Runs required
This is largely my summary from Colin Jones's e-mail on 26 August
2014. The runs needs to be ready by the January 2018.
Minimum years of model run time
The minimum number of years that we are required to run the model
for is
- 500 years of pre-industrial
- 160 years of historical run (1850-2010)
- 90 years of the Reprensentive Concentration Pathway (RCPs)
(2010-2100). This includes several runs, including 1% increase in
CO2 and 4 * CO2.
so 750 years.
UKESM1-Hi
Ideal
The sub-models we probably want to run are
- IO server
- ATMOS + JULES (N216)
- Miniature ATMOS + JULES + GLOMAP-mode full chemistry (N96)
- NEMO-CICE (ORCA025)
- Miniature NEMO-CICE + MEDUSA (?)
where each point is a separate executable. At time of
writing (8 September 2012), the miniature versions don't
exist and won't for a while.
Offline oxidants option
Instead of the miniature ATMOS + JULES + GLOMAP-mode full
chemistry, this is
- IO server
- ATMOS + JULES + GLOMAP-mode simplified chemistry (using offline
oxidants) (N216L85)
- Miniature ATMOS + JULES + GLOMAP-mode full chemistry (N96L85)
- NEMO-CICE (ORCA025), and MEDUSA here or extra executable as
above.
where the input for the offline oxidants would be come from a
N96 run.
Further options
If the above options are still too expensive, further options
for the very long pre-industrial run are
- we could use simplified chemistry
- 500 years at N96 + a further 100 years at N216
And further options are
- Use N144/ORCA025 instead of N216/ORCA025. Richard Wood
is particularly against this, because there's much more
experience of N216 than N144, such as more work in tunning
N216. N144 would require a lot of work from Richard's AOIL
team.
UKESM1-Lo
Ideal
The sub-models we probably want to run are
- IO server
- ATMOS + JULES + GLOMAP-mode full chemistry (N96)
- NEMO-CICE + MEDUSA (ORCA1)
Pre-industrial spin up
When determining how much spin-up, it's necessary to consider
the time scales different parts of the Earth System work on.
For a boundary layer this might be of order of a day, whereas
for the deep oceans this is much longer, say of order a 100 years
(?).
It is for the deep
oceans that we require the longest spin-up.
Fortunately, the deep oceans aren't that sensitive to fine detail
in the atmosphere and so don't require 500 year of coupling
with the full UKESM1 at N216.
The spin-up of the ocean is likely to be something like the
stages below
- Take 30 years of data from N96 run (I'm guessing mostly SSTs),
and use this input
repeatedly on a ORCA1 run without MEDUDA for maybe 1000 years.
30 years is chosen,
so that it's
not biased by any one freak year. (I'm not sure why a climatology
is not used, but maybe the variability is important.)
- Attach a N96 run to this and run with MEDUSA for say 500 years
- Increase resolution of ocean to ORCA025, and run for several
hundred years but less than years in previous stage.
- Attach a N216 run to ocean (ORCA025) and run for even less
years
The reality is that the spin-up will probably never be quite as
long as we'd like. So we'll take branches of our spin-up, while
keeping the spin-up going, and use results from the one the
latest branches.