Or so it would be, if mere statement, restatement and endless repetition by those in authority were sufficient to convince us all of something that our senses and memories do not.

The claim that “1997 has been the warmest year on record” (NOAA News Jan. 8, 1998 Vol. II, No. 2) is the big news of the start of 1998, with the full institutional authority and weight of the British Meteorological Office, the Climatic Research Unit (CRU) of the University of east Anglia, and the U.S. National Oceanographic and Atmospheric Administration (NOAA) behind the claim, a claim which is presented as an accomplished fact, a definitive statement whose finality brooks no discussion or debate.

So confident were they of their authority in such matters that the British Met Office and CRU even announced that verdict on 1997 five weeks before 1997 was even finished, just in time for the start of the Kyoto climate conference in December 1997. Hardly a scientific approach.

What is always ignored in such pronouncements is the fact that the surface record is only one way of measuring global temperature. There are two others - the satellite record and the radio sonde record, both of which are mutually consistent, and which show 1997 to be the 8th coldest of the last 19 years.

So which data source do we believe? The surface record which shows 1997 to be the warmest ever? - or the satellites and sondes which show 1997 to be a colder than average year?

The surface data comprises the combined average of tens of thousands of thermometers world-wide in every country, recording temperatures in standard white louvered boxes called Stevenson Screens, usually mounted one metre above ground.

The boxes are mostly placed where there are suitably trained people to read them, such as at post offices in town/city centres, airports, pilot stations, lighthouses, radio/tv stations, farms, and cattle stations. By far the majority are located in towns and cities.

But can we trust data, whose accuracy is known to be highly variable? (Anita McConnell, Endeavour, v.16, No.2, 1992 provides an excellent discussion about the problems of air temperature measurement). Is it scientifically sound for such problematic data to be subjected to esoteric statistical manipulation to extract real or imaginary trends? Trends which in some cases may really only be an artefact of the statistical program itself?

The Urban Heat Island Effect is the first major source of error, caused by the tendency of concrete, roads, and buildings to heat up to high temperatures in the daytime and slowly release that heat during the night, resulting in a higher daytime and even higher nightime temperature than would exist in a nearby rural area. The effect increases with the size of the urban area, so that as towns undergo their natural growth over time, so too does the measured temperature increase in step with that urban growth. This gives a false impression of long-term warming.

For example, here are rural `greenfields' temperatures for Australia. The post-1910 data used the standard Stevenson Screen.

But if we look only at the six state capital cities of Australia, we get quite a diffferent story -

From the above, it is clear that If the majority of land-based readings are taken in growing towns and cities, we end up with a long-term warming creep in the averaged data, a creep which is impossible to correct since there is no absolute point of reference against which to apply corrections.

Some groups (eg. Jones et al, Nature 347, No.6289, p,169) have attempted to cross-check the general surface network by comparing it with selected `rural’ stations to determine the size of the heat island error, if any. However, the definition of the term `rural’ then becomes critical here as such groups typically regard towns of several thousand people as `rural’. One group even defined `rural’ to mean towns of up to 50,000 people ! (Easterling & al. Science 277, 18/Jul/97 p.364).

But is this sound science? Other research by Dr Robert Balling of the Univ. or Arizona on heat islands has found that the effect even pervades communities of only a few hundred people, let alone 50,000. To be truly `rural’ a site needs to be strictly `greenfields’, where there is no urbanisation whatsoever. Such sites are few and far between, but they do exist, and most of them do not show warming as shown in the Australian study above.

Dr Thomas Karl of the NOAA, the author of the latest NOAA announcement, himself investigated the temperature history of the continental USA back in 1989, using over 6,000 sites and subjecting them to intense quality control against urbanisation errors and found no warming at all (Geophys. Res. Letters, v.16, p.49-52 Jan 1989). The temperature of the continental USA has changed little since then. From his latest announcement, it is clear that he believes the rest of the world has warmed, even if the USA has not.

Since land only occupies 30% of the earth’s surface, the network of white boxes can only measure temperature on land, not the 70% of the planet represented by the oceans. Islands can give some indication of oceanic temperature, but in many cases these are widely scattered and usually located in the largest town on such islands (e.g., Papeete, the capital of Tahiti, shows a post-war warming, but adjacent islands, such as Rapa, do not, so Papeete’s record must be regarded as contaminated by urbanisation).

