The paper by Karl et al. (2015) published today in Science is an ‘express’ report and not up to the standards of a comprehensive paper. It is a highly speculative and slight paper that produces a statistically marginal result by cherry-picking time intervals, resulting in a global temperature graph that is at odds with all other surface temperature datasets, as well as those compiled via satellite.
Key pitfalls of the paper:
* The authors have produced adjustments that are at odds with all other surface temperature datasets, as well as those compiled via satellite.
* They do not include any data from the Argo array that is the world’s best coherent data set on ocean temperatures.
* Adjustments are largely to sea surface temperatures (SST) and appear to align ship measurements of SST with night marine air temperature (NMAT) estimates, which have their own data bias problems.
* The extent of the largest SST adjustment made over the hiatus period, supposedly to reflect a continuing change in ship observations (from buckets to engine intake thermometers) is not justified by any evidence as to the magnitude of the appropriate adjustment, which appears to be far smaller.
1. They make 11 changes (not all are explained) producing the ERSSTv4 Sea Surface Temperature (SST) dataset that includes new estimates for the different way SSTs are measured from ships (intake or buckets). They also add 0.12°C to each buoy to bring their measurements in line with those taken from ships. These issues have been raised before by the UK Met Office when compiling their HadSST3 ocean surface temperature dataset, see, ‘A review of uncertainty in in situ measurements and data sets of sea surface temperature’
2. The greatest changes are made since 1998, which is interesting because this is when we have the highest quality of data and global coverage using several methods. Only this analysis finds any increase in global annual average surface temperature over this “hiatus” period. The authors have produced a dataset that is at odds with other surface temperature datasets, as well as those compiled via satellite.
3. The authors start their trend estimates in 1998 and 2000. This has long been considered unwise as 1998 is a very strong El Nino year and 1999-2000 is a much cooler La Nina period. The difference between them distorts their trend estimates. For example, their 1998-2014 trend is 0.106+/- 0.058°C per decade. Starting two years later (during La Nina influenced years) yields a trend of 0.116 +/- 0.067°C per decade as one would expect from starting at a lower temperature. Ignoring these caveats the authors say their analysis produces twice as much warming for 1998-2014 than earlier estimates. Their conclusion is, ironically, based on inbuilt biases in their analysis.
Their Fig 1 shows that when using their updates it is only with the use of these inappropriate start and end points that the “hiatus” is reduced.
“I believe their estimates of the error in their decadal trend figures are far too small. They quote the error in a 15-year period to a precision of one thousandth of a degree C. In their report the authors admit that their error analysis is not definitive and that looking at them another way invalidates their trend conclusions,” said Dr David Whitehouse, science editor of the GWPF.
5. Note that trends that include 2014 and 2015 must be treated with caution due to a recently persistent very warm feature in the NE Pacific that is affecting global SST estimates.
6. In addition, they do not include any data from the Argo array that is our best coherent data set on ocean temperatures. The authors state this is because Argo temperature data is not surface data. However, ship-derived temperatures can be from as much as 15 m below the surface. The Argo array samples 5 m below the top of the ocean. From 2004 to 2013 it shows considerable variation and little trend. The non-ARGO data aptly demonstrates the problem of starting trend analysis in 1998 or 2000.
7. Their conclusions are also at odds with satellite data that shows no trend in the past 16-years or so.
8. Extending a change in ship observations (from buckets to engine intake thermometers) to the present time had the largest impact on the SST adjustments over the hiatus period, per Karl et al 2015:
“Second, there was a large change in ship observations (i.e., from buckets to engine intake thermometers) that peaked immediately prior to World War II. The previous version of ERSST assumed that no ship corrections were necessary after this time, but recently improved metadata (18) reveal that some ships continued to take bucket observations even up to the present day. Therefore, one of the improvements to ERSST version 4 is extending the ship-bias correction to the present, based on information derived from comparisons with night marine air temperatures. Of the 11 improvements in ERSST version 4 (13), the continuation of the ship correction had the largest impact on trends for the 2000-2014 time period, accounting for 0.030°C of the 0.064°C trend difference with version 3b.”
Ref (18) is a 2011 paper by Kennedy et al. It states (paragraph 3.1):
“Dating the switchover from uninsulated canvas buckets to insulated rubber buckets is problematic as it is not clear how quickly the practice of using insulated buckets was adopted. … Based on the literature reviewed here, the start of the general transition is likely to have occurred between 1954 and 1957 and the end between 1970 and 1980.”
A 2010 review article “Effects of instrumentation changes on SST measured in situ” by Kent, Kennedy, Berry and Smith states that “Models of corrections for wooden and uninsulated canvas buckets show the adjustments to be five to six times greater for the canvas buckets.”
So post 1980 adjustments to bucket measurements should be very small (under 0.1 C) Moreover, by 2000 ship measurements were a minority of total measurements and all types of bucket were a small proportion of ship measurements (see figs 2 and 3 of Kent et al. 2010). These facts imply that post 2000 adjustments warranted by use in some ships of bucket measurements should be negligible.
“The justification given for the change that had the largest impact on trends for the 2000-2014 time period – continuing to adjust ship SST measurements by reference to night marine air temperature (NMAT) data, ‘which have their own particular pervasive systematic errors’ (Kennedy 2014) – i.e. that some ships still continue to take bucket observations, appears to support only a very small adjustment,” said Nic Lewis, an independent climate scientist.
In summary
This is a highly speculative and slight paper that produces a statistically marginal result by cherry-picking time intervals, resulting in a global temperature graph that is at odds with those produced by the UK Met Office and NASA.
Caution and suitable caveats should be used in using this paper as evidence that the global annual average surface temperature “hiatus” of the past 18 years has been explained.