My preferred SST dataset has been HadSST3, for reasons noted at the end.  However, no new data has been provided for either February or March, so I have been looking at alternatives.  This post will feature ERSST5, with some comparisons with HadSST4 which has now been updated through March 2021.   First the usual contextual introduction.

Overview

The best context for understanding decadal temperature changes comes from the world’s sea surface temperatures (SST), for several reasons:

• The ocean covers 71% of the globe and drives average temperatures;
• SSTs have a constant water content, (unlike air temperatures), so give a better reading of heat content variations;
• A major El Nino was the dominant climate feature in recent years.

The Current Context

The various ERSST sources and history are described at the home page NOAA Extended Reconstructed Sea Surface Temperature (ERSST), Version 5.  A major distinction is the practice of interpolation, which involves infilling 2° by 2° grid cells missing sufficient observations in a month.  The values are anomalies from average anomalies for the period 1971 to 2000.  HadSST3 reports only on 5° by 5° grid cells observed in a month, and compares to a baseline 1961 to 1990. HadSST4 is the same as v.3, except that the older data from ship water intake was re-estimated to be generally lower temperature than shown in v.3.  The effect is that v4 has lower average anomalies for the baseline period 1961-1990, thereby showing higher current anomalies than v3. Clive Best has a fuller analysis comparing HadSST3 and HadSST4 in this post HadSST4 and knock on effects.

The chart below shows SST monthly anomalies as reported in HadSST4 starting in 2015 through March 2021. Though the anomaly values are higher than those reported in HadSST3, the patterns of sst changes remain the same. After three straight Spring 2020 months of cooling led by the tropics and SH, NH spiked in the summer, along with smaller bumps elsewhere.  Now temps everywhere are dropping the last six months, with all regions well below the Global Mean since 2015, matching the cold of 2018, and lower than January 2015. A small upward bump in March still leaves all regions the same as March 2015.

 

A global cooling pattern is seen clearly in the Tropics since its peak in 2016, joined by NH and SH cycling downward since 2016.  In 2019 all regions had been converging to reach nearly the same value in April.

Then  NH rose exceptionally by almost 0.5C over the four summer months, in August 2019 exceeding previous summer peaks in NH since 2015.  In the 4 succeeding months, that warm NH pulse reversed sharply. Then again NH temps warmed to a 2020 summer peak, matching 2019.  This has now been reversed with all regions pulling the Global anomaly downward sharply.

Note that higher temps in 2015 and 2016 were first of all due to a sharp rise in Tropical SST, beginning in March 2015, peaking in January 2016, and steadily declining back below its beginning level. Secondly, the Northern Hemisphere added three bumps on the shoulders of Tropical warming, with peaks in August of each year.  A fourth NH bump was lower and peaked in September 2018.  As noted above, a fifth peak in August 2019 and a sixth August 2020 exceeded the four previous upward bumps in NH.

And as before, note that the global release of heat was not dramatic, due to the Southern Hemisphere offsetting the Northern one.  The major difference between now and 2015-2016 is the absence of Tropical warming driving the SSTs, along with SH anomalies reaching nearly the lowest in this period. Presently both SH and the Tropics are quite cool, with NH coming off its summer peak.  Note the tropical temps descending into La Nina levels.  At this point, the 2016 El Nino and its NH after effects have dissipated completely.

 

ERSST5 reports only Global SSTs, unlike HadSST4 which also shows results for NH, SH and the Tropics (latitudes 20N to 20S).  The graph shows in green ERSST5 anomalies are much more volatile with both higher and lower extremes, compared to the blue HadSST4.  NH is added since it appears to vary similarly to ERSST, with the notable contradiction in 2016.  Also, the 2019 peak is much higher than 2015-16 in ERSST, whereas HadSST Global shows them comparable.  Both datasets show SSTs dropping sharply since summer 2020, and now below the mean anomaly for the period (only ERSST mean is shown).

A longer view of SSTs

The graph below  is noisy, but the density is needed to see the seasonal patterns in the oceanic fluctuations.  Previous posts focused on the rise and fall of the last El Nino starting in 2015.  This post adds a longer view, encompassing the significant 1998 El Nino and since.  The color schemes are retained for Global, Tropics, NH and SH anomalies.  Despite the longer time frame, I have kept the monthly data (rather than yearly averages) because of interesting shifts between January and July.

 

In the longer record the 1998 El Nino stands as one bookend and 2019 as the other.  Note that HadSST Global warming events appear more as extended periods of slightly higher anomalies, while ERSST events appear as sharp peaks and valleys.  Note also that present Global SSTs are matching the mean since 1995.

