Getting rid of the spurious blips - another look at global sea surface temperatures

Kevin Cowtan | Source: U York

Once again, Kevin Cowtan has brought his skills to climate science, working with Robert Rohde and Zeke Hausfather. They decided to explore those pesky ups and downs in the temperature record, which a lot of people (scientists mostly) have expressed concerns about. This new paper is largely addressing bucket bias in sea surface temperature and is very detailed. Taken with his other work on temperature records, this has to cement Kevin Cowtan's place among the "serious climate nerds" (h/t ykw!)

A hybrid check on sea surface temperature bias corrections

The authors analysed sea surface temperature, the main source of the temperature blips, from a new perspective. Their analysis can be seen mainly as a check of the bias corrections used in other sea surface temperature records. Instead of reanalysing data from ships and buoys, they compared weather stations on the coast and on islands with the measurements taken on ships when they passed close to the coast. They subjected this to further analysis and called the result a hybrid SST (sea surface temperature).

I can only imagine how much work this must have entailed. There are hints in the paper. Not only did they get the temperature records from land and nearby sea, they made adjustments in their analysis to compensate for the fact that with global warming, the land surface is warming faster than the sea surface, plus more.

They used their results to assess the bias correction that needs to be made when the sources for sea surface temperature changed, such as from buckets to engine intake, and to buoys (see below). The end result was a different check on sea surface temperatures and additional evidence that:

• Some of the odd blips in the temperature records were not what actually happened - particularly the upward WWII blip and the drop down around 1910
• The NOAA sea surface temperature record from 1997 onwards is probably closest to reality. On the other hand, the Cowtan17 analysis indicates ERSST v4 is too warm in the earliest years (1860 to 1900 or so) and too cool in the early 20th century (1910 to late 1930s).
• Climate models reflect reality even more closely than previous records suggest. 

There's an excellent article on Kevin Cowtan's website which explains the research, and accompanying provisos. The paper and supporting information contain a lot more detail, including all the ifs and buts and maybes. Co-author Zeke Hausfather has a  Twitter thread about the paper, too.

Challenges in the historical record of sea surface temperature

The authors begin by pointing out that getting a record of sea surface temperature is more challenging in many ways than putting together land temperature records. The difficulty with sea surface temperature is that information sources change much more than those on land.

On the land, apart from getting as many records together as possible (thank you CRU and other early collectors, and more recently ISTI), the main issues to contend with are adjusting for changes in instrument design and location. Location changes can be identified from station records or inferred from abrupt changes in the record compared with neighbouring records. Technological change hasn't happened all that often in the past 150 years or so. The main ones include the introduction of the Stevenson screen way back, and the more recent shift to automatic weather stations with resistance probes replacing mercury thermometers.

On the sea, the problems include the different sources for temperature readings: buckets of differing materials being dipped into the sea, engine room intakes, sensors on the ships hull and, more recently, drifting buoys and satellites. Within all that, scientists have to account for things like changes in the height of ship decks, interruptions to the consistency of records caused by world wars (where the data source changed from predominately merchant ships to predominately naval vessels), and more. The marvel is that researchers have worked through all these difficulties and developed records of sea surface temperature going back many decades.

Questionable peaks and troughs in the SST records - WWII and all that

One period about which most scientists who've worked on the subject have had most issue with are the years of the second world war (WWII). Some data sets show a peak in temperature that has not been easily explained by weather or climate change phenomena. In addition, previous records show a drop in the temperature around 1910 that looks a bit odd. In this paper, the authors did not find the spike that exists in ERSST v5 and to a lesser extend in HadSST3. Neither did they find the drop in temperature in the early 1900s.

In the top chart below, the hybrid record is shown in blue. The different series are a bit hard to distinguish so you might want to click on the image to enlarge it.

Figure 1 | Comparison of the coastal hybrid temperature reconstruction (using all coastal stations and fitting the global mean of the coastal temperature differences only) to co-located data from HadSST3 and ERSSTv5 for the period 1850-2016. Spatial coverage is that of HadSST3 for all of the records, with coastal cells weighted by ocean fraction.The shaded region is the 95% confidence region for the HadSST3 anomalies including combined bias adjustment and measurement and sampling errors. The lower panel shows the adjustment applied to the raw data in the HadSST3 and coastal hybrid records. A comparison with the ERSSTv4 ensemble is shown in Figure S7. Source: Cowtan 17 Figure 12.

To help see the difference, the chart below compares the Cowtan17 hybrid record with NOAA's ERSST v4 record. As discussed, the two are very similar in the most recent decades, but differ much more in the period prior to the early 1940s.

Figure 2 | Comparison of coastal hybrid temperature reconstruction to the ERSSTv4 ensemble. The dotted line is the ensemble median, while the shaded region is the 95% range of the ERSSTv4 1000 member ensemble from Huang et al (2016). DOI: 10.1175/JCLI-D-15-0430.1 . Source: Cowtan17 supporting information Figure S7

The table below highlights further that the Cowtan17 analysis is closer to the NOAA data set for the period after WWII since 1997 than it is to the Hadley record (HadSST3). The trend of HadSST3 is lower than that found in Cowtan17 and ERSST v4.

Table 1: Trend in sea surface temperature since 1997. Source: Cowtan17 Supporting Information Table S2.

The analysis supports the CMIP5 models

Another thing the analysis suggests is that there is less of a difference between observations and the blended mean from CMIP5 model runs. This is shown in the chart below, from Kevin Cowtan's briefing paper, where the green line is the CMIP5 blended mean.

Figure 3 | Comparison of global temperature records based on either the UK Met Office sea surface temperature record (HadSST3), or our coastal hybrid record. The smoothed records are compared to the average of climate model simulations from the CMIP5 project. The lower panel shows the differences between each set of observations and the models. Source: Kevin Cowtan's blog article.

Constraints and provisos

The authors of Cowtan17 show a lot of restraint and go into quite a bit of discussion of uncertainties and provisos. They present their findings not as the be all and end all of temperature reconstruction, but as a suggestion of where to investigate further. Kevin Cowtan wrote in his briefing:

However we do not necessarily trust our new record, because of the assumptions we had to make in constructing it. The most important result of our work may therefore be to identify places where extra attention should be given to addressing problems in the existing sea surface temperature records. A secondary result is that caution is required when trying to draw conclusions about any differences between the models and the observations, whether it be to identify internal cycles of the climate system or problems in the models, because the differences that we do see are mostly within the range of uncertainty of the observations.

Just the same, this paper has a lot of merit, looks at the data differently, and shows that the spurious peaks and troughs from years gone by may indeed be out of whack. It also supports the records in recent times, which seems to me to add weight to their findings.

What deniers are saying about Cowtan17

Nothing. At least nothing at WUWT or anywhere else that I've seen. Either they all missed the paper because it came out in the holidays, or they haven't figured out what to say about it.

References and further reading

Cowtan, K., Robert Rohde, and Zeke Hausfather. "Evaluating biases in Sea Surface Temperature records using coastal weather stations." Quarterly Journal of the Royal Meteorological Society (2017). DOI: 10.1002/qj.3235 (pdf here)

• Evaluating biases in Sea Surface Temperature records using coastal weather stations - overview, briefing document, links to methods and data and more from Kevin Cowtan
• Twitter thread from Zeke Hausfather