Even within the land areas, vast areas of desert, tundra and mountains have also not been monitored. Thus we have a problem of `geographical spread’ where one region, such as central England, has been subject to overkill from hundreds of boxes, whereas a vast area like central Australia would be lucky to have just one.

In the face of this unevenness of data, the procedure has been to divide the world into grid squares and to estimate the historical temperatures of each square by averaging the temperatures of all the sites within that square (eg. Jones, J.Clim. v.7, p.1794, 1994). Global temperature is then determined from averaging all the grid squares. Where a grid square only has one site, that site’s temperature becomes the temperature applicable for that entire square.

In some cases, there are no sites within a grid square (e.g.. parts of Australia, vast areas of ocean etc.), in which case the temperature for such a square must be estimated from the temperatures recorded in the nearest neighbouring squares. For example, the lone site at St Helena Island in the South Atlantic Ocean has to represent most of the South Atlantic for sheer lack of sites anywhere else in the region. Any problems with the St Helena site (as once occurred when the box was storm damaged and subsequently moved downhill, only to record warmer temperatures at the new location) results in the local error infecting most of the South Atlantic.

In Australia, the whole of central Australia (almost one third of the continent), is represented by the instrumental record at Alice Springs, an urban site (Torok & Nicholls, Aust. Met. Mag., 1996, 4-Dec-96, p.251), (or is it a rural site? - It only has 40,000 people which puts it squarely within some scientist’s definition of `rural’).

Any site needs to be properly maintained if it is to give a continuously accurate record. This means keeping the box clean and white, keeping the louvers clear, keeping the instruments clean, and avoiding changes to the local micro-environment in the immediate vicinity of the box itself.

Consider cleanliness as an example. The boxes are painted white for a good reason, which is to reflect sunlight so that the box itself does not heat up and give false readings. this means that the box needs an occasional paint job and to be regularly cleaned to maintain the proper degree of whiteness. If this is not done, the box gets progressively dirtier over time. A dirtier box is a warmer box. Lack of simple cleaning and painting will cause the measured temperatures to show a warming creep over time, a creep which might get some impressionable scientist a thousand miles away, looking only at the numbers recorded from such a box, to get all excited about a climatic warming at the site.

Then there’s the louvers. These allow the instruments inside the box to be properly ventilated while being shielded from direct radiation from outside. But such louvers are traps for blown dust and dirt, while spiders find them ideal for cobwebs. The louvers should be cleared regularly to maintain proper air flow to the instruments inside. If this is not done, the box will lose ventilation efficiency and the interior will get warmer over time. Again we have a box-induced warming creep, giving a distant scientist the impression of climatic warming.

Having said all that, how well do poorer countries maintain their boxes? Can they afford the paint, the cleaning etc.?   Do they even care?   USA stations certainly do and so too does Europe and Australia, but that’s barely 6% of the planet. For example, a country like Russia which hardly ever pays its officials these days and cannot afford to render spent nuclear submarines safe, would hardly be expected to put a priority on maintenance of weather boxes. The ragged state of many records from 3rd world countries suggest not only bad maintenance, but also bad personnel training, if any, and a generally low priority on data collection.

No amount of elaborate computing and statistical massaging at CRU or the NOAA can turn bad data into good. If the starting data is bad, all subsequent analyses will inherit those faults.

Where there are boxes, even `greenfields’ ones, we also have people. People typically change the micro-environment to suit their own preferences, such as growing bushes, trees, erecting a shed maybe, or turning a bare piece of ground into a vegetable patch. Most such changes occur over time, such as tree or bush growth, and have two effects. One of them is to reduce the visible skyline of the box (a problem highlighted by radiation expert Doug Hoyt), which reduces the ability of the box to radiate its heat to space, but instead is subject to increased infra-red radiation from the growing obstructions. The shrinking skyline introduces yet another warming creep into the measured temperatures.

The other effect of the shrinking skyline is that the obstructions themselves act as a wind break. This was most evident at Low Head Lighthouse in Tasmania (photo) where a box mounted on a headland in a perfect spot, exposed to the prevailing north-westerly wind, ended up in a mini sun trap caused by nearby bushes growing high enough to screen off the wind. The result was a rocketing daytime temperature, a warming not reflected in neighbouring sites. This effect again imposes a warming creep, even in `greenfields’ sites such as Low Head Lighthouse.

To counterbalance such a wide array of warming creep error factors, there are precious few instances where a cooling creep might occur, so the argument that the errors would somehow cancel themselves out over thousands of records simply does not apply here. The vast majority of such errors are warming ones, and as such they tend to be cumulative over the thousands of records, giving the impression of a climatic warming where none may actually exist.

