Climate Etc.

Unknown and uncertain sea surface temperatures

by Tony Brown

Over the years I have become intrigued at the manner in which historic surface temperature records- that were never intended to be more than a broad reflection of the micro climate around them- have been used as if they were highly accurate scientific data and subsequently used to inform policy. I have written two articles about their historic accuracy, both of which can be accessed through this link.

The oceanic equivalent-sea surface temperatures (SST’s) have even more dubious origins than land temperatures, but are also perceived as a scientific record that accurately informs our global understanding of SST’s back to the middle decades of the 19th Century. In some ways they are even more of a key measure than land temperatures as the ocean constitutes 70% of our globe.

In this article we tackle the methods used to calculate SST’s-most famously by Hadley/Met office who compiled the graph of SST anomalies from 1850 shown here. Methodology is described by the Met Office Hadley Centre here:

The SST data are taken from the International Comprehensive Ocean-Atmosphere Data Set, ICOADS, from 1850 to 1997 and from the NCEP-GTS from 1998 to the present. HadSST2 is produced by taking in-situ measurements of SST from ships and buoys, rejecting measurements which fail quality checks, converting the measurements to anomalies by subtracting climatological values from the measurements, and calculating a robust average of the resulting anomalies on a 5° by 5° degree monthly grid. After gridding the anomalies, bias corrections are applied to remove spurious trends caused by changes in SST measuring practices before 1942. The uncertainties due to under-sampling have been calculated for the gridded monthly data as have the uncertainties on the bias corrections following the procedures described in the paper.”

The Wikipedia defines sea surface temperature here:

“Sea surface temperature (SST) is the water temperature close to the oceans surface. The exact meaning of surface varies according to the measurement method used, but it is between 1 millimetre (0.04 in) and 20 metres (70 ft) below the sea  surface.”

However, the complexities of defining sea surface temperature are elaborated upon here:

SST is a difficult parameter to define exactly because the upper ocean (~10 m) has a complex and variable vertical temperature structure that is related to ocean turbulence and the air-sea fluxes of heat, moisture and momentum. Definitions of SST provide a necessary theoretical framework that can be used to understand the information content and relationships between measurements of SST made by different satellite and in situ instruments. 

To ascertain how the basic SST data is physically collected, refined, and subsequently used as the basis for information utilised by Governments all over the world, and in conjunction with CRU as the definitive record of global land/sea temperatures, it is worth starting out on our voyage of discovery by reading the Wikipedia article referenced above in full, which continues;

“There are a variety of techniques for measuring this parameter that can potentially yield different results because different things are actually being measured. Away from the immediate sea surface, general temperature measurements are accompanied by a reference to the specific depth of measurement. This is because of significant differences encountered between measurements made at different depths, especially during the daytime when low wind speed and high sunshine conditions may lead to the formation of a warm layer at the ocean’s surface and strong vertical temperature gradients (a diurnal thermocline). Sea surface temperature measurements are confined to the top portion of the ocean, known as the near-surface layer.”

The methodology for collecting temperatures from the sea historically covers measurements from ships (buckets and engine intakes), buoys floating on the ocean surface, weather ships, and in recent years the employment of an array of diving ‘Argo’ buoys. Satellite measurements commencing around the 1970’s are outside the scope of this article. All methods measure at different depths, from the top ‘skin’ of the ocean at 0.01 mm through to hundreds of metres below the surface, and therefore one method cannot be easily compared to another, as temperature is likely to vary considerably according to depth.

“The temperature of the world’s ocean is highly variable over the surface of the ocean, ranging from less than 0°C (32°F) near the poles to more than 29°C (84°F) in the tropics. It is heated from the surface downward by sunlight, but at depth most of the ocean is very cold. Seventy-five percent of the water in the ocean falls within the temperature range of −1 to +6°C (30 to 43°F) and the salinity range of 34 to 35. “ (reference)

The earliest measurements of the world’s oceans, rather than coastal waters, were taken from ships, with an early example of this being when Benjamin Franklin suspended a mercury thermometer from a vessel while travelling between the USA and Europe when conducting his survey of the Gulf Stream in the late eighteenth century. Temperatures can however be relatively consistent if there is a current-such as the Gulf Stream- operating throughout the varying levels, as exemplified in the examples given below. These cover the first reliable reference to systematic and detailed ocean temperature sampling for scientific purposes that the author is aware of. They come from around the 1820’s, when expeditions were mounted by the Royal Society to investigate the ‘unprecedented’ melting of the Arctic.

