AMIP with Partitioned Fluxes

[LenkaNovak]

The Climate Modeling Alliance

Software Design Issue 📜

Purpose

We need a stable interface with robust checks, so once standalone ClimaAtmos is complete, we can immediately test our AMIP configuration. We will also finalize all large software components that need input from scientists (i.e. partitioned surface fluxes).

Cost/Benefits/Risks

b: streamlining AMIP delivery.

People and Personnel

• Lead: @LenkaNovak
• Collaborators: @valeriabarra
• Reviewers: @juliasloan25

Components

• there are two parts:
• 1. update package versions / fix outstanding bugs in existing code (AMIP Roadmap non-optional requirements)
  • once this is complete, the AMIP run can be used as a benchmark again
• 2. enhancements (other features used by AMIP setups)
  • fluxes: revise surface fluxes (also the order of timestepping (e.g. radiation) relative to flux computation in Atmos) + include the option to calculate partitioned fluxes
  • check for masking
  • bring back Slack reports

Results and Deliverables

• utility to calculate fluxes from multiple surfaces and then combine https://github.com/CliMA/ClimaCoupler.jl/milestone/5
• physical correctness summary to weekly Slack updates https://github.com/CliMA/ClimaCoupler.jl/milestone/6

QA

• conservation tests pass (e.g. Flux partitioning #360 )
• coupled comparable to atmos for performance
  • Atmos v Coupler like for like comparison (using the same dt and apply_limiters=false):
  • Atmos dt=200s (cpu=32, 22 nodes) → 23h25m
  • Coupler dt=200s (cpu=32, 1 node) → 20h50m
  • Coupler dt=200s, ct_cpl=600 (cpu=32, 2 nodes) → 20h45m
  • Coupler + topo, dt=100s (cpu=32, 1 node) → ~40h20
  • Coupler + topo + edmf, dt=100s (cpu=32, 1 node) → 70days in 20h17
• physical tests (see Buildkite runs for different comparisons, also Update Longruns #355, and Slack #coupler-report )
  • reasonable comparison with observations (TOA fluxes, precipitation, circulation metrics), given the Atmos setup (benchmark)

Task Breakdown And Schedule

1. fig outstanding bugs in existing code (AMIP Roadmap non-optional requirements)

• SIC clean up #282 [25 April]
• masking - ensure all tests are implemented and passing for all masking options Check masks #286 [1 May]
• use most recent ClimaAtmos [15 May]
  • this will be done in Update ClimaAtmos to v0.11.0 in ClimaCoupler #281 in a separate folder. Here we will update those API changes to the main driver, and we'll work together with @valeriabarra to update the general environment and tests.
• enable a surface flux calculation from the atmosphere (from averaged surface states) and partitioned flux calculation from the Coupler (using individual states) surface fluxes #311 [1 August]

SDI Revision Log

• tasks to accomplish the Coupler release were split into two SDIs (WIP)

CC

@tapios @simonbyrne @cmbengue

[LenkaNovak]

Complete