Support non-zero `SurfacePressure` by xylar · Pull Request #433 · E3SM-Project/Omega

xylar · 2026-06-11T17:47:50Z

This PR moves SurfacePressure ownership from VertCoord to OceanState so it participates in IO, and defines all Omega pressure fields as relative (gauge) pressure.

Changes

Move `SurfacePressure` into `OceanState`

Move SurfacePressure/SurfacePressureH arrays from VertCoord to OceanState
Register SurfacePressure in the State and Restart field groups so it is included in IC and restart streams
Share the device array back to VertCoord so it can be used when computing 3D pressure
Exchange halos and sync device→host after the initial-state stream read
Remove the placeholder TODO zero-fill from VertCoord
Call initSurfacePressure() in OceanInit after the stream read so the host mirror and halo are ready before the model advances

A consequence of this work is that SurfacePressure must always be present in the Omega initial condition.
It will be filled with fill values, not zeros, if it is absent, and the State field group read will fail for files that are missing SurfacePressure.

Define all pressure fields as relative (gauge) pressure

Omega's pressure fields (SurfacePressure, PressureInterface, PressureMid) are defined as relative pressure — absolute pressure minus the standard atmosphere (101325 Pa). Motivations:

TEOS-10 expects relative (gauge) pressure; Eos.h already passes Pressure * Pa2Db with no offset, consistent with this definition.
z-tilde pseudo-height ẑ = p / (ρ₀ g) equals zero at the sea surface under standard atmospheric forcing only when p is relative pressure. Under absolute pressure the surface pseudo-height would already be ≈ −10 m, which is conceptually strange.
The CF standard name sea_water_pressure is defined for absolute pressure, so it is removed from PressureInterface and PressureMid; no CF standard name exists for relative pressure.

Specific changes:

VertCoord.cpp: remove sea_water_pressure CF standard name from PressureInterface and PressureMid; update long names to say "relative pressure (gauge pressure)"; set minimum valid value to −AtmRefP (≈ −101325 Pa, since absolute pressure ≥ 0 implies relative pressure ≥ −standard atmosphere)
OceanState.cpp: remove surface_air_pressure CF standard name from SurfacePressure; update long name to say "relative pressure"
Eos.h: document that Pressure inputs to the TEOS-10 specific-volume, freezing-temperature, and Brunt-Väisälä functors are relative pressure
VertCoord.h/VertCoord.cpp: clarify that computePressure computes relative pressure given relative SurfacePressure
User guide (OceanState.md, VertCoord.md): describe all pressure fields as relative, note that SurfacePressure is 0 Pa for standard atmospheric forcing and represents the atmospheric anomaly in coupled runs
Developer guide (OceanState.md): note that SurfacePressure is relative pressure, typically ≈ 0 Pa for free-surface runs

Checklist

This merge adds 5 constexpr fill value constants (FillValueI4, FillValueI8, FillValueR4, FillValueR8, FillValueReal) in namespace OMEGA, wrapping PIO_FILL_* from SCORPIO.

This merge adds private fillWithValue<T>() template helper (using if constexpr on T::value_type) and a call to it at the end of attachData<T>(). Arrays are now auto-filled with the declared fill value at attach time.

This merge removes all local fill value declarations and replaces references with the centralized constants.

In some cases we definitely don't want to overwrite fields we attach with fill values. Fix 2 such cases in Tracers and OceanState

We don't want fill values at every boundary edge becausee these are also valid. Instead, we want zeros except for edges between two inactive cells (e.g. both adjacent cells are below bathymetry)

Rather than assuming that the NormalVelocity read from an initial condition has correct masking, we ensure that it has fill values for inactive layers, zeros at boundary edges (both adjacent to land and to bathymetry), and leave its values unchanged for active, non-boundary layers.

We need to ensure that it is zero at active boundary edges.

