Sunday, April 1, 2012


date: Fri, 28 Feb 1997 10:39:59 +0000 (GMT)
from: "N.W.Arnell" <>
subject: Re: thresholds paper..
to: (Mike Hulme)


How likely is tolerable climate change...

I think this is a very good - and revealing - paper. I like the
approach taken, and methods proposed. My comments fall into
two areas.

1. Signal/noise ratios and defining climate change scenarios

I very much like the comparison of 30-year time slices from the HADCM2
transient run. This is exactly the sort of information I was seeking at
the DoE Global meeting. Have you done it for other seasons, or for
months or locations? I guess the problem is even worse with decades!

The variability you show emphasises the dangers of simply comparing two
n-year periods or, more worringly, comparing results from scenarios
based on different periods (eg rescaled 2080s compared to 2020s). The
only solution appears to lie in the use of ensembles...

I'm not so concerned about the similarities (or otherwise) between
future changes and past n-year variability. In practice, most impact
assessments use (or should use) a well-defined baseline period, and it
is change relative to that which is important, even if it is small
(but systematic). Managers are planning for the future, not the past,
although an idea of future vs past variability would give them some
clues about their ability to manage future changes. One significant point
that you have highlighted, however, is that changes in estimated
system performance due to climate change may be of a similar magnitude
to changes which appear simply because different data periods are used
(as a footnote, water companies in Britain are currently reviewing their
system yields, supposedly using consistent baseline data - they are
in for some surprises!!).

2. Identification of critical thresholds

Although thresholds appear attractive in water management and are reasonably
easy to define (eg implementation of a hosepipe ban / water level in a
reservoir below 10% of usable storage...), it can be very
difficult to work out or generalise the risk of them being exceeded. In
water resources terms, for example, it is not just summer temperature and
precipitation that is important, but rather the state of storages at the
beginning of the summer season: these are an integration, in Britain, of
climate over winter. Unfortunately, the winter-end/summer start storage can
depend very much on how climate was distributed over winter. It is not
just the rainfall total that is important, but also (although probably to
a lesser extent) how it was distributed over the winter. A full assessment
of the risk of threshold exceedance would therefore be multivariate, with
axes defining, for example, cumulative winter net rainfall, the distribution
of that rainfall through winter (eg December total as % of the DJF total??),
summer precipitation and summer temperature. The "danger zone" would
be multi-dimensional.

...but this is a quibble, and perhaps reflects problems peculiar to the water
system, which is by nature integrative over time.

I promised Ferenc Toth that I would prepare a paper on thresholds and water -
even though I had to withdraw from the workshop at the last minute....! - and
I am attempting to work some of these issues up. I will try to include some
numerical examples. I'll let you see a draft once I have made some proper


Nigel Arnell

No comments:

Post a Comment