the exposure.” No evidence was offered and the statement appears to have been a guess — but it offers some evidence that shelter variations during the 1909-28 period were probably at their worst.
And the assumption that contamination ceased in 1960 is equally dubious. It is based on the fact that after 1960 there is no relative trend between Albert Park and two nearby comparators.
One chosen comparator is Riverhead Forest, itself affected by a “forest heat island” effect. NIWA comments (RR, footnote 29) that “forested sites in NZ are often noticeably warmer than non-forested ones.” For obvious reasons, Hessell excludes all “Forest Service stations” which were “planted adjacent to or amongst newly planted exotic forests.” If the trees at Riverhead were growing as fast as those at Albert Park, then the lack of relative trend would demonstrate that both sites were equally contaminated.
The other comparator is Mangere, which was built on the sewage treatment ponds in the 1950s. Apart from being in the midst of a seething stew of chemical reactions, the Mangere area must surely have experienced the highest urbanisation index in New Zealand during the relevant 1950-80 period. The population of Manukau City rocketed 1200% — from 15,700 in 1951 to 190,000 in 1981. If Mangere and Albert Park showed the same warming trend during 1960-76, then the latter was almost certainly affected by UHI.
So all the available indications are that Albert Park was affected by UHI/shelter throughout its 1909-76 life, rather than just the 1928-60 period admitted by NIWA. If shelter alone caused contamination of maximum temperatures of 0.18°C/decade as found by NIWA (RR, p 37), then maxima were 1.21°C higher than they ‘should’ have been over the life of the site. If Hessell is right, and minima are masked by UHI, we cannot know what they ‘should’ have been.
Even if minima were totally unaffected, Albert Park’s raw data needs to be reduced by a very substantial 0.7°C.
The above discussion is confined to Albert Park and assumes its sole contamination problem was shelter, not UHI. No such assumption could be made about the two sites which were established in Manukau City during the period of its greatest expansion. This was also the period that Auckland International airport and the Mangere sewerage plant were established.
NIWA shows that Albert Park was 0.6°C warmer than Mangere at the changeover in 1976. That’s no doubt because Albert Park had accumulated non-climatic warming for 67 years, while it had been happening at Mangere for only 25 years. Significantly, their respective ongoing warming trends during the overlap period were identical.
NIWA did not check Mangere against Te Aroha or any other UHI-free station. If UHI in fact contributed non-climatic mean warming of 0.09°C/decade at Mangere/Auckland Aero, then another 0.3°C needs to be subtracted from the Auckland total.
What effect would an Albert Park correction have on the series? Quite a lot. NIWA says that omitting Auckland would reduce the series trend by more than 0.1°C. But it doesn’t stop there. The trend at Albert Park was used as a comparator to increase the warming trend at:
Masterton in 1942 (RR, p 50) and 1920 (p 52)
Wellington in 1928 (p 67) and 1912 (p 70)
Nelson in 1932 (p 90) and 1920 (p 92)
Hokitika in 1928 and 1912 (p 114)
Dunedin in 1912 (p 162)
So all of that ‘derivative’ warming also needs to be undone.
NIWA’s Schedule of Adjustments for the 7SS, tabled in Parliament on 19 February 2010, suggests that a key reason for moving the Auckland weather station from Albert Park to a less built-up site (Mangere) was because there were indications of urban warming.
So when it came to merging the records of the Albert Park and Mangere sites, NIWA knew that the late readings of the former were likely contaminated and unreliable. But they simply ignored this fact. In Figure 3 (RR, p 18), they compare the shelter-inflated readings from Albert Park (without adjustment) with the readings from Mangere. It’s no surprise that they find a base level difference of about 0.7°C — the prima facie UHI/shelter figure suggested above.
If proper allowance was made for UHI/sheltering, the base level temperatures of the two stations would be very similar. No significant adjustments would be necessitated by the site change.
Ignoring UHI, NIWA subtracts 0.65°C from the Albert Park record from March 1976 all the way back to commencement of the series in September 1909. This one adjustment changes the raw data by a time-weighted aggregate of 41.67°C and accounts for nearly half the warming shown by all the adjustments in the NZT7 (see Schedule of Adjustments).
The Role of the BoM
The 2010 Review is the first time NIWA scientists have acknowledged any detectable non-climate influences in the temperature trend of any of the seven stations. It is also the first time their efforts have been open to influence by any outsider.
The Australian Bureau of Meteorology (BoM), which NIWA retained as a consultant for the Review, has long been conscious of the potential depredations of UHI on long temperature series. As NIWA’s trans-Tasman counterpart, it maintains the Australian Temperature Record, based on “high quality data sets (HQDS)” of about 100 stations — which exclude all data from the major cities.
During the consultancy, on 13 October 2010, the BoM issued a media release stating that they always knew city night-time temperatures (i.e., minima) were affected by UHI and added: “We can now confidently say that the reason our cities are warmer, and warming faster, than the surrounding countryside during the day is (also) urbanisation.”
The BoM research team analysed data from 70 sites from towns with populations of 500 to 100,000, as well as a handful from large cities.
The results were described at a meeting of the American Meteorological Society on 26 January 2011. It found that daytime (maximum) temperatures in Sydney and Melbourne were increasing at rates of 0.08°C and 0.07°C per decade faster than the average of the 73 sites.
BoM offers no solutions or corrections for UHI-affected sites. It just steers clear of them.
What should have been done?
As NIWA says above, it’s not easy to determine how much historical records were affected by UHI/shelter at any time. So it is not easy to effect a cure.
Hessell divided all stations with continuous records (and no major site changes) during 1930-79 into two categories. Those which seemed to have shelter/UHI problems were “A” stations, whilst the rest were “B”. He found that the average increase in mean temperatures at “A” stations was about five times that of the “B” category. The “B” stations showed no overall warming trend at all.
He paired nine urban stations with adjacent rural ones, and found all the former to be significantly warmer. After using further statistical measures, he concludes firmly:
It is thus considered that urbanisation is a significant factor in influencing the temperature regime of a district, and that this is an important contributor to temperature trends in growing built-up areas.
The usual response is to simply by-pass UHI-affected sites and focus on rural readings. But there were no rural stations in 1909. The 7SS include the four largest cities, while Masterton, Hokitika and Nelson are also substantial urban areas. All were growing vigorously during the early-mid twentieth century.
Perhaps Lincoln might qualify as rural. But, ironically, it was cursed by repetitive shelter belt problems during 1910-43.
Hessell was adamant that “the factors which have affected the reported temperatures at Auckland are common to many sites in New Zealand” and indicted Christchurch Gardens and Kelburn in particular. He regarded Dunedin as so unreliable that he did not even test it.
At the commencement of its Review, NIWA ought to have transparently tested all seven stations, at decadal intervals, for indications of gradual non-climatic temperature change. If NIWA or BoM had any disagreement with the tests or findings of the long-respected Hessell paper, they ought to have challenged them with a peer-reviewed journal paper of their own.
Where UHI/shelter problems were identified, NIWA should have made an earnest attempt to identify and correct for the impacts. Where this could not be done with 95% confidence, the defective site ought to have been dropped from the series.