I’ve been looking at the Steig stuff today and I intend to put up a new post on the topic (or update the old one) in the next day or so. I think it’s more or less figured out.

Re: Peter (#48), Thanks for the information, but it wouldn’t help the readers much. I’m also not sure that it would interact properly with the front end of R as well.

“remember not to accidently copy something else into the clipboard”

Suggest you install a clipboard extender:
href=”http://www.google.com/search?hl=en&q=Clipboard+Extender&btnG=Search”>

You write

Precision .. can be thought of as measuring to a high precision and then rounding the results to the nearest degree… Assuming that there is no bias in the instrument, this will usually have the result of increasing the variability of the values thus requiring a larger number of observations to achieve a given level of precision. If the instrument is poor, the number of observations needed to estimate a result with a small error bar could be very large.

1) But in climate studies, we’re not observing the same thing more than once, so how can the above statement apply?
I.e. if we measure the temp in N different locations, we do not have N samples of 1 thing, we have 1 sample each of N different things (i.e. with different estimated values and distributions).
Likewise, if you measure the temp M times (in the same location), separated in time (let’s say a day between each sample), again you do not have M samples of the same thing, you have 1 sample each of M different things (i.e. with different estimated values and distribitions).

2) Furthermore we do indeed have biases, limited precision, and random measurement errors, for every instrument, every where, every time.
How is that taken into account?

(As can be seen, I do not understand how these .322 average global temperatures (deviations) can make any sense whatsoever.)

3) Simple example
Assume that the measured temps (at some point in time) in 7 different locations are 0, 1, 3, 2, 4, 1, and 0.
Assume further that each thermometer has an in-precision of 0.5, and that there is no systemic error (bias), nor any other measurement error, in any of the thermometers.
i) What is the average temperature at this particular time? (2? 1.6? 1.57? 1.571?)
ii) What is the (estimated) error, taking also instrument in-precision into account?

If and when you get some time…

Thanks.

Thank you. I shall raise the topic of error propagation with Steve when he’s ready, the correct procedure as you note. However, I would infer that you have a lot to contribute and I hope that you do if it happens.

The lecture you gave me differs in no material aspects from ones I gave to staff 30 years ago. But, there has been an observable change in climate “science”. That change can briefly be summarised as obtaining confidence intervals that are an order of magnitude or more better than reality, while failing to authenticate their derivation. The start of this error obfuscation is the omission of calibration data.

Given the satellite involvement in the Steig paper, one can study satellite errors from literature. If you wish to give yourself a fright, try

http://landsathandbook.gsfc.nasa.gov/handbook/handbook_htmls/chapter8/chapter8.html … and …
http://landsathandbook.gsfc.nasa.gov/handbook/handbook_htmls/chapter8/htmls/cal_actions.html

Although this is specific to Landsat, the scope for errors as decribed (but not individually quantified) is huge. To quote,

A major objective of the Landsat-7 program is to upgrade the radiometric quality of the data to be commensurate with the other sensors in the Earth Observing System (EOS). Unlike its predecessors, a specific goal of the Landsat-7 program is to achieve radiometric calibrations of the data to ± 5% uncertainty over the 5 year life of the mission. Pre-launch, the mission design supports this requirement through hardware design changes, and instrument characterizations. Post-launch or on-orbit, this 5% requirement is supported by a monitoring and calibrations program, and the implementation of any necessary changes to the ground processing of the data.

From whence does one contrive a 0.1 deg C warming increment? Maybe it just rolls easily off the tongue, as does the so often seen +/- 5%. This is more the type of error analysis to which I was referring. And this is only the tip of the iceberg, so to speak.

Your link to #328 took a little searching since it links to this page. My guesses for Byrd and Siple would be series 817 and 324 of the satellite recon (which were simply calculated by minimizing the sum of the absolute difference between the lats and lons). I located more accurate co-ordinates for Byrd (80.007S, 119.404W), but you may be right that extra rounding could move things around.

I don’t have the time right now to do any more with it right now, but I will try to look further at things later.