Climate Cryptography

Here's an intentionally provocative post. Suppose you have temperature data T(x,t) over space x and time t, and the slope of your regression line indicates an average rate of increase of 0.6 degrees per 100 years. Suppose you need to allow others to verify that 0.6 number, but for whatever reason, you would not like to give out the raw data T. What to do? Choose a "private key" g(x), a function of space alone, which you keep secret. Publish and release to the public only "temperature anomaly" data T1(x,t) = T(x,t)-g(x).
http://lwf.ncdc.noaa.gov/oa/climate/research/anomalies/anomalies.html http://www.cru.uea.ac.uk/cru/data/temperature
You can claim that T1 is a normalization of the data, accounting for local climate variations, or some such. It's clear that T can not be recovered from T1 without g. So your data is safe. However, it's also clear mathematically that the regression line through T1 will yield that same 0.6 number you obtained. Cryptography has allowed you to keep your data private, while permitting others to "verify" your results!

Update 9/8/2008 : more detail.