“An observation which it is interesting to mention here, and which gives a proof of the very little difference between the temperature of the surface and that at some depth, is mentioned in the Voyage of Captain Graah, p. 21. He says,” The 5th of May, 1828, in lat. 57° 35′ N., and 36° 36′ w., Gr., the temperature of the surface was found 6°-3 (46°-2 Fahr.), and at a depth of 660 feet 5°5 + K. (44°-5Fahr.).” This proves that there is no cold submarine current in the place alluded to the S.E.. of Cape Farewell. A still more conclusive experiment is recorded by Sir Edward Parry in the account of his first voyage, June 13, 1819 : in lat. 57° 51′ n., long. 41° 5′, with a very slight southerly current, the surface temperature was 40J° Fahr. ; and at 235 fathoms 39°, a difference of only 1J°.” (reference).

Any measurement of the ocean’s surface, or deeper sub surface, prior to the middle of the 19th century is generally considered of relatively limited scientific value as there was little consistency in the data collection.

In 1853 Lieutenant MF Maury helped organise the Brussels international Maritime conference, whereby all participating countries agreed to adopt common methods to monitor meteorological and marine information-of which SST’s were a small part of the total. His 1855 book ‘The Physical Geography of the Sea” was considered required reading.

However, it was not until after World War II that the science of measuring SST gained momentum as new and more reliable methods of measuring them came about, given a further boost during the International Geophysical year in 1957/8. However, even in more recent years the ocean has continued to yield surprises, for example in the mid 1990’s it was discovered that deep ocean currents were both much stronger and much more variable than previously realised. In this context the development of increasingly accurate SST’s (when gathered from scientific sources), albeit still spatially incomplete, can be said to have arisen only over the last half century. Data collected prior to that has a big question mark over it as we shall discover.

 Chapter 5 of the book ‘Descriptive Physical Oceanography’ by M P M Reddy describes some of the methodology used, but is perhaps even more interesting for the general history preceding it.

Proper scientific expeditions, such as those mounted by the Royal Society in the 1820’s, were probably able to retrieve broadly accurate (to a few degrees)  sea temperatures (subject to all relevant caveats mentioned in the article about land temperatures) for a tiny stretch of ocean during a brief window of time before the ship moved on. Then the observer would be sampling a piece of ocean that might display completely different temperature characteristics, for as any swimmer will testify, water does not always mix very well, and as already observed, that depth plays a key part in temperature.

However, the majority of earlier records were not taken under strict scientific conditions, but on a much more casual ad hoc basis by members of the world’s navies, together with fishermen. The method of sampling was quite simple, whereby a wooden or canvas bucket was attached to a length of rope marked off in fathoms, the ensemble thrown overboard, the bucket subsequently raised and then a thermometer stuck in it to record the temperature of that small portion of the ocean. (see figure).  A method that, with small variations, persisted for 140 years.

Commenting on disparities in data this paper observes;

“…A new paper by David Thompson and other NOAA atmospheric scientists in Nature reports a different explanation (for accuracy). Most of the wartime measurements of sea temperatures factored into the global average came from US warships, which unlike the British navy tended to log engine room water intake thermometer readings as representing the temperature of the sea. The hardy jack tars who returned to meteorological duty as the war wound down instead relied as always on the time honoured method of throwing a bucket over the side, hauling it in, and putting it on deck for a thermometer wielding chief or officer to measure. The late Victorian change from oaken buckets to galvanized steel was compounded before World War II, when not just British, but Dutch and Japanese hydrographers were issued porous and hence cooling-prone canvas seawater scoops, a bad idea since the wind is generally brisk on a moving vessel.  Inevitably, the seawater sampled tended to cool – evidently measurably, in the time it took to present it on deck for measurement.”

The difficulty of keeping such a fragile instrument as a thermometer in one piece, let alone calibrated, can be imagined. On serious scientific expeditions it might be kept locked in the Captain’s cabin but in many other circumstances it might remain on a hook outside in all sorts of weather

This link  shows a ships barometer and thermometer from around 1855. The thermometer used in the bucket would often have been a robust standalone version of the instrument on the left of the main picture.

To put the problems inherent in recording ‘bucket’ temperatures in this fashion into their proper context, I can do no better than record the conversation I had some years ago with someone who had served in the British Navy in the 1940’s and 50’s when the bucket readings were still common (they finally finished in the early 60’s).

He matter of factly pointed out that the water was taken from all sorts of depths (greatly dependent on the strength and disposition of the person involved) and the water left in the container (not always an approved bucket) for indefinite periods of time, which included periods of hot sunshine and the cool of the night. Similarly, the quality of thermometer was not always of the highest, calibration infrequent and thermometers left in the ambient temperature on deck before often cursory readings were taken of the water sample, thereby compounding uncertainty.

