Guest post by Bob Tisdale
Date: August 16, 2012
Subject: Southern Hemisphere GISS LOTI Land Surface Temperature Anomaly Data
From: Bob Tisdale
To: James Hansen – GISS
Dear James:
I discovered what appear to be an atypical upward step and a recent abnormal increase in variability in the land portion of the GISS Land-Ocean Temperature Index (LOTI) dataset for the Southern Hemisphere. I found this yesterday while preparing one of the final chapters of my book Who Turned on the Heat? The Unsuspected Global Warming Culprit, El Niño-Southern Oscillation.
I used the ocean mask feature of the KNMI Climate Explorer to isolate the Land Surface Temperature portion of the LOTI data for the Southern Hemisphere, then smoothed it with a 13-month running-average filter. Since I’m using the satellite-based Reynolds OI.v2 sea surface temperature dataset as the primary source of data for my book, the graph starts in November 1981, and the GISS LOTI data through the KNMI Climate Explorer was only available through March 2012, and that explains the end month. I used the base years of 1982 to 2011 for anomalies to try to minimize any seasonal components. As shown in Figure 1, there appears to be an upward shift in the data during the 1998/99/00/01 La Niña, around the year 2000. The timing of the shift does not agree with what would be an expected response to a major ENSO event.
Figure 1
Comparing the Southern Hemisphere LOTI data without the ocean data to scaled NINO3.4 sea surface temperature anomalies as a reference for the timing of ENSO events, and shifting the NINO data upwards by 0.3 deg C after January 2000, helps to highlight the upward step I was seeing. See Figure 2.
Figure 2
I found this odd, to say the least, so I started looking for explanations. I checked to see if there was a problem with the land-mask feature of the KNMI Climate Explorer. I had never encountered one before, but I checked anyway. Figure 3 compares the GISS land-only surface temperature anomaly data (with 250km smoothing) to the LOTI data with the ocean data masked. Both datasets show the unusual rise.
Figure 3
I checked NOAA’s GHCN and the UK Met Office’s CRUTEM3 land surface temperature anomalies for the Southern Hemisphere, and they did not display the shift, as shown in Figure 4.
Figure 4
When I compared the four versions of the Southern Hemisphere land surface temperature anomalies, Figure 5, a few other things stood out. It appears the two GISS datasets pick up an additional warming trend after 2000 that does not exist in the GHCN and CRUTEM3 data, and the two GISS datasets appear to have much greater year-to-year variations in recent years.
Figure 5
Regarding the trends, Figure 6 shows the GISS LOTI data for the Southern Hemisphere, with the ocean data masked, for two periods: November 1981 to December 1999 and January 2000 to March 2012. The trend for the GISS data starting in January 2000 is 3 times greater than the earlier period. But if we look at the average of the GHCN and CRUTEM3 data, Figure 7, the trend from January 2000 to present is considerably less than the earlier period.
Figure 6
HHHHHHHHHHHHH
Figure 7
The GHCN and CRUTEM3 datasets include less data in Antarctica, and of course, they don’t have the 1200km smoothing employed by GISS to infill areas with missing data. So for the next two graphs, Figures 8 and 9, to try to isolate the cause, I excluded all land surface temperature data south of 60S, to remove the Antarctic data. The GISS data still has a higher linear trend during the latter period, while the average of the GHCN and CRUTEM3 data continues to warm at a lesser rate after 2000.
Figure 8
HHHHHHHHHHHHH
Figure 9
Could the 1200km smoothing be the cause? I wouldn’t think so, but I checked. Figure 10 shows the linear trends for the two periods using the GISS land-only temperature data (250km smoothing) for the Southern Hemisphere, and, as you’ll note, I’ve excluded the Antarctic data. The period starting in January 2000 has a much higher trend than the earlier period. That’s really odd since the GHCN and CRUTEM3 data in Figure 9 should be using data similar to, if not the same as, GISS, but their trend is less during the later period. And as you’ll note, the GISS trend for the period starting in 2000 is about 2.8 times higher than the average of the other two datasets.
Figure 10
That made me wonder if the coverage of the GISS land-only temperature data (250km smoothing) was in fact similar to the GHCN and CRUTEM3 datasets, so I used the map-making feature of the KNMI Climate Explorer to run a quick comparison of spatial coverage. I set the map type so that it would display grids and areas where data existed, and I set the contours so that the grids and areas wouldn’t offer any distractions by changing colors. I plotted the Southern Hemisphere for each January starting in 1982, and animated the sequence of maps. This was just a cursory look. It appears that GISS excludes grids or stations in the Southern Hemisphere sooner than the other two and GISS seems to have less coverage than GHCN and CRUTEM3 as time progresses. You may need to click on the animation to view it.
Animation 1
James, you should look into this matter. I don’t have the time or the inclination to carry this investigation any further.
Some persons might think GISS has been manipulating data to acquire a higher land surface temperature anomaly trend in recent years. They also might assume GISS has been reducing coverage in recent years to create a little more variability, thereby increasing the chances for new record temperatures with each El Niño. And the way that all suppliers of temperature data appear to use data for a grid one year but not the next, and then have data for that grid reappear a year or two later, may lead some persons to think data is being cherry picked for use. We wouldn’t want people to think those things.
