South Pacific SST Patterns

I’ve moved to WordPress.  This post can now be found at South Pacific SST Patterns

##################
INTRODUCTIONThere are visitors here who probably wondered why, in my prior post about the South Pacific Hot Spot, I did not use a longer-term dataset such as ERSST.v3b or HADISST. And there are other visitors who would assume to know the reason why I hadn’t bothered. This post is for both. I had mixed emotions about starting it because it really would not prove anything; it would simply illustrate very basic facts about Sea Surface Temperature (SST) reconstructions.

The Sea Surface Temperature (SST) dataset used for the Hadley Centre and NOAA/NCDC SST reconstructions contains large areas with little to no data, especially in the South Pacific. And this low sampling continued through the first few decades of the satellite era. So if I used the longer dataset, instead of the satellite-based OI.v2, I would have had to assume the assumptions made by the researchers during the reconstruction of South Pacific SST data were correct.

Keep in mind, this post is about patterns, not time-series graphs. When looking at basin-wide SST data in a time-series graph, the different methods used to infill the gaps create differences in linear-trends, but the year-to-year wiggles tend to follow one another, though there are some occasional exceptions. I’ll provide long- and short-term graphs with linear trends at the end of this post for those interested.

Sea Surface Temperature (SST) before the advent of buoys and satellites was measured by ships. The majority of ships traversing the oceans kept to shipping lanes. This limited the spatial coverage of SST measurements, especially in the South Pacific. Figure 1 is a collection of COADS SST data coverage maps available from the National Center for Atmospheric Research (NCAR). More on COADS later. And here are the links to the full-size individual maps:
http://www.cgd.ucar.edu/cas/guide/Data/coads.sst.f1.html
http://www.cgd.ucar.edu/cas/guide/Data/coads.sst.f2.html
http://www.cgd.ucar.edu/cas/guide/Data/coads.sst.f3.html

NCAR describes the maps as “Percentage of non-missing data in each time period is plotted. The minimum number of observations needed per month per grid box was 1.” Non-missing data is an interesting but descriptive explanation. It definitely gets the point across that missing data is the norm in many areas. Each of the maps illustrates the monthly SST coverage for a 20-year period, starting in 1861 and ending in 1997. Zero to 10% coverage is in white, while complete coverage (90 to 100%) is shown in gold. Even the last period from 1981 to 1997 shows major gaps in the data for the South Pacific.

http://i48.tinypic.com/2i7spwy.png
Figure 1

Assume for example, after posting the discussion of the South Pacific Hot Spot, I wanted to take a longer look at South Pacific Sea Surface Temperature (SST) anomaly maps during El Nino events, extending the maps back in time, with hope of determining if the El Nino-induced hotspot appeared regularly during Decembers from the 1950s through the 1970s. The first official El Nino was in 1951/52. Knowing that the spatial coverage of South Pacific SST measurements is extremely poor, I would first use the KNMI Climate Explorer to create a map of ICOADS SST anomalies for December 1951. ICOADS is short for International Comprehensive Ocean-Atmosphere Data Set. The KNMI link for the COADS data will bring you to the ICOADS 2 Degree webpage here:
http://www.esrl.noaa.gov/psd/data/gridded/data.coads.2deg.html

ICOADS SST data is the basis for the SST reconstructions available through the Hadley Centre and NCDC. The NCDC uses the data for its ERSST.v2 and ERSST.v3b datasets:
http://www.ncdc.noaa.gov/oa/climate/research/sst/ersstv3.php

The Hadley Centre’s HADSST2 data is also based on COADS data until 1998:
http://hadobs.metoffice.com/hadsst2/

And here the Hadley Centre notes that the ICOADS data is used to supplement their Met Office Marine Data Bank (MDB) used in their HADISST dataset:
http://www.ncof.gov.uk/hadisst/index.html

And here’s a paper that says that MDB data is a part of COADS:
ftp://ftp.wmo.int/Documents/PublicWeb/amp/mmop/documents/JCOMM-TR/J-TR-13-Marine-Climatology/REV1/joc1166.pdf