His incredulous laughter as I recounted the great importance scientists attached to readings such as his is with me still. This is not to say of course that every SST was collected in this manner, but far too many for the general record to be considered to be scientifically robust and meaningful.

This from a 1947 paper “A new bucket for measurement of sea surface temperatures”

“It has been known for many years that the standard method of measuring sea surface temperatures by taking a sample with a canvas bucket is liable to serious errors.”


In this 1963 book H F Herdman commented that too often the sample is taken in a canvas bucket and the temperature read after an appreciable time.

This internal Met Office memorandum from 1985 by the highly respected D E Parker and is entitled “A comparison of bucket and non-bucket measurements of sea surface temperatures” concerns the differences of temperatures between water collected in insulated or non-insulated buckets. The author seems to fondly believe that, despite all the evidence to the contrary, water taken from buckets can be parsed to tenths of a degree. It also usefully shows the grid system used to measure data over the 70% of the globe that is ocean, and illustrates the precision which Hadley believes they have. Essential reading for all serious students of SST’s as it gives an indication of the early development of the SST data base.

There was a variation on the bucket method whereby the thermometer was put in the bucket before lowering over the side, with, in later years, a rubber cushion for protection. A reversing thermometer-generally used for greater depths than the surface -was developed in 1874 and in use from 1900 to 1970.

Generally, the SST data collected was of variable quality because of methodology and instrumentation quality, to which can be added lack of ‘spatial’ data-measurements were intermittently gathered over a tiny proportion of the world’s oceans as observed here:

‘Maps of mean temperature have also been made from ICOADS data. The data are poorly distributed in time and space except for some areas of the northern hemisphere. In addition, Reynolds and Smith (1994) found that ship temperature data had errors twice as large as temperature errors in data from buoys and AVHRR. Thus, space data processed by Reynolds are more accurate, and better distributed than ICOADS.’

A little more insight into the creation of sea surface temperature records was provided by Richard Verney (in the comments), here writing about the method that eventually superseded buckets — taking measurements from engine intakes:

“It seems to me that the Met Office and CRU do not understand how ship’s data is taken. Sea water temperature reported by ships is taken from the sea water drawn for cooling the engine. Where this is drawn from depends upon the design and configuration of the ship and whether the ship is proceeding in ballast or is laden. Ships try and avoid lengthy ballast voyages since these are not revenue earning legs. With a laden ship, the sea water drawn for cooling is drawn about 10m below the surface (could be anywhere between say 7m and 13m but about 10m is typical).

What does this mean? It means that ships are measuring sea temperature at a depth of about 10m whereas Buoys measure sea temperature at a depth of about 1m (to 3m). (Authors note; buckets measure temperatures at various depths, according to the disposition of the thrower, satellites at .01mm) Generally, the greater the depth, the cooler the water.

It follows from this (i.e. the depth differential) that one would expect ships engine intake data to record a lower temperature (not a higher temperature) when compared to the similar measurement taken by Buoys (measuring nearer the surface). This means that in order to make a like for like comparison, one should either adjust the Buoy temperature downwards, or the ship’s temperature upwards. There is no case for adjusting the Buoy temperature upwards since this further exacerbates the difference between the depth at which the data is taken.

Further, a not insignificant number of ships may have a tendency to deliberately under record the sea temperature. Many ships carry liquid cargoes that need to be heated (various chemicals, palm oils, veg oils etc.). In simple terms, the ship owner gets paid for heating these cargoes.…The ship owner gets paid for heating when he heats. Of course with very warm tropical seas, cargoes cool slower and the natural prevailing sea water temperature may be sufficient to keep the cargo free flowing such that much heating may not be required. It is therefore in the ship owner’s interest to record sea water temperatures slightly lower than those truly prevailing so that he can claim and charge for heating when in fact no heat is being applied. I am not saying that the practice is uniform throughout the shipping industry but it certainly does occur. Thus a number of ships are recording/reporting a lower temperature than that actually experienced.

Accordingly, for these two reasons, there is reason to believe that temperature records provided by ships under assess/under record the sea surface temperature. That being the case, sea temperatures may have decreased even more than the ‘team’ (or those closely connected with them and/or supporters thereof) is prepared to accept.”

This interpretation is confirmed by the  Seafriends site, of  Dr. J. Floor Anthoni, from which comes this excerpt concerning ocean temperatures:

“Ocean surface temperatures have been measured by ships for several centuries. First it was done by collecting surface water in a bucket while steaming on, but later the engine’s cooling water inlet was used. Unfortunately this made a difference, because the water inlet is at some depth under water. Today this may serve to advantage because satellite can measure only the top few centimeters of the sea because infrared radiation is rapidly absorbed by water. Because water continually evaporates from the sea, the surface film is somewhat colder than a few meters down.”

