Frank Lansner’s post Did GISS discover 30% more land in the Northern Hemisphere? at Jo Nova’s blog created a recent stir. Watts Up With That ran a similar post by Frank, Tipping point at GISS? Land and sea weight out of balance. Both posts spawned rebuttals/explanations, including Zeke Hausfather’s post The GISTemp Land Fraction at Lucia’s The Blackboard and my post Notes On The GISTEMP Ratio Of Land To Sea Surface Temperature Data, which Anthony Watts co-posted as GISS land and sea ratios revisited, on the same day that he ran Frank Lansner’s post.
Basically, Frank Lansner’s post contends that GISS has increased the ratio of land to sea surface data with time, from zero percent early in the 20th century to near 70% in recent years. With a close examination of the graph that was being presented as a reference, Frank’s land surface data looked unusual, and I believe Frank’s observations are skewed by his choice of base years, and possibly by his smoothing method. I discussed this with him in a detailed comment that I posted at WattsUpWithThat and Jo Nova’s website. Refer to:
OVERVIEW OF THIS POST
There are lingering beliefs that there’s something unusual about the way GISS handles land surface data. In an effort to dispel those misunderstandings, the land surface data contribution to combined land and sea surface temperature data will be illustrated in this post, using a very simple method. Sea surface temperature data will be subtracted from GISS, Hadley Centre, and NCDC combined (land and sea) surface temperature products. The remainders, which are the contributions of the land surface temperature data to the combined products, will be compared. Personally, I was surprised with the results. But first, we need to eliminate the effects of known differences between the GISS and the other two global temperature datasets.
GISS treats the polar regions differently than the Hadley Centre and NCDC. GISS has better land surface temperature data coverage than the Hadley Centre and NCDC in the Arctic and Antarctic. And the Hadley Centre and NCDC include Arctic and Southern Ocean sea surface temperature data as seasonal sea ice melts, while GISS deletes sea surface temperature from areas where there is seasonal sea ice. The treatment of polar data by GISS was discussed in GISS Deletes Arctic And Southern Ocean Sea Surface Temperature Data, which was also co-posted at WattsUpWithThat, GISS Deletes Arctic And Southern Ocean Sea Surface Temperature Data. So, due to those differences, this post will only examine the global temperature data between the latitudes of 60S and 60N. These latitudes represent approximately 85% of the surface area of the globe.
Note: It is also known there is little to no Antarctic land surface temperature data prior to the 1950s. But this can’t explain the results Frank Lansner was reporting.
GISS and Hadley Centre combined (land and sea) surface temperature anomaly data were downloaded from the KNMI Climate Explorer:
The KNMI Climate Explorer was also the source of Hadley Centre sea surface temperature (SST) anomaly data (HADSST2) used in its combined product. It also served as the source of the two SST components of the GISS combined product, HADISST from January 1880 to November 1981 and Reynolds (OI.v2) from December 1981 to present. The method employed to merge the two SST datasets used in the GISS product is discussed under Step 4 of the GISS current analysis webpage. The base years (1982 to 1992) used for splicing are different than those presented by the KNMI Climate Explorer for the GISS combined product, so I shifted the merged SST anomaly data to account for this.
I wanted to compare NCDC data in this post also, but it is not available through the KNMI Climate Explorer, and since the NCDC does not break down its standard combined surface temperature product into the desired latitude band (60S-60N) on its Global Surface Temperature Anomalies page, I used a second source. The NCDC also has SST, LST, and combined temperature anomaly data available through its ERSST Version 3/3b webpage, and it is available in multiple latitude bands, including 60S-60N. Scroll down to their link ftp://eclipse.ncdc.noaa.gov/pub/ersstv3b/pdo under the heading of “ASCII Time series Tables”.
Figure 1 compares the combined global (land and sea) surface temperature anomalies of the standard NCDC product and the data available through the ERSST.v3b webpage. The trends are identical at 0.057 deg C/decade. The difference appears to be caused by the use of different base years. The standard NCDC product uses 1901 to 2000 for base years while the data available through the ERSST.v3b webpage appears to be based on the NCDC climatology of 1971 to 2000. So this post uses the NCDC SST and combined (LST and SST) data that’s available through the ERSST.v3b webpage:
Links to the data presented in Figure 1:
“Standard” NCDC Global combined product:
NCDC Global combined product through ERSST.v3b webpage:
Note: There are two different series of data available through the ERSST.v3b webpage. Those ending with .gv3.asc are recent additions, and since they have a slightly different trend and I have not found any mention of them in any other webpage, I have not used them in this post.
NON-POLAR SST AND COMBINED (LST & SST) SURFACE TEMPERATURE ANOMALY COMPARSONS
Figures 2 through 4 are comparison graphs of GISS, Hadley Centre, and NCDC SST and Combined (SST&LST) datasets for the latitudes of 60S-60N. All data has been smoothed with a 13-month running-average filter. They are being provided as references.