As illustrated in Figure 2, the COADS SST data for December 1951 is sparse, very much so in the South Pacific. Based on the poor data distribution, is it possible to determine if the December South Pacific hotspot was a reoccurring effect of El Nino events, at least in 1951? Possibly. If I looked at the SST anomaly patterns presented by the reconstructions, I would have to assume that the assumptions made by the researchers while preparing their reconstructions represented reality.
http://i46.tinypic.com/15zm49f.png
Figure 2

LOADS OF INFILLING

But for this post, I decided to take a look at what the individual datasets (ERSST.v2, ERSST.v3b, HADSST2, and HADISST) actually showed for December during the El Nino years since 1951. What follows are simple visual comparisons—nothing statistical—just visuals. Figure 3 illustrates maps of the South Pacific SST anomalies for December 1951 for each of those four datasets. Also illustrated is the COADS data in the upper left-hand corner. The NCDC’s ERSST.v2 occupies the left-center cell, and their ERSST.v3b the right-center cell. The Hadley Centre’s two datasets occupy the lower cells, with HADISST on the left and HADSST2 on the lower right. The Hadley Centre HADSST2 data appears to be a 5-degree grid version of the 2-degree grid COADS data, with the missing data clearly visible.
http://i49.tinypic.com/4t4mrr.png
Figure 3

One thing for sure, the El Nino-induced hotspot in the mid-to-high latitudes of the South Pacific exists in all four of the reconstruction datasets in December 1951. The patterns vary, but the hotspots are there.

Note the differences in the patterns between the NCDC’s ERSST datasets and the two datasets from the Hadley Centre, however. The NCDC’s areas of positive and negative anomalies are much better defined when compared to the HADISST data. Also note that the HADSST2 data does not infill like the others. Is one better? Impossible to say.

Referring to Figure 3, I always find it remarkable that the NCDC and Hadley Centre can infill that much missing data. There may be other variables employed, but they are not discussed in the overviews. Also, the one month snapshots are deceiving. The researchers can also infill over time. That is, readings from the months before and after December 1951 could be used to help infill the missing data, especially if the coverage differed. If we look at South Pacific COADS data for the 5-month period from October 1951 to February 1952, (including two months before and after December 1951), Figure 4, the spatial coverage is much better, though there are still large areas of missing data. And whether there is a hotspot there would be a matter of debate.
http://i46.tinypic.com/2vafs7t.png
Figure 4

Papers can be found through the links above for discussions of the methods employed by the researchers. And as discussed in those papers, the methods used in the reconstructions do differ, so in that respect, the SST patterns that result from the infilling reflect the researchers’ beliefs on how SST anomalies should respond in certain conditions.

PRE-SATELLITE ERA SST ANOMALY PATTERNS – 1957 TO 1977
Again, for example, I produced maps for each of the datasets for the December South Pacific SST anomalies during El Nino events, to see if we could determine if the South Pacific hotspot was a reoccurring effect of ENSO. Figures 5 through 12 are for the period of 1957 to 1977. The satellite-based SST data begins in 1979, so this data represents the 20+ years prior to then. The ONI Index was used as reference for “full-fledged” El Nino events. While there are differences in the patterns from El Nino to El Nino and between the datasets, and if we assume that the reconstructions represent actual South Pacific SST anomalies, it could easily be stated that the hotspot presently being experienced is not unusual.

Also, note the differences between the first and second years of the multiyear 1968/69/70 El Nino, but the similarities between the first and second years of the multiyear 1976/77/78 El Nino.
http://i46.tinypic.com/2unub02.png
Figure 5 - 1957
###################
http://i50.tinypic.com/fmoy9z.png
Figure 6 - 1963
###################
http://i47.tinypic.com/18zssl.png
Figure 7 - 1965
###################
http://i50.tinypic.com/9h0sck.png
Figure 8 - 1968
###################
http://i47.tinypic.com/3322f44.png
Figure 9 - 1969
###################
http://i47.tinypic.com/32zuomt.png
Figure 10 - 1972
###################
http://i46.tinypic.com/bjbszo.png
Figure 11 - 1976
###################
http://i45.tinypic.com/1krit.png
Figure 12 - 1977

SATELLITE ERA SST ANOMALY PATTERNS – 1982 TO 2006

For the satellite era comparisons, Figures 13 through 21, I’ve also added NCDC OI.v2 SST anomaly maps to the graphics, in the upper right-hand cell. Not surprisingly, the SST anomaly patterns of the OI.v2 and HADISST data tend to agree. And in most cases, the patterns of the very fragmented HADSST2 data appear to agree with those of the two satellite-based datasets.