Notwithstanding the method of collection, it must also be recognised that the few readings that were taken came from shipping lanes that represented a tiny fraction of the oceans’ surface, and here we have another factor already touched on, for as well as the quality of the information there is an equal concern with the quantity of the data, as relatively few readings were taken, and the geographical coverage is much sparser than even the inadequate land temperature record.

This graphic – Reynolds 2000- shows the traffic of ships used for collecting SST Data during the week 1-8th January 2000 during a time of maximum economic activity. (Width of the lanes greatly exaggerated for pictorial purposes). It shows very poor coverage even then but a snapshot from 1850 would reveal magnitudes of lower activity. Observations and calculations of SST anomalies are shown in this WG2 analysis from the IPCC

Today’s hi tech version of these older methods of measurement are Argo buoys:

“Argo is a global array of 3,000 free-drifting profiling floats that measures the temperature and salinity of the upper 2000 m of the ocean.  This allows, for the first time, continuous monitoring of the temperature, salinity, and velocity of the upper ocean, with all data being relayed and made publicly available within hours after collection.”

This currently short lived experiment had a controversial beginning as the buoys initially recorded ocean heat content that was dropping (although not strictly under the same criteria as SST’s) This was an apparent anomaly as ocean temperatures were expected to show a rise, commensurate with computer models. This deviation was explained in this article which also provides useful information of what % of the whole-such as thermal expansion and glacier melt- is attributed to each aspect of sea level change.

“In 2004, Willis published a time series of ocean heat content showing that the temperature of the upper layers of ocean increased between 1993-2003. In 2006, he co-piloted a follow-up study led by John Lyman at Pacific Marine Environmental Laboratory in Seattle that updated the time series for 2003-2005. Surprisingly, the ocean seemed to have cooled.

Not surprisingly, says Willis wryly, that paper got a lot of attention, not all of it the kind a scientist would appreciate. In speaking to reporters and the public, Willis described the results as a “speed bump” on the way to global warming, evidence that even as the climate warmed due to greenhouse gases, it would still have variation. The message didn’t get through to everyone, though. On blogs and radio talk shows, global warming deniers cited the results as proof that global warming wasn’t real and that climate scientists didn’t know what they were doing.”

However, the interpretation given above has itself been superseded by this paper from November 2010  ‘Recent energy balance of earth’ by Knox and Douglass, who after researching the data from the Argo floats, show that for the most recent period 2003-8 ocean heat content was indeed shown to be still cooling-not warming.

Abstract. A recently published estimate of Earth’s global warming trend is 0.63 ± 0.28 W/m2, as calculated from ocean heat content anomaly data spanning 1993–2008. This value is not representative of the recent (2003–2008) warming/cooling rate because of a “flattening” that occurred around 2001–2002. Using only 2003–2008 data from Argo floats, we find by four different algorithms that the recent trend ranges from –0.010 to –0.160 W/m2 with a typical error bar of ±0.2 W/m2. These results fail to support the existence of a frequently-cited large positive computed radiative imbalance.

Considerable debate on the re-affirmation that oceans appear to be generally cooling sparked a lively debate on this blog post.

To date, what with the adjustments cited in the paper, the shortness of the project, and the difficulty in obtaining on going data, this method of collecting sea temperatures is as yet unproven.

Conclusions on Sea Surface Temperatures

With land temperatures we observed we were often comparing apples and oranges. Similarly it can be seen that with SST’s we are mixing a great variety of incompatible methods of collection, can observe that the number of samplings of the ocean are minute in terms of physical numbers, and recognize that the methodology itself is potentially severely flawed. In addition, the limited understanding we have of ocean temperatures drops exponentially the further back in time and the more remote the area, as the measuring points are so limited. To compound the problems, where the data is sparse it is statistically infilled from areas where it may still be sparse.

None of this will stop Hadley (and others) parsing global SST’s to a fraction and elaborating on the robustness of the answer, which the IPCC and  National Governments will then take as proof positive to enact yet more measures to guard against warming.

The basic historic temperature data,  land or surface, used in good faith by climate scientists, statisticians and analysts does not appear to meet basic quality control measures and are not fit for purpose-that of consistently determining temperatures to tenths of a degree.  Historic Sea Surface Temperatures in particular are highly uncertain and should not be considered as any sort of reliable measure.


220,000 log books of the Royal Navy from 1669 to 1976 are being studied for meteorological information that can give an insight into climate change.

A project to recover worldwide weather observations made by Royal Navy ships around the time of World War I.

Metrology of thermometers is the science of measuring as described here

Fascinating article about various types of historic measurements taken at sea

Bob Tisdale has an excellent web site dealing with all aspects of ocean temperatures