Minutes of the group meeting held in Cambridge on November 2nd 2006
Present: Clare, Erik, Roisin, Meike, Nick, Marie and Corinne.
General introduction and goals
Corinne presented the main objective of the group:
"to understand and quantify the interactions between marine
biogeochemistry and climate (detect hot spots and non-linearities)"
To achieve this, a lot of model developments are needed. The
timeline for model developments need to meet the timeline for climate
model developments if we are going to have an impact in the near
future. In order to be included in the next round of climate model
simulations, our green ocean model component needs to be ready
in summer/fall 2007.
Any single component to be included in future climate simulations
need to show that it causes a change in a climate-relevant variable
by at least 10% globally or 50% locally. The competition for processes
and computer time will be large, so we need to meet that target.
Model developments
Model versions so far:
- PISCES-T
- PlankTOM 5.0 (the Jena NCCL9 version)
- PlankTOM 5.1 (5.0 + DMS cycle, Meike's version)
- PlankTOM 5.2 (5.0 running on NEMO at UEA)
- PlankTOM 6.0 (5.0 + N2-fixers, Nick's version)
- PlankTOM 9.0 (not optimized, running on NEMO at UEA)
- PlankTOM 10.0 (not optimized, running on NEMO at UEA)
We decided to maintain PlankTOM 5.2 and PlankTOM 10.0 as core models.
No decision is taken regarding PISCES-T.
This means that developments by individual members of the group will be
included in one central code, which will be used to do climate test runs.
Each group member is responsible to verify their results for the core model
runs when they will be finished. Each group member can keep their own version of the code to pursue
their individual work. PlankTOM 10.0 will be modified by Erik to include:
- Meike's DMS cycle (ready in Dec 2007)
- Leticia's river input (ready)
- Manfredi's biomass/heat impact (ready)
- Nick's N cycle (ready in Dec 2007)
- Roisin's macro-zooplankton's best intial guess (current status to be used)
- Alex Baker's dust parameterisation with dust from Leads, and enhanced Si deposition
- Xavier's nutrient supply from sediments?
Ideally, developments will be done over a period of ~2 months before the end of winter.
Model evaluation
Meike presented a general strategy to evaluate model results more formally and
to keep track of model improvements. The central point was that several tests
are needed to properly evaluate model results, as different tests will assess
different aspects of the model. Some tests are quite complex to implement (and
understand), and we may decide at a later date to hire someone to do this.
For the immediate, the most useful tests included an estimate
of cost function, model efficiency, and model bias.
Meike to add equations here:
Model output
- All state variables (3D) and baseline information (such as T, light, etc)
should be saved on monthly averages.
- 3D fluxes should not be saved, but they should be computed
offline using shared ferret scripts
- Exceptions to this include the variables requested by Jim Orr for model
intercomparison and some of the more complex 3D variables, such as PFT-dependent primary
production
- 2D fluxes should be saved every day. In addition to what we already have, we need:
integrated pp, DMS flux, and MLD.
- We discussed the possibility of saving the 3D variables for the surface every day,
but did not come to a decision on this.
Things to do
- All: prepare your individual modules for implementation in PlankTOM-core
- Erik: implement the various components in PlankTOM-core
- Clare: get the restart ready (include macro-zooplankton, DMS and DMD (from Kettle DMS and DMSp data)
- Clare: work on web site to include ferret scripts for offline calculation of processes
(move "How to implement the code on cluster" to a separate file and include "How tos" from
our individual files. Also make a list of the namelist variables in one ferret file.)
- Clare: add global integrated diagnostics and model evaluation in Erik's post-processing file
(totals.output)