CONTRIBUTION OF LAND SURFACE TEMPERATURE ANOMALY DATA TO NON-POLAR SURFACE TEMPERATURES
As discussed in the overview, to determine the contribution of land surface temperature anomaly data, the SST anomaly data was subtracted from the combined data. Before subtraction, the SST data was scaled by a factor of 0.755 to represent the ratio of ocean to land between the latitudes of 60S-60N. The remainders for each dataset are shown in Figures 5 through 7. Note that these graphs represent the remainder of subtraction, not the actual land surface temperature anomalies. To convert them back to land surface anomalies, they would have to be scaled.
COMPARISON OF LAND SURFACE TEMPERATURE CONTRIBUTIONS
Figure 8 compares the remainders resulting from the subtraction of the scaled SST data from the combined (LST&SST) GISS, Hadley Centre, and NCDC products. As shown in Figures 5 through 7, the land surface residuals are noisy, so for this comparison, the data was smoothed with a 37-month running-average filter. While there are slight differences in the yearly and decadal variations in Figure 8, the linear trends for the three datasets are basically the same, differing only 0.001 deg C/decade.
This suggests, for the latitudes of 60S-60N, the differences between the combined products from GISS, Hadley Centre, and NCDC result from differences between the SST data they employ. Refer to An Overview Of Sea Surface Temperature Datasets Used In Global Temperature Products. And the most significant differences in SST anomalies occur before 1940. From 1880 to 1940, the SST anomaly data used by the Hadley Centre and NCDC have significant dips and rebounds, as shown in Figure 3 and 4, while the dip and rebound is much less pronounced in the SST data used by GISS, Figure 2.
SO HOW DOES GISS LAND SURFACE DATA DIFFER FROM THE HADLEY CENTRE AND NCDC?
The Hadley Centre and NCDC land surface temperature anomaly datasets represent continental land masses only, and on a global basis, both of those datasets exclude Antarctica. The land surface data presented by GISS, on the other hand, includes continental land mass data plus much more. First, looking at Figure 9, continental land mass data is extended out over the oceans in the GISS land surface temperature product with 1200km radius smoothing. (Figure 9 is a trend map available through the GISS Global Maps webpage, and it shows the regional changes in temperature anomaly from 1880 to 2009. I’ve cropped the map to show the latitudes, 60S-60N, used in this post.)
Second, GISS also includes data from island surface stations and from “Ship stations,” and these values are also extended out over the oceans. This GISS dataset is a carryover from the methods developed by GISS back in the 1980s, when SST datasets were incomplete. They were attempting to simulate global temperature anomalies without using SST data. This is explained further in a WUWT comment from Zeke Hausfather to Frank Lansner in which Zeke quotes from a correspondence from Dr. Reto Ruedy of GISS:
In part it reads, “The curve NCDC and most likely you are computing shows the mean temperature over the land area (which covers about 1/3 of the globe, a large part of it located in the Northern hemisphere).
“None of our graphs represents that quantity. We could obtain it by creating a series of maps, then averaging just over the land areas (similar to what we do to get the US graph).”
It continues, “Since our interest is in the total energy contained in the atmosphere which correlates well with the global mean surface temperature, all our graphs display estimates for the global mean, the ones based on station data only as well as the ones based on a combination of station and ship and satellite data. Obviously, the latter is the more realistic estimate and we keep the first one mostly for the following historical reason:
“When we started out in the 1980s analyzing available temperature data, historic ocean temperature data were not yet available and we did the best we could with station data. As soon as ocean data compilations became available, we used them to refine our estimates (calling it LOTI). But we kept the earlier estimates also, mostly for sentimental reasons; they are rarely if ever mentioned in our discussions (see also the ‘note’ in the ‘Table’ section of our main web site).”
And continuing this post, the “‘note’ in the ‘Table’ section of [the GISTEMP] main web site” reads, “Note: LOTI provides a more realistic representation of the global mean trends than dTs below; it slightly underestimates warming or cooling trends, since the much larger heat capacity of water compared to air causes a slower and diminished reaction to changes; dTs on the other hand overestimates trends, since it disregards most of the dampening effects of the oceans that cover about two thirds of the earth’s surface.
And again, LOTI represents the GISTEMP combined land and sea surface data and the dTs represents the land surface data.
The GISS Global-mean monthly, seasonal, and annual means dataset does not represent continental land mass temperature anomalies as many believe. It, therefore, cannot be employed in analyses like the one Frank Lansner is attempting to perform.
Dr. Ruedy’s statement that GISS could create a continental land temperature anomaly dataset similar to NCDC “by creating a series of maps, then averaging just over the land areas,” is another way of saying they could create it by masking the areas where the land surface data extends out over the oceans. And this was noted in the post Notes On The GISTEMP Ratio Of Land To Sea Surface Temperature Data.
Then there’s the similarity in the linear trends of the land surface contributions for the three combined datasets, Figure 8. It confirms the findings of the independent researchers who are creating land surface temperature anomaly datasets: the results are pretty much the same as the GISS, Hadley Centre, and NCDC data.