The current standard practice is to record the daily maximum and minimum temperatures at each site and to average these out over a month and a year to give a mean temperature. The mean temperature for each grid square is then obtained by averaging all the mean temperatures from the individual sites within the grid square (if there are any).

But it was not always so. Prior to adoption of these standards, it was common for data to be collected in a variety of ways, such as measuring the temperature every 6 hours, or twice a day at fixed times, or at other times convenient to the people collecting the data who often had better things to do.

In many cases, particularly at isolated sites, there have even been cases where the people at the site neglected to record the temperature on many occasions but filled in fictitious temperatures in their log books in order not to lose the small stipend they receive for such work. Or the cases of seaside tourist resorts where a tourist operator (who also happened to be the town weather recorder), bumped up the temperature a few degrees in the hope of getting more visitors from the city. Such stories abound in the corridors of weather bureaus, but scientists naively imagine that the people collecting the data are as scrupulous about it as they are.

Falsification of data of all kinds for economic reasons became a way of life in Soviet Russia, so temperature data would have been no exception. One could imagine what a local soviet official in some faraway Siberian village would do during the bad old days of Communist central planning, when he knows that his village’s fuel supply is allocated according to how cold his community is? Of course he will record lower temperatures in his log than his instruments show in order to justify an increased fuel allocation. With the demise of communism, such a life-or-death motive would no longer exist and temperatures would then be recorded correctly. This would give the impression of a post-communist warming - which is exactly what has happened at many Siberian sites, a warming which is not evident outside Russia at similar latitudes such as Alaska and Finland. But the scientists who analyse such data are oblivious to how it was collected. To them, only the numbers themselves matter, regardless of how they were collected, or what political factors may have influenced the collection process.

A further procedural error has been the conversion in the last few decades from manual reading to remote automatic reading of the temperatures in the boxes. During the manual days, opening the door of the box increases ventilation, thus cooling the instruments prior to being actually read. Today, with automatic monitoring, the box door is not opened and so the temperature recorded will be slightly higher, thus giving an impression of warming of perhaps one or two tenths of a degree.

Dr Hugh Ellsaesser of Lawrence Livermore has even suggested that the sharp warming of the 1920s may have been partly caused by changes in the measurement procedures from that of fixed times to one of maximums and minimums only. If thousands of sites world-wide all change their procedures over a period of only a few years, a distinct up or down jump will show up in the long-term aggregate data. There is such a sharp jump during the 1920s at the same time when procedures were changed to the present system, and it was a warming jump of about a quarter of a degree.

But claims today that we are +0.5 deg.C warmer than 100 years ago include that 1920s upward jump, a `warming’ which is in all probability a largely artificial one coincident with the procedural changes.

Over the 70% of the planet represented by the oceans, the only indication of temperature history comes from small islands and ships. In some cases, the air temperature is measured in the usual way from a Stevenson Screen located on an island or a ship, while in others the sea surface temperature (SST) itself is taken as a proxy for the atmospheric temperature overlying the ocean.

This suffers even more from geographical spread problems than the land temperatures. The use of islands to represent vast ocean areas is one problem already mentioned. In addition, the temperature recorded on an island is often from the only town on that island and thus affected by its own urban heat island and other warming creep errors already mentioned.

In the case of ships, the instruments are generally maintained properly and the micro-environment is not subject to much change. However, the ship is constantly travelling so that a temperature taken during the day in one location may be 200 nautical miles from the next temperature taken during the night.

Ships travel on well-established routes so that vast areas of ocean, are simply not traversed by ships at all, and even those that do, may not record weather data on the way. Attempting to compile a `global mean temperature’ from such fragmentary, disorganised, and geographically unbalanced data is more guesswork than science.

As to sea surface temperatures (SST), this data is even more fragmentary than the air temperature readings. Prior to around 1940, SST was collected by throwing buckets over the side of a ship, hoisting it on deck and dipping a thermometer in it.

Get the idea?

Long-term climatic data gathered by buckets is little short of junk.

Bucket data is only useful for immediate weather prediction purposes, not for long-term statistical climatic analysis. Any other data collected in such bizarre ways would be laughed out of any other scientific forum - but not from Greenhouse forums, where such data is tortured with esoteric precision until it confesses to anything you want it to.

In 1989, MIT did an analysis of SST bucket data (brave fellows), but could only find a +0.2 deg.C warming between 1860 and 1940, hardly the stuff of catastrophes (Newell & al, MIT Technology Review, Nov/Dec 1989, p.80).