- Move SurfacePressure/SurfacePressureH arrays from VertCoord to OceanState, which already owns the other state fields (NormalVelocity, PseudoThickness) - Register SurfacePressure in the State and Restart field groups in OceanState::defineFields() so it is included in IC and restart streams - Share the device array back to VertCoord via setSurfacePressure() so its pressure-computation kernels continue to work without duplication - Add OceanState::initSurfacePressure() to exchange halos and sync device->host after the initial-state stream read - Remove the placeholder TODO zero-fill from VertCoord; the field now arrives pre-filled from the IC file or defaults to zero from host allocation - Call initSurfacePressure() in OceanInit after the stream read so the host mirror and halo are ready before the model advances Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

- Call initSurfacePressure() in initStateTest() to mirror the production initialization sequence after the IC stream read - Test that SurfacePressureH extent matches NCellsSize - Test that SurfacePressure reads as zero when not present in the IC file, confirming the default-zero behavior expected for uncoupled runs Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

xylar · 2026-06-11T17:48:08Z

This branch is based off of #428

xylar · 2026-06-11T17:55:29Z

##Testing

Testing was with E3SM-Project/polaris#610

Polaris `omega_pr` suite

Baseline workdir: /lcrc/group/e3sm/ac.xylar/polaris_1.0/chrysalis/test_20260610/omega-pr-fix-fill-values
Baseline build: /lcrc/group/e3sm/ac.xylar/polaris_1.0/chrysalis/test_20260610/omega-pr-fix-fill-values/build
PR build: /lcrc/group/e3sm/ac.xylar/polaris_1.0/chrysalis/test_20260610/omega-pr-surface-pressure-init2/build
PR workdir: /lcrc/group/e3sm/ac.xylar/polaris_1.0/chrysalis/test_20260610/omega-pr-surface-pressure-init2
Machine: chrysalis
Partition: compute
Compiler: oneapi-ifx
Build type: Release
Log: /lcrc/group/e3sm/ac.xylar/polaris_1.0/chrysalis/test_20260610/omega-pr-surface-pressure-init2/polaris_omega_pr.o1231363
Result: All tests passed

A new realistic global initial condition is needed with SurfacePressure included before CTests can be expected to pass. I have not yet done this, and am hoping that the utility in E3SM-Project/polaris#562 might be used.

xylar · 2026-06-11T17:56:08Z

Documentation hasn't yet been updated.

xylar · 2026-06-12T16:31:42Z

We need to add to this that we subtract standard_atmosphere from pressure when computing SpecVol.

xylar · 2026-06-13T11:07:24Z

I will instead define pressure fields clearly as relative pressure and drop the CF standard name sea_water_pressure, see E3SM-Project/polaris#610 (comment)

Remove the incorrect CF standard name `surface_air_pressure` (which applies to absolute atmospheric pressure) and update the long name to explicitly say "relative pressure", since SurfacePressure is defined as pressure relative to the standard atmosphere (101325 Pa). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

All three pressure fields (SurfacePressure, PressureInterface, PressureMid) are relative pressure — absolute pressure minus the standard atmosphere (101325 Pa). This clarification is needed because: - The CF standard name sea_water_pressure applies to absolute pressure, so it is removed and left blank for PressureInterface and PressureMid - The TEOS-10 EOS expects relative pressure, consistent with how Eos.h already passes pressure (Pa2Db * Pressure with no AtmRefP offset) - The z-tilde pseudo-height is 0 m at the sea surface under standard atmospheric forcing only when p is relative pressure Changes: - VertCoord.cpp: remove sea_water_pressure CF name, update long names and min valid values (-AtmRefP, since absolute pressure >= 0) for PressureInterface and PressureMid; clarify computePressure comment - VertCoord.h: clarify computePressure declaration comment - Eos.h: note that Pressure inputs to TEOS-10 functors are relative - userGuide/OceanState.md: describe SurfacePressure as relative pressure, typically 0 Pa for standard atmospheric forcing - userGuide/VertCoord.md: note that p in the pseudo-height formula is relative pressure; update PressureInterface/PressureMid descriptions - devGuide/OceanState.md: note SurfacePressure is relative pressure Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

xylar · 2026-06-13T12:27:21Z

Updated Testing

Again with E3SM-Project/polaris#610

Polaris `omega_pr` suite

Baseline workdir: /lcrc/group/e3sm/ac.xylar/polaris_1.0/chrysalis/test_20260611/omega-pr-baseline/
Baseline build: /lcrc/group/e3sm/ac.xylar/polaris_1.0/chrysalis/test_20260611/omega-pr-baseline/build
PR build: /lcrc/group/e3sm/ac.xylar/polaris_1.0/chrysalis/test_20260613/omega-pr-surface-pressure-init/build
PR workdir: /lcrc/group/e3sm/ac.xylar/polaris_1.0/chrysalis/test_20260613/omega-pr-surface-pressure-init
Machine: chrysalis
Partition: compute
Compiler: oneapi-ifx
Build type: Release
Log: /lcrc/group/e3sm/ac.xylar/polaris_1.0/chrysalis/test_20260613/omega-pr-surface-pressure-init/polaris_omega_pr.o1235245
Result: All tests passed

xylar · 2026-06-13T12:30:39Z

@cbegeman, @andrewdnolan and @alicebarthel, sorry for the see-sawing on relative vs. absolute pressure. I have settled on relative pressure and that is documented here now.