Final note: As you know, GISS, in effect, deletes sea surface temperature data in areas of seasonal sea ice and replaces it with much-more-variable land surface temperature data. This, of course, creates a warming bias at the poles in the GISS data. Refer to the zonal-mean graph in Figure 11 that compares the linear trends of the Reynolds OI.v2 data and the version of it with the GISS modifications, for the period of January 1982 to October 2011. It’s from my most recent post that discusses this subject: The Impact of GISS Replacing Sea Surface Temperature Data With Land Surface Temperature Data.
Figure 11
Because of that monumental bias, when I present GISS Land-Ocean Temperature Index data, I usually limit the latitudes to exclude polar data. Now, with this find in your land surface temperature data, I’ve had to switch to an average of the GHCN and CRUTEM3 data for that chapter of Who Turned on the Heat? The Unsuspected Global Warming Culprit, El Niño-Southern Oscillation.Sorry to say, but with all of the biases toward warming, your GISS LOTI data, in my eyes, is becoming more and more unsuitable for research.
Sincerely,
Bob Tisdale
SOURCE
The data and the maps used in this post are available through the KNMI Climate Explorer.
Ah, perhaps I should have looked at your graph before. I’m talking about individual nights, you’re talking about month-long averages.
At any rate, I’m more intrigued with the down step in water content than the clouds. Less humidity and less low level cloudiness suggest more clear nights, and more efficient cooling. (Dew fall slows the rate of temperature decline because of the heat released while condensing.) Not really what Bob’s graphs show, but so be it. Unfortunately, I don’t have time to check data from around the world.
Drier conditions could lead to a delay in cloud formation and higher cloudbase. Those could leard to warmer Tmax values. Don’t have time to check that either.
Darren Potter says:
August 17, 2012 at 2:01 pm
Right. My comment (thought) was directed at NASA. The author of the NASA statement just might be trying to admit the truth, that the data is unjustifiably manipulated, without actually saying it.
At any rate, I’m more intrigued with the down step in water content than the clouds. Less humidity and less low level cloudiness suggest more clear nights, and more efficient cooling.
Anthropogenic aerosols seed more persistent clouds. That is, clouds that take longer to precipitate out. Reducing these aerosols would decrease these persistent clouds and hence reduce the atmospheric water content. It also explains the reduction in low level clouds and the increase in higher level clouds. Water vapor that would have formed aerosol seeded clouds now migrates higher in the atmosphere before forming clouds.
Which doesn’t explain the step down in 1998. I’ve suggested before that the Russian Financial Crisis in 1998, which resulted in the shutdown of much of the aerosol polluting Soviet era industry, was the cause, but have no evidence to back this up.
Which is why I’d like to see separate NH and SH versions of that graphic.
Sam Yates says: “Hm. This is interesting, but I have a question about figure 11. Is it dealing only with summer temperatures (for the respective hemispheres, of course), or is it year-round?”
That’s year round.
Sam Yates says: “Even with the land masked, you know, those two charts are measuring two different quantities: sea surface temperature and air temperature.”
Yes and no. The Reynolds data is sea surface temperature data. The GISS Land-Ocean Temperature Index data with the land masked is sea surface temperature data from about 52S to 52N until the data diverge. North of 52N and south of 52S it transitions to land air temperature.
Sam Yates says: “In fact, come to think of it even if it is only seasonal data you’d still expect to see a deviation between the two graphs, because even during the austral and boreal summers the sea ice doesn’t melt completely, so air temperatures over the ocean are STILL going to deviate from ocean temperatures–just not as much.”
Depends on the hemisphere, doesn’t it? And in the Arctic, with open waters, GISS should not be able to extrapolate land surface data across the open ocean to the remaining sea ice. Consider also where the sea ice remains near the coast. There’s the change in albedo when the snow near the weather stations melts and leaves ground cover. Would the albedo at the stations be similar to the albedo of the sea ice then? GISS shortcuts the process.
Regardless, as I noted in the post, because they replace sea surface temperature data with land surface temperature data, I rarely use GISS polar data because of the seasonal bias.
Bob:
There is this little thing about testing a theory against other known databases, which you did. The mere fact that the “jump” doesn’t not appear in those data sets should be a good indicator of the trustworthiness of those questionable data sets.
This is not the first time Mr Hansen has been caught with his hands in the cookie jar and manipulating data. The fact that he is doing it to the satellite data, which was up until recently was disproving CAGW, creates problems for their continued implementation of draconian control over the populace. HE who controls the Oil controls the world.
Excellent post Bob… just another pin in the hot air filled balloon of CAGW.
I have been trying to conceive of some physical mechanism that would hold the Great Lakes at a new higher equilibrium after a single step shift, and I am completely unable to do so. The lakes are fed and flushed by freshwater sources, and unless said sources (rivers) were also suddenly warmer and (more or less) stably so, one would have expected gradual “reversion to the mean”. I suspect data corruption; it is the only available consistent possible source of the consistency!