But post-1940, things seemed to improve.

SST was now measured directly from water intakes beneath the hulls of ships. Of course, the depth of this intake varies with the state of loading and the size of the ship. The ships still travel their same favourite routes well away from regions of the ocean which still lack any data at all, and there are no scientific checks on the accuracy, calibration, or drift of most of the instruments used. So we are a bit better off than with the buckets, but not much.

The 1989 MIT study also analysed the post-1940s data but could find no warming at all.

But finally in the 1980s and 1990s we have satellites to measure SST. Unlike the other satellites which measure atmospheric temperature to an accuracy of 0.01 deg.C, the SST remote sensing is nowhere near as accurate as that. This is because of a little-known effect of water surfaces known as thermal Skin Effect. The immediate surface of the water cools through evaporation and direct contact with the air, causing the top millimetre of water to be several tenths of a degree cooler than water even a few centimetres below.

Unfortunately, satellites sensing in the infra-red can only see the immediate water surface, not the water even a few centimetres beneath. This is because infra-red radiation at these wavebands (around 10 microns) cannot penetrate water at all, and so the satellite can only `see’ that top millimetre.

Of course, we could apply a uniform correction to eliminate skin effect, but that itself raises another complication. As wind speed increases, the skin effect reduces. Once the wind is up to a stiff breeze, the surface turbulence destroys the skin effect altogether, making a uniform correction invalid. The scientists on the ground analysing the SST data, cannot know if the satellite collected its data in calm conditions or in breezy conditions, and thus cannot apply the right correction to take account of skin effect.

For this reason, SSTs taken from satellites are only accurate to within a few tenths of a degree, good enough for immediate meteorological purposes or detecting an El Nino, but not suited to measuring global climatic changes of a few tenths of a degree when the SST satellite data is itself subject to errors of a similar magnitude.

The only way surface data can be used with any confidence is to exclude all town/city and airport data - no exceptions. Only rural sites should be used, and by `rural’ is meant strictly `greenfields’ sites where there is no urbanisation of any kind near the instrument.

This would reduce the available stations to only a tiny fraction of those presently used.

Once `greenfields’ sites have been identified, the station history of each site needs to be examined thoroughly, including old photographs, details of site moves, records of maintenance, procedural changes and a thorough on-site inspection of the micro-environment. Only then can meaningful corrections to data be contemplated. Any such corrections should be independently reviewed by inter-disciplinary scientists, not the questionable `peer review' by fellow specialiasts as used at present. In-house review by fellow `peers' is hardly likely to convince a skeptical public.

Once the sites have been quality tested in this way, it is highly likely that the result would be little or no global warming this century, particularly as the USA itself shows no warming after high-quality control being applied to its surface data.

The satellite data and surface data diverge mostly at low latitudes. I suspect therefore that the sea surface data are in error. My hypothesis is as follows. SST is measured by at least four different methods: by buckets, by intake measurement of cooling water, by buoys, and by island stations (where air temperature acts as a proxy for a vast area of the ocean). Each method measures a different depth of the "surface".

Somehow, these different data streams are merged together. The details are not known to me. But if the proportion of each data stream changes with time, then there is the possibility of introducing an artificial temperature trend.

This matter needs to be investigated in some detail. I suggest the formation of a "Team-B", which is independent of the analysts who have constructed the historic ocean temperature record.

There is one point which I have never seen mentioned; the unreliability of mercury-in-glass thermometers. particularly before 1945.

In the early days, mercury-in-glass thermometers were universally used in meteorology. The British ones, which covered the EMPIRE would have been the most widely distributed, and thus figure most frequently in the old records. They would have been made by Negretti and Zambra, London, and most of them would have been calibrated only in 1°F.

It was insufficiently appreciated at the time that glass is an unstable material which gradually shrinks, thus causing an increase in thermometer readings. Current Standards insist that such thermometers have to be re-calibrated at regular intervals.

The greatest increase in surface temperature this century (what's left of it) took place between 1910 and 1945, a period when it would be very difficult for many Met station personnel to return their thermometers for re-calibration to London, even if this was required, because of the world wars and economic collapse. An error of as large as 0.4°C (0.7°F) is quite likely from this cause alone.

Of course, all this might be concealed, as immediately after the war many of them would have renewed their equipment, made changes which caused a slight rise anyway, but subsequently calibrated regularly, so that the temperature stayed level.

