A comphrehensive comparison of GISS and UAH global Temperature data

I originally started this comparison by looking at the differences between OI.v2 SST and the UAH MSU Lower Troposphere Temperature (TLT) anomalies for the same ocean segments. Since OI.v2 SST data has been used by GISS for their GISTEMP product since December 1981, I decided to add another comparison: GISS Land Surface Temperature (LST) versus UAH MSU TLT for the same continental land segments. Then I added one last comparison, which is the subject of this post.

Note 1: The data illustrated in the following graphs are as I downloaded them from the KNMI Climate Explorer website. I made no effort to offset either dataset in the comparative graphs so that the two curves rested on one another. The graphs will show that GISTEMP anomalies are higher than UAH MSU TLT anomalies. This is a function of base years. Focus on the trends and the shapes of the curves, not the location of the curves.

Figure 1 is a comparison of Global GISS Surface Temperature (GISTEMP) and UAH MSU Lower Troposphere Temperature (TLT) anomalies. Both datasets have been smoothed with 12-month running-average filters.

http://i41.tinypic.com/34ryski.jpg

Figure 1

Similar graphs always create speculative comments about the basis for the differences between GISS and UAH data. In this post, I’ve segmented the globe, Figure 2, to locate the areas with the largest differences, in an effort to narrow the possible reasons for those divergences. The coordinates used are listed on the graphs. I’ve plotted the data and added the linear trends, but I have not speculated about the causes for the differences in the data for the smaller global areas.

http://i40.tinypic.com/511opd.jpg

Figure 2

Note 2: GISTEMP data through the KNMI Climate Explorer is available with 250 km and 1200km smoothing. The graphs in the post use the 1200 km smoothing, which is the smoothing presented by GISS in their GISTEMP product. Figure 2, however, is the May 2009 GISS Global Temperature Anomaly map with 250km smoothing. Grey areas indicate locations with no data. These are the areas infilled by the 1200 km smoothing.

Note 3: Also keep in mind that the MSU TLT data reaches to 82.5N and 70S. The approximate locations of those latitudes are shown in Figure 2. UAH also fills in the polar data. On the other hand, MSU data has better global coverage in other areas where surface station data is lacking.

Note 4: And for the last note before looking at graphs and EXCEL-calculated trends, keep in mind that GISTEMP and UAH MSU TLT represent datasets made up of different variables. GISTEMP is composed of Sea Surface Temperature (SST) and of Land Surface Temperature data based on surface station readings. The UAH MSU TLT data represents the temperature of the lower troposphere.

COMPARISONS

Figure 3 illustrates GISTEMP and UAH MSU TLT anomalies for the Arctic, 65N to 90N. The GISTEMP linear trend for the period is 0.0595 deg C/decade while the UAH MSUTLT data has a linear trend of 0.0461 deg C/decade. Note how the GISS data exaggerates (or the UAH MSU data suppresses) the variations, especially after mid-2004.

http://i43.tinypic.com/1zp1q8j.jpg

Figure 3

The North America Plus datasets, Figure 4, also include the Eastern North Pacific and the majority of the North Atlantic. The trends are significantly lower than the Arctic datasets, as would be expected. The linear trend for the UAH MSU TLT data (0.0185 deg C/decade) is greater than the trend for the GISTEMP data (0.0159 deg C/decade).

http://i43.tinypic.com/4in94x.jpg

Figure 4

The South America Plus datasets, Figure 5, also show a UAH MSU TLT linear trend (0.063 deg C/decade) that is higher than the GISTEMP trend (0.05 deg C/decade). These datasets also include major portions of the eastern South Pacific and western South Atlantic. Both linear trends are again significantly lower than the North American Plus datasets. Note the dominance of the ENSO signal in the South American Plus data.

