However, if one simply invalidates the years 1993 and 1994 for Pinatubo and 1998 for super El Nino, which is not a totally unreasonable thing to do, then a single linear trend fits better than a step change, as I recall.

So, the data are not that robust against censoring of extreme residuals known/believed to be connected to a physical process.

I suspect you may be seeing the same sort of thing with UAH/RSS.

Rich.

Now, the difference between the two is 0.13 C (for Feb 09), i.e. a difference of more than a tenth of a degree C. Hence, why would they claim results with a precision of one thousands of a degree?

More appropriate statements would be:
- UAH: Global average temperature anomaly for Feb 09 is approx 0.4, +/-0.2 (or more)
- RSS: Global average temperature anomaly for Feb 09 is approx 0.2, +/-0.2 (or more)

The very strong signal from these data sets is of a highly significant positive coefficient, which we all can derive via standard or exotic (R!) software. The UAH and RSS sets are very similar, which is reassuring.

However, as I have written in other postings, I am of the opinion that fitting a simple linear model to long and possibly complex data sets such as these and expecting to derive something really sensible and insightful is very optimistic. Of course, the highly suggestive coefficients generate a sense of security. There can be no doubt whatsoever as to their statistical significance. Whether or not they are of real practical value or import is, I suggest, open to question.

What I advocate is a careful appraisal of the data to see whether a simple linear model over the whole time span of the data is indeed a sensible approach. In the current case I am sure that it is not.

What I’d recommend to those who like to dissect data is that you do exactly that, and fit a three part composite linear fit. Here are some recommendations: Fit the Global UAH data from 1978 to 1993 to a linear model, and repeat the process for 1994 to 1997 and then for 1998 to Feb 2009. What you will find is that the time coefficient for each of these periods cannot be distinguished from zero, but that the mean values are very different. What has happened is that the data underwent remarkably clear step changes. Next, repeat the process for the RSS data (I worked with data averaged over the full set of locations to cut down the amount of work). You will find the same step change scenario.

So, what has been going on? I have no idea, unfortunately, but the evidence for these abrupt changes is really quite strong. Something must have caused them. The possibility of instrument problems is remote, since UAH and RSS work independently. The remarkable agreement between the two sets is strongly suggestive of two very reliable technologies both detecting rapid changes in the signals they are receiving, which occur virtually at the same time.

If you are wary about choosing “arbitrary” subsets of the data, just consider what observers carrying out analyses in “real time” would have found. They would have been unable to find significant changes over the time periods I’ve suggested (or similar time periods), but if their data sets included the points of abrupt change they would have found substantial (positive) coefficients. This is exactly what fitting a long term straight line does.

I suggest that these ideas are worth following up by other readers of this blog, and I look forward to reading what you have made of it.

What I’ve found is evidence for the following abrupt changes (approximate values):

_____________UAH/Global ____ RSS/Average

Late__ 1993 _____+ 0.09 _____0.13
End___ 1997 ____ + 0.18 _____ 0.20

Some formatting difficulties -sorry

This is the briefest summary I can think of. I can supply the rest of the stats stuff if anyone wants it.

Robin

Thank you for the lead-in reference. I’m digging deeper and becoming more fascinated by the capacity of the early satellites to produce large errors, mainly because there are sometimes many indirect steps in some procedures and the errors from each are not always combined. I’ll give examples later this month.

OTOH, they did some magnificent work for their primary missions.

I’m working on a second post focusing on monthly trends. The UAH trends vary much more than RSS and the surface data sets. Here’s a preview:

In recent years, UAH has shown much higher anomalies in February than in later months (e.g. May), leading to a wide divergence of monthly and seasonal trends.

Eric,

Why not stick with just

following the science

No amount of money or political connectedness can ultimately triumph over the science. To me, (an individual not connected to anything), it is disturbing that there are serious and traceable flaws in the established science that drives the political discussion. The most glaring one being the absence of the medieval warm period and little ice age from the famous hockey stick temperature reconstruction. This does not mean that we should not take the issues seriously. However, the climate science community must do some vetting of it’s knowledge base. Currently, this vetting is not happening for fear of loosing public support for the measures that are thought necessary to avert climate change. However, this is extremely unwise. The measures necessary to change the energy track of the world have a large impact. We are not going to convince the Chinese unless we have science that can be audited in the way that Steve McIntyre is trying to do here.

Will

UAH is a compilation of temperature records from satellite data, most recently from the instruments aboard the Aqua (and Terra) satellites.

Taking Earth’s temperature
Describes Advanced Microwave Scanning Radiometer (AMSU) aboard NASA’s Aqua satellite. References microwave sensor aboard TIROS satellite (1978). MSR apparently looks at the intensity of microwaves emitted by oxygen molecules. In space, each microwave sounding unit self-calibrates every cycle. Describes daily calibration and steps to remove four sources of error.

Spencer, Roy, PhD, and Dr. John R. Christy. “Taking Earth’s temperature .” Educational. UAH News: Your Official UAH News Source. http://www.uah.edu/News/climatebackground.php.

Science Writers’ Guide To Aqua
Aqua’s Instruments (duplicated on TERRA)
The Aqua satellite has six instruments onboard to measure and monitor Earth’s hydrologic cycle.
These include (pages 3-4):
• AIRS (Atmospheric Infrared Sounder)
• AMSR-E (Advanced Microwave Scanning Radiometer – EOS)
• AMSU (Advanced Microwave Sounding Unit)
• CERES (Clouds and the Earth’s Radiant Energy System)
• HSB (Humidity Sounder for Brazil)
• MODIS (Moderate-Resolution Imaging Spectroradiometer)

Measurements (Page 5):
Note: ” | ” represents columns in the document’s table. Far left column represents “region” of interest.
Region | Measurement | Instrument(s) Used
Atmosphere | Aerosol Properties | MODIS, CERES
| (Composition, Size, Distribution)
| Atmospheric Humidity | AIRS, AMSR-E, AMSU, HSB, MODIS
| Atmospheric Temperature | AIRS, AMSU, MODIS
| Cloud Properties | MODIS, CERES
| Greenhouse Gases | AIRS
| Precipitation | AIRS, AMSR-E, HSB
| Radiative Energy Fluxes | AIRS, AMSR-E, CERES,
| (Emitted Thermal and Reflected Solar Radiation) | MODIS
Cryosphere | Sea Ice | AMSR-E, MODIS
| Snow Cover and Depth | AMSR-E, MODIS
Land | Fire Occurrence | MODIS
| Land Cover and Land Use Change | MODIS
| Surface Temperature | AIRS, AMSR-E, MODIS
| Surface Wetness | AMSR-E
| Volcanic Effects | MODIS
Ocean | Ocean Color | MODIS
| Phytoplankton and Dissolved Organic Matter | MODIS
| Sea Surface Temperature | AIRS, AMSR-E, MODIS
| Sea Surface Wind Speed | AMSR-E

Gutro, Rob, Krishna Ramanujan, and Jim Closs. “Science Writers’ Guide To Aqua.” NASA Goddard Space Flight Center, March 31, 2002. http://eospso.gsfc.nasa.gov/ftp_docs/Aqua_Sci_Writer%27s_Guide.pdf.