But note the disagreement between the NCDC’s ERSST datasets and the satellite-based data. It almost appears that the methods used by the NCDC suppress negative anomalies and amplify positive anomalies, many times changing the appearance of the pattern by doing so. This was the unusual finding of this post.

And again, as in my earlier discussion of the South Pacific Hot Spot, the hotspot does not appear as consistently between the 1982/83 and 1997/98 El Nino events. Is this a result of volcanic aerosols from the eruptions of El Chichon and Mount Pinatubo?
http://i45.tinypic.com/29qi440.png
Figure 13 - 1982
###################
http://i50.tinypic.com/jrxlw6.png
Figure 14 - 1986
###################
http://i50.tinypic.com/16k767p.png
Figure 15 - 1987
###################
http://i45.tinypic.com/346t8r8.png
Figure 16 - 1991
###################
http://i45.tinypic.com/f44ks7.png
Figure 17 - 1994
###################
http://i47.tinypic.com/650ugn.png
Figure 18 - 1997
###################
http://i46.tinypic.com/iyikjo.png
Figure 19 - 2002
###################
http://i50.tinypic.com/103z8li.png
Figure 20 - 2004
###################
http://i45.tinypic.com/1zdq5cg.png
Figure 21 - 2006

Not all of the datasets have been updated for December 2009, which is why I have not presented it here.

A QUICK LOOK AT SOUTH PACIFIC TIME-SERIES GRAPHS
Though this is a post about SST patterns, there are those who would be interested in seeing a comparison time-series graph of the various datasets. Figure 22 is that spaghetti graph, showing the South Pacific SST anomalies and trends for the various SST datasets from January 1950 to April 2007, which is when the COADS data on the KNMI Climate Explorer ends. As illustrated, the year-to-year variations are similar for all of the datasets, but there are substantial differences in the trends. The linear trend of the HADSST2 data almost doubles that of the HADISST data. Some of that additional variability in the HADSST2 data resulted when the Hadley Centre changed SST data suppliers in 1998. The switch created a step change in the HADSST2 data that does not exist in the other datasets. Refer to The Step Change in HADSST Data After the 1997/98 El Nino and in Met Office Prediction: “Climate could warm to record levels in 2010”.

The upward step changes in South Pacific SST anomalies that occurred in the late 1970s (the Great Pacific Climate Shift) and after the 1997/98 El Nino are very obvious.
http://i50.tinypic.com/w990lf.png
Figure 22

And in Figure 23, I’ve added the NCDC’s OI.v2 SST data to the spaghetti graph of shorter-term SST anomaly data for the South Pacific. Note that the two satellite-based SST datasets, HADISST and OI.v2, have significantly lower trends than the other datasets. The third lowest was the ERSST.v3b.
http://i47.tinypic.com/3005csw.png
Figure 23

CLOSING

I started this post off by saying it really would not prove anything; it would simply illustrate very basic facts about Sea Surface Temperature (SST) reconstructions. And it did illustrate something well known to many: that the SST sampling is poor in decades past and, putting aside satellite measurement which is excluded purposely in some datasets, SST sampling really didn’t get any better until recently.

Where did we wind up with the South Pacific hotspot?

Even if we were to rely solely on one-month (December) snapshots of the raw ICOADS data for the period before the satellite era, upper left-hand corner cells in Figures 5 through 12, it appears as though the hotspots in the South Pacific were a regular December phenomenon during El Nino events.

What should also be apparent is that when studying SST data outside of the shipping lanes, one is relying on infilled data and on the assumptions made by the researchers to infill the missing data. And as you will note, there are differences in patterns between datasets.

SOURCE

SST anomaly data and maps are available through the KNMI Climate Explorer:
http://climexp.knmi.nl/selectfield_obs.cgi?someone@somewhere