Tisdale on comparing global climate data with SST

There are a numerous blogosphere posts about the global coverage, or lack thereof, of the GISS and Hadley Centre land plus sea surface temperature datasets. Few include the NCDC product. This post provides sample maps from 1880 to 2010 comparing NCDC merged land+sea surface temperature data to those of GISS (GISTEMP with 1200km radius smoothing–LOTI) and Hadley Centre (HADCRUT). It also shows GHCN and ERSST.v3b maps, which are the sources for the NCDC merged product. And this post illustrates how NCDC deletes infilled data in early years, discusses why they delete the data, and shows the very limited impact of the NCDC’s deletion of that data in early years.

The NCDC merged land+sea surface temperature anomaly data is now available through the KNMI Climate Explorer. (Many thanks to Dr. Geert Jan van Oldenborgh.)

Figure 1 compares the July 2010 temperature anomaly map of the NCDC merged Land+Sea Surface Temperature to those of the GISS and Hadley Centre products. I’ve used the base years of 1901 to 2000 for all datasets. These are the base years of the NCDC data, not their dot-covered maps. And the contour levels of the maps were set for a range of -4.0 to 4.0 deg C.

As illustrated, the NCDC does not present data over sea ice. Also, there is a sizeable area of east-central Africa without data during July 2010. And the NCDC does not present Antarctic data. The infilling methods employed by the NCDC provide greater land surface coverage than the Hadley Centre product but less coverage than GISS. The methods used by NCDC are discussed in Smith et al (2008) Improvements to NOAA’s Historical Merged Land-Ocean Surface Temperature Analysis (1880-2006), and in Smith and Reynolds (2004) Improved Extended Reconstruction of SST (1854-1997).

http://i35.tinypic.com/2uy1x6o.jpg

Figure 1

GISS includes more Arctic surface station data than NCDC and Hadley Centre. This can be seen in the maps that compare the NCDC GHCN data, the Hadley Centre CRUTEM3 data, and the GISS land surface data with 250km radius smoothing, Figure 2. GISS includes Antarctic surface stations (not illustrated), which are not included in GHCN. And of course, GISS Deletes Arctic And Southern Ocean Sea Surface Temperature Data and extends land surface data out over the oceans to increase coverage in the Arctic and Antarctic.

http://i36.tinypic.com/e7glzq.jpg

Figure 2

The GISTEMP combined land plus sea surface temperature dataset with 250km radius smoothing is used to show how little Arctic Ocean sea surface temperature data remains in the GISS product. Refer to the bottom cell in Figure 3. The NCDC and Hadley Centre, on the other hand, include Arctic Ocean Sea Surface Temperature data during seasons with reduced sea ice.

http://i34.tinypic.com/55jcxe.jpg

Figure 3

Figures 4 through 8 provide global coverage comparison maps for NCDC, Hadley Centre and GISS surface temperature products from 2010 to 1880. Januarys in 2010, 1975, 1940, 1910, and 1880 are shown. Note how the coverage decreases in early years. The exception is the SST data presented by GISS. Keep in mind that the three SST datasets prior to the satellite era basically use a common source SST dataset, ICOADS. Refer to An Overview Of Sea Surface Temperature Datasets Used In Global Temperature Products. The HADSST2 data in the Hadley Centre maps represents the locations of the SST samples. The HADISST and ERSST.v3b datasets used by GISS and NCDC are infilled using statistical methods.

http://i37.tinypic.com/1215rag.jpg

Figure 4

####################

http://i34.tinypic.com/29yqvsl.jpg

Figure 5

####################

http://i35.tinypic.com/24dplkj.jpg

Figure 6

####################

http://i34.tinypic.com/28mixyb.jpg

Figure 7

####################

http://i33.tinypic.com/28w1bus.jpg

Figure 8

####################

The decrease in land surface coverage is not surprising, but the NCDC uses ERSST.v3b SST data for the oceans and that dataset provides complete coverage for the oceans even in early years. This can be seen in Figures 9 through 13. They include the same Januarys as the maps above, but they present the NCDC merged product and the GHCN land surface data and ERSST.v3b sea surface data used by NCDC. The NCDC infilled much of the Sea Surface Temperature data in early years. Why do they then delete so much of it? The answer follows the maps.

http://i36.tinypic.com/24b2sup.jpg

Figure 9

####################

http://i37.tinypic.com/2isd2e8.jpg

Figure 10

####################

http://i33.tinypic.com/bgbfxt.jpg

Figure 11

####################

http://i35.tinypic.com/2vcezhf.jpg

Figure 12

####################

http://i38.tinypic.com/11uuavo.jpg

Figure 13

####################

In Smith et al (2008), Improvements to NOAA’s Historical Merged Land-Ocean Surface Temperature Analysis (1880-2006), the NCDC describes why they delete data from their merged product. On page 6, under the heading of “Sampling cutoffs for large-scale averaging”, they write, “The above results show that the reconstructions can be improved in periods with sparse sampling. However, there can still be damping errors in periods with sparse sampling. Damping of large-scale averages may be reduced by eliminating poorly sampled regions because anomalies in those regions may be greatly damped. In Smith et al. (2005) error estimates were used to show that most Arctic and Antarctic anomalies are unreliable and those regions were removed from the global average computation. Here testing using the simulated data is done to find objectively when regions should be eliminated from the global average to minimize the MSE [global mean-squared error] of the average compared to the full data.” Smith et al then go on to describe the criteria for deleting the data in poorly sampled regions.

Of course, the question that comes to mind is, what impact does deleting the all of that SST data have on the long-term trends? Answer: very little. Figures 14 and 15 compare SST data for the NCDC merged product and the source SST data in the North and South Pacific. The coordinates (illustrated on the graphs) were chosen to capture large portions of those ocean subsets, while making sure they were free of influences from land surface data and sea ice. As shown, the NCDC merged data become much more volatile during periods of reduced coverage, but there is little impact on the long-term trends.

http://i38.tinypic.com/11h3vk1.jpg

Figure 14

####################

http://i37.tinypic.com/1juvdy.jpg

Figure 15

Makes one wonder, doesn’t it?

SOURCE

The data and maps are available through the KNMI Climate Explorer:

http://climexp.knmi.nl/selectfield_obs.cgi?someone@somewhere