All this is rather speculative, but when you are using an instrument which is calibrated in degrees to confirm a change of only 0.4 degrees the results are bound to be questionable.

The effect of all this on the temperature record is difficult to assess. It could be argued that in the end, truth will out, so that any temporary errors in equipment will eventually be ironed out, unaffecting long term trends. So urbanisation is more important. Coverage is, however, involved. A few wrong measurements in poorly covered areas can make a big difference to averages, and there is sometimes no certainty that an established bias will ever be corrected.

Dr Vincent Gray
New Zealand

Subject: Global Warming
Date: Sat, 18 Apr 1998 10:05:59 -0400
From: "J.WILLIAMS." <100354.554@compuserve.com>
To: "J.Daly" <daly@microtech.com.au>

Sir,

I am a navigator in the Merchant Navy who has been involved in making meteorological observations from British and Australian merchant ships for over 35 years. I am an interested observer of this whole "Global Warming" issue via print and internet pages such as yours.

It is clear to me that nobody could possibly constuct an historical record of oceanic air and sea temperatures to any kind of scientific standard of accuracy from ships' meteorological and other log books.

The standard Met. Office issue stevenson screen (sometimes two) enclosed issue wet and dry bulb thermometers were (and are) hung on bridge wings with more thought given to the convenience of the observer than anything else. It was (and is) known that temperatures should be taken from thermometers hung to windward in clear air. Half a gale of wind from the starboard side with cold rain at 0200 is a powerfull reason not to open the starboard wheelhouse door for anything at all! Also, with a relative wind from astern, the temperature must be affected by heat from the engine room, a very powerful heat source, among others. Some men were (few are) very conscientious in shifting the screen to windward; others were (most are) not. In any event there was and can be no standard position relative to ship structure and heat sources such as is possible in a shore installation.

Even to pretend to read a thermometer to a precision of 0.1 on board ship is asking a very great deal under at least some circumstances. Personally, I would guess that more than 50 per cent of temperature observations as recorded in Met. logs are in error by more than 1 degree.

Wet bulb temperatures I suspect will be worse. The water was commonly taken from the bridge kettle filled from the ships ordinary fresh water tanks. I have spent hours scraping hard white residues from the bulbs of these thermometers when changing the wicks. It was productive of much bad language to find, at observation time, that the water reservoir had dried up. With time pressing, it is problematical whether enough time was left for the newly filled thermometer to reach the proper reading.

Everything on board ship is covered, more or less, with salt deposits. This must include the dry bulb thermometer. Salt is hygroscopic. Ergo, dry bulbs cannot have been completely dry. Does this mean that there could have been errors in dry bulb readings due to evaporation from the thermometers??

Sea water temperatures were taken from samples obtained by means of a Met Office issue rubber "bucket" which looked like a short (about 300mm x 70mm dia.) length of very heavy duty rubber hose with a (I think) wooden plug in the bottom and a metal strop on top to take a length of line. Some had thermometers permanently fitted in a guide set in the bucket so that they couldn't be removed. The breakage rate was considerable and any handy thermometer was used when this happened. Later buckets were not so fitted, the thermometer was removed while taking the water sample.

The bucket was left on deck or hung up where it had last been used. It adjusted to ambient temperature. In sunlight in the tropics it could be bloody hot. It was thrown over the side and allowed to trail astern. How far aft it went depended on the ship's draft and the length of the line but any sample must have been contaminated by heat from boundary layer friction from the ship's passage and heat from the engine room. Where exactly in the water body the sample was taken from is not possible to say. Some men just threw the whole lot over, in which case the bucket sunk below the surface to some extent, while others paid out the line slowly and the bucket just skimmed and bounced along the surface. How long it was left there depended on the man, the weather and how much time pressure the man was under. (Call from Radio Officer :"You'd better hurry up, I'm off watch in five minutes.....")

Sea temperatures now are taken from engine cooling water intakes which is why in modern ships the Met. Office doesn't get sea water temperatures at night since engine rooms are unmanned at night. Engine rooms can be very hot places and the sensors are set in steel pipework at some distance from the ship's side. The draft of the ship changes and the ship rolls and pitches. Temperatures are taken from VDU screens calibrated to whole degrees. How accurate they are and how often they are calibrated I do not know. I suspect not very and not very often since a degree or two error is not of concern from an engineering point of view

All in all, I would not place too much weight on Met. information from ships being of the standard of accuracy that seems to be required.

John Williams.