Please review and modify as needed (e.g. rebasing after #428 goes in). Please get @sbrus89 or someone else with the privileges to merge when you're happy, and once #428 has been merged.

cbegeman

I agree with the code changes you made here. I will circle back for testing.

Adds `surfacePressure` to Omega initial conditions ... ...but does not add it to yaml files. Thus, `surfacePressure` is not used by the model. This relies on E3SM-Project/Omega#433. Adding `surfacePressure` to the yaml files for Omega-supported tests (that do not use `State`) will be done in a further PR. This PR also adds `surfacePressure` --> `SurfacePressure` mapping from MPAS-Ocean --> Omega. We need to add `SurfacePressure` explicitly to to manufactured solution initial state, since it does not use the `State` field group.

xylar and others added 18 commits June 10, 2026 07:39

Add a design document for fixing fill values

e480c98

Add fill value constant

86d832a

This merge adds 5 constexpr fill value constants (FillValueI4, FillValueI8, FillValueR4, FillValueR8, FillValueReal) in namespace OMEGA, wrapping PIO_FILL_* from SCORPIO.

Fill fields with fill values during attachData()

bb6330c

This merge adds private fillWithValue<T>() template helper (using if constexpr on T::value_type) and a call to it at the end of attachData<T>(). Arrays are now auto-filled with the declared fill value at attach time.

Update fill values in all fields

26278a2

This merge removes all local fill value declarations and replaces references with the centralized constants.

Determine fill values automatically by array type

1967d50

Drop manually assigned fill values

51fe005

Make detection of fill values in VertCoord reads less brittle

51da799

Add FillOnAttach argument to attachData()

a115ef9

In some cases we definitely don't want to overwrite fields we attach with fill values. Fix 2 such cases in Tracers and OceanState

Reorder 2 cases so we attach first, then fill in the array

53a8d97

Fill fluxes on edges first with zeros over full physical range

b2cbb2b

We don't want fill values at every boundary edge becausee these are also valid. Instead, we want zeros except for edges between two inactive cells (e.g. both adjacent cells are below bathymetry)

Add fill-value CTests

366f447

Update the docs

9e3ca79

Fix bounds on DivHU computation to avoid inactive layers

8539edd

Mask NormalVelocity in time stepper

fce8ffa

We need to ensure that it is zero at active boundary edges.

Correct comment on where BottomGeomDepth gets read from

71bfe6a

xylar marked this pull request as draft June 11, 2026 17:55

xylar mentioned this pull request Jun 11, 2026

Add surfacePressure to Omega initial conditions E3SM-Project/polaris#610

Merged

2 tasks

xylar requested review from alicebarthel, andrewdnolan and cbegeman June 11, 2026 18:05

Update the docs based on non-zero SurfacePressure

2233adf

xylar commented Jun 12, 2026

View reviewed changes

Comment thread components/omega/src/ocn/OceanState.cpp Outdated

xylar commented Jun 12, 2026

View reviewed changes

Comment thread components/omega/src/ocn/OceanState.cpp Outdated

xylar force-pushed the omega/support-surface-pressure branch 2 times, most recently from 32511f1 to d18f6bd Compare June 12, 2026 20:04

xylar force-pushed the omega/support-surface-pressure branch from d18f6bd to 6ef1c89 Compare June 13, 2026 11:27

xylar force-pushed the omega/support-surface-pressure branch from 6ef1c89 to 18c5e81 Compare June 13, 2026 11:50

xylar assigned sbrus89 Jun 13, 2026

cbegeman reviewed Jun 15, 2026

View reviewed changes

cbegeman mentioned this pull request Jun 16, 2026

Require state vars in write_initial_state_dataset, including surface pressure E3SM-Project/polaris#615

Open

1 task

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Support non-zero `SurfacePressure`#433

Support non-zero `SurfacePressure`#433
xylar wants to merge 21 commits into
E3SM-Project:developfrom
xylar:omega/support-surface-pressure

xylar commