cc: tom@ocean.tamu.edu
date: Mon Aug  7 14:14:19 2000
from: Phil Jones <p.jones@uea.ac.uk>
subject: Re:
to: "Michael E. Mann" <mann@holocene.evsc.virginia.edu>, Rob Harris <rnharris@mines.utah.edu>


 Dear Rob,
    Mike's forwarded your email to him to me. The diagram on the web site
 is for illustrative purposes. All we have done, as alluded to by Tom, is
 to regress each series against a common instrumental series (the NH average
 for land areas N or 20N, for the April-Sept average). As this is simple
 regression it doesn't change anything in a series, except changing the level.
 It also brings each series into better accord with each other. Regression
 for each series is over the period 1881-1960 except for Tom's odd periods
 given in his Science and Ambio papers.
    I am producing another plot where I compare Mike's and Tom's series
 instrumental calendar year temps and another where I compare the growing
 season series.
    None of the proxy compilations are complete over 1961-90, but this
 regression with an instrumental series which is with respect to 61-90
 brings them all to this level. You can get almost the same result with
 Mike's if you just subract 0.12 from each value. This is the offset between
 Mike's base period and 1961-90. The regressions's have
 their biggest offset with Tom's series - made worse by the fact that on
 the web page we used Tom's series without two series (Sargasso and a Lake 
 core in the central US). Using Tom's full series with these two brings
 Tom's series back to be more like Mike's. We will correct this, but
 everyone is away at the moment.

    I hope this explains the problem. Comparing the series with an instrumental
 record means you have to adjust one to the other. We adjusted the proxy
 series and used a common instrumental series for each. By the way the
 correlation between this NH series and the NH land+marine series for the
 whole NH ( ie land N of 20N for April-Sept and the whole NH for the calendar
 year) is over 0.95.

 Cheers
 Phil

