from: Arnell <arnell61atXYZxyznternet.com>
subject: RE: An idea for Science or Nature
to: 'Mike Hulme' <m.hulmeatXYZxyz.ac.uk>
From: Mike Hulme[SMTP:email@example.com]
Sent: 30 June 2000 08:27
I agree that there is some interesting material here. A number of issues
occur to me:
- it needs to be clear that these response curves are unique (?) to HadCM2
regional patterns - different GCMs would yield different response curves if
only because P patterns are different. It would be nice to repeat this
whole exercise with all DDC GCMs and see how far these response curves are
robust, or not, across different GCMs (By the way, I assume we are using 3
different time-slice patterns rather than one end-of-the road scaled
pattern? The latter approach, however, is methodologically sounder).
I agree - I've provided Martin with revised figures. I've rescaled the HadCM2 ensemble mean 2080s scenario to different temperature changes, and applied these to the 2020s, 2050s and 2080s worlds. This should correct for the slice-slice variations in precipitation (which drive the shape of Martin's water response surfaces).
I've attached the spreadsheet for you to see.
- do we fully understand why the shapes of the responses are different? It
is not immediately obvious to me why malaria and hunger are functionally
different from water and coasts - but maybe you've thought about this.
I think the water curves have the shape they have because the precipitation changes from slice to slice. The revised data shouldn't do that. The shape of the coast curves should reflect the relative pace of climate and "baseline" change (because the climate change is smooth). Actually, I would have expected malaria to do something similar to water, because it too is dependent on temperature.
- are we happy using IS92a, a now dated scenario? And of course, although
we have altered the climates due to progressive targets of -10% etc., we
have not changed the underlying worlds in which these impacts will be
played out (i.e., GDP, adaptive capacity, etc.). Are we not learning that
the underlying worlds are just as important for impact as the climate itself?
- and are these response curves with or without adaptation? I guess
actually a bit of both, but it would be nice to show two sets of response
surfaces - one w and and one w/o adaptation.
Not possible in the water sector: it's the old "use/resource" ratio again...
- with regard to 'realistic' emissions targets I have an idea about how to
show a more generic set of sensitivities, but it would take some time to
work-up. The point being that we should not be pretending that we are in
the business of 'realistic' targets (we are not qualified to say), more
that we want to explore the sensitivity of impact to a whole 'range' of
targets - some Annex 1 and some both Annex 1 and II. Thus a 2-D plot of T
change by 2100, overlain with impact response envelopes by 2100, where the
X-axis: date by which reductions achieved, 2010, 2020, 2030, 2040, etc.
Y-axis: size of reduction, -10%, -20%, -30%
We (Xianfu and I) will work up a mock-up of this for next week.
Overall, I just wonder whether we have a watertight case to go to print
now? There is certainly scope for methodological improvement.
I suspect that our response surfaces would vary significantly with climate model: I'm repeating my rescaling exercises with HadCM3 and the DDC, but my results with the "raw" DDC and HadCM results show that the variation is huge (at least in terms of water, which is precipitation driven...).
Some more minor points:
- I would choose different colours - blue for coasts, red for malaria,
brown for hunger, orange for water
- the MAGICC curves should be with *no* aerosols, not just constant 1990
aerosols. But I can re-do these.
See you Monday....
this is my message to Martin....
I've attached a spreadsheet which shows the number of people at risk of increased water stress by 2025, 2050 and 2085, under different temperature increases.
1. I did it by rescaling the HadCM2 ensemble mean 2080s scenario back from 3.1oC (the model estimate of global temperature change at that time horizon). In fact, I rescaled the simulated runoff - I would get different answers (slightly) by rescaling the input climate change and re-running the hydrological model. The temperature-response function is therefore a "conditional" function. I haven't repeated the exercise with any other GCMs, but that would be easy.
2. The "performance indicator" is as before: the number of people living in stressed countries (using more than 20% of their resources) in which resources fall by more than 10%.
3. I've actually done the calculations by country (meaning that big countries hide lots of variation) and by watershed, and both sets of numbers are in the spreadsheet. Arguably the watershed-scale stuff is more accurate (1300 units rather than around 150), but I had to make a few heroic assumptions to downscale the baseline water use to the watersheds.
4. There is a big jump at between 1 and 1.25o change - this is largely due to India (and South Africa, Mexico and Iran) seeing resources reduced by more than 10%. The effect is less obvious at the watershed scale because India is split up into several basins, not all of which suffer.