Home > climate models, uvic_escm > UVIC_ESCM: Main Flow Chart

UVIC_ESCM: Main Flow Chart

2012 February 18

“Begin at the beginning,”, the King said, very gravely, “and go on till you come to the end: then stop”
-Lewis Carroll

Last week’s dump of the entire calling structure was an interesting exercise, but didn’t go very far in revealing the structure of UVic_ESCM. This week I focus on just the major components of the “main” routine. The graphic to the left is my first pass at a high-level summary.

Before continuing, you should be aware of ClimateSight Kate’s previous work, as a research student of Steve Easterbrook, which covers much of this ground already. In particular, this diagram of UVic_ESCM’s structure.

I want to dive deeper, though. The goal is to get into enough detail in the model to know how it’s put together and how to intelligently modify it.

The main routine is pretty straightforward.

1. Initialize Surface Boundary Condition (SBC) indices
2. Initialize tracers
  a. Ocean
  b. Atmosphere
3. Setup Ocean
4. Setup Atmosphere
5. Setup Land
6. Calculate time steps
7. Check consistency of SBC
8. Initial Global Sums
9. Main Loop
  a. atmos
     1. gasbc
     2. embm
  b. land
     1. glsbc
     2. mtlm
  c. ocean
     1. gosbc
     2. mom
     3. embmout
     4. mtlmout
  d. global sum
10 Final Global Sum

Or, graphically:

Advertisements
  1. anonymous
    2012 February 19 at 3:49 am

    would there be any advantage wrt computing time if the subcomponents were run separately for longer periods (f.e. a week) of simulation time, how would the energy and water transfer be handled in this case (would diurnal rythms be lost?), and how much loss of accuracy would this make?

  2. 2012 February 19 at 8:04 am

    Thanks for the comment. Please understand that I have nothing authoritative to say about the UVic_ESCM model at this time. Anything I do say is subject to correction as I learn more. I am an explorer in unfamiliar territory.

    would there be any advantage wrt computing time if the subcomponents were run separately for longer periods (f.e. a week) of simulation time

    I would *think* so, but, of course, something else would be lost. I don’t know enough yet to guess what would be lost.

    how would the energy and water transfer be handled in this case

    At first glance, the transfer between components appears to be through MTLMout (land/ocean) and EMBMout (atmos/ocean). I believe that the components on different sides of those subcomponents can be run at different time steps. The land model, for instance, appears to run at 1 hour time steps while the ocean and atmos models run at 12 hour time steps.

    how much loss of accuracy would this make

    In a few weeks, I might know enough to experiment in this direction. For now, I too am in the dark.

  3. anonymous
    2012 February 19 at 8:57 pm

    Thank you sir for the response, I’d imagine if the components exchanced data only on weekly basis on simulation time, this would lead to a delayed response on all variables, which then would be problematic wrt incoming radiation. I’m just guessing here too, I’ve only done proper analysis in the style of Foster & Rahmstorf (monthly data). I’ve tried analysing daily data of NH ocean ice, this went quite off-side (taking WAGs of the various albedos, estimating the increase of snowfall in the autumn). I’m surprised to hear the land component runs with such small steps, as f.e. carbon uptake by plants in spring/summer is somewhat regulated by the atmophere (sunshine%, precipitation, soil moisture, nutrients). But then there likely are plenty other effects that might require such a small step.

  1. No trackbacks yet.
Comments are closed.