http://i42.tinypic.com/ohpnb5.jpg

Figure 5

The Europe Plus datasets show the highest trends of those examined in this post. This should be due to the impact of the North Atlantic on Europe. As illustrated and discussed in my post “Putting The Short-Term Trend Of North Atlantic SST Anomalies Into Perspective”, the linear trend of the North Atlantic SST anomalies is more than 2.5 times the dataset with the next highest trend. The GISTEMP trend (0.429 deg C/decade) for the Europe Plus dataset is slightly higher than the UAH MSU trend (0.379 deg C/decade).

http://i39.tinypic.com/x29niv.jpg

Figure 6

The difference in linear trends is greatest in the Africa Plus datasets, Figure 7. The GISTEMP linear trend at 0.194 deg C/decade is more than twice the linear trend of 0.093 deg C/decade for the UAH MSU data.

http://i43.tinypic.com/2iszbjt.jpg

Figure 7

For the Asia Plus subsets, Figure 8, the GISTEMP linear trend (0.256 deg C/decade) is also higher than the UAH MSU linear trend (0.179 deg C/decade). The Asia Plus datasets have the second highest linear trends of the areas illustrated in this post.

http://i41.tinypic.com/maudqf.jpg

Figure 8

The comparison of the Australia Plus datasets, Figure 9, illustrates another occasion when the GISTEMP linear trend (0.076 deg C/decade) is less than the USH MSU linear trend (0.096 deg C/decade).

http://i39.tinypic.com/2mrwtja.jpg

Figure 9

The first thing that stands out in the comparison of Antarctic datasets is the difference in the signs of the linear trends. The GISTEMP data show a positive trend of 0.048 deg C/decade, while the UAH MSU data show a negative trend, -0.091 deg C/decade.

http://i42.tinypic.com/an27et.jpg

Figure 10

The Antarctic datasets are also the noisiest of those illustrated in this post. But the real curiosity is the timing of the mid-to-late 1990s spike in the GISTEMP Antarctic data. At first glance, it appears to be a result of the 1997/98 El Nino. But the spike is more than a year early. In Figure 11, scaled NINO3.4 SST anomalies have been added to the comparative graph of Antarctic Plus GISTEMP and UAH MSU TLT data. The spike in the GISTEMP Antarctic data is not a response to the 1997/98 El Nino.

http://i41.tinypic.com/ngqeqq.jpg

Figure 11

Figure 12 illustrates the GISTEMP Surface Temperature and the two components of it: GISTEMP Land Surface Temperature, and OI.v2 SST data for the Southern Ocean. The source of the anomalous spike in the mid-1990s is the GISTEMP Land Surface Temperature data, not the SST data.

http://i40.tinypic.com/64dqft.jpg

Figure 12

Note 5: The GISTEMP Surface Temperature data from 90S to 60S is clearly dominated by the GISTEMP Land Surface Temperature data, though the surface area of the Southern Ocean (20.3 million sq km) is greater than the land mass of Antarctica (14.0 million sq km). This appears to be a function of Southern Hemisphere sea ice area, which can vary from 1.5 to 16.5 million sq km over the course of a year. During the winter, sea ice area increases. The land surface area then becomes greater than the sea surface area, making it the dominant dataset.

CLOSING COMMENT

I do not recall any discussions of a 1996 spike in the GISTEMP Antarctic surface temperature data. I have double-checked to assure I downloaded the data correctly. However, I have not tried to confirm whether or not the 1996 spike occurs in the individual Antarctic surface station data available from GISS:

http://data.giss.nasa.gov/gistemp/station_data/

A gif animation of the annual GISTEMP maps, Figure 13, does show elevated Antarctic surface temperatures in 1996.

http://i39.tinypic.com/2mwdopx.gif

Figure 13

SOURCE

THE GISTEMP Surface Temperature, GISTEMP Land Surface Temperature, UAH MSU TLT, and OI.v2 SST data are available through the KNMI Climate Explorer website:http://climexp.knmi.nl/selectfield_obs.cgi?someone@somewhere