1995 is a reasonable (ENSO neutral) starting point prior to the first El Nino.  The sharp Tropical rise peaking in 1998 is dominant in the record, starting Jan. ’97 to pull up SSTs uniformly before returning to the same level Jan. ’99.  For the next 2 years, the Tropics stayed down, and the world’s oceans held steady around 0.2C above 1961 to 1990 average.

Then comes a steady rise over two years to a lesser peak Jan. 2003, but again uniformly pulling all oceans up around 0.4C.  Something changes at this point, with more hemispheric divergence than before. Over the 4 years until Jan 2007, the Tropics go through ups and downs, NH a series of ups and SH mostly downs.  As a result the Global average fluctuates around that same 0.4C, which also turns out to be the average for the entire record since 1995.

2007 stands out with a sharp drop in temperatures so that Jan.08 matches the low in Jan. ’99, but starting from a lower high. The oceans all decline as well, until temps build peaking in 2010.

Now again a different pattern appears.  The Tropics cool sharply to Jan 11, then rise steadily for 4 years to Jan 15, at which point the most recent major El Nino takes off.  But this time in contrast to ’97-’99, the Northern Hemisphere produces peaks every summer pulling up the Global average.  In fact, these NH peaks appear every July starting in 2003, growing stronger to produce 3 massive highs in 2014, 15 and 16.  NH July 2017 was only slightly lower, and a fifth NH peak still lower in Sept. 2018.

The highest summer NH peak came in 2019, only this time the Tropics and SH are offsetting rather adding to the warming. Since 2014 SH has played a moderating role, offsetting the NH warming pulses. Now September 2020 is dropping off last summer’s unusually high NH SSTs. (Note: these are high anomalies on top of the highest absolute temps in the NH.)

What to make of all this? The patterns suggest that in addition to El Ninos in the Pacific driving the Tropic SSTs, something else is going on in the NH.  The obvious culprit is the North Atlantic, since I have seen this sort of pulsing before.  After reading some papers by David Dilley, I confirmed his observation of Atlantic pulses into the Arctic every 8 to 10 years.

But the peaks coming nearly every summer in HadSST require a different picture.  Let’s look at August, the hottest month in the North Atlantic from the Kaplan dataset.
The AMO Index is from from Kaplan SST v2, the unaltered and not detrended dataset. By definition, the data are monthly average SSTs interpolated to a 5×5 grid over the North Atlantic basically 0 to 70N. The graph shows August warming began after 1992 up to 1998, with a series of matching years since, including 2020.  Because the N. Atlantic has partnered with the Pacific ENSO recently, let’s take a closer look at some AMO years in the last 2 decades.

 

This graph shows monthly AMO temps for some important years. The Peak years were 1998, 2010 and 2016, with the latter emphasized as the most recent. The other years show lesser warming, with 2007 emphasized as the coolest in the last 20 years. Note the red 2018 line is at the bottom of all these tracks. The black line shows that 2020 began slightly warm, then set records for 3 months. then dropped below 2016 and 2017, peaked in August and is now below 2016. Note this year is starting out among the coolest analog years.

Summary

The oceans are driving the warming this century.  SSTs took a step up with the 1998 El Nino and have stayed there with help from the North Atlantic, and more recently the Pacific northern “Blob.”  The ocean surfaces are releasing a lot of energy, warming the air, but eventually will have a cooling effect.  The decline after 1937 was rapid by comparison, so one wonders: How long can the oceans keep this up? If the pattern of recent years continues, NH SST anomalies may rise slightly in coming months, but once again, ENSO which has weakened will probably determine the outcome.

Footnote: Why Rely on HadSST3

HadSST3 is distinguished from other SST products because HadCRU (Hadley Climatic Research Unit) does not engage in SST interpolation, i.e. infilling estimated anomalies into grid cells lacking sufficient sampling in a given month. From reading the documentation and from queries to Met Office, this is their procedure.

HadSST3 imports data from gridcells containing ocean, excluding land cells. From past records, they have calculated daily and monthly average readings for each grid cell for the period 1961 to 1990. Those temperatures form the baseline from which anomalies are calculated.

In a given month, each gridcell with sufficient sampling is averaged for the month and then the baseline value for that cell and that month is subtracted, resulting in the monthly anomaly for that cell. All cells with monthly anomalies are averaged to produce global, hemispheric and tropical anomalies for the month, based on the cells in those locations. For example, Tropics averages include ocean grid cells lying between latitudes 20N and 20S.

Gridcells lacking sufficient sampling that month are left out of the averaging, and the uncertainty from such missing data is estimated. IMO that is more reasonable than inventing data to infill. And it seems that the Global Drifter Array displayed in the top image is providing more uniform coverage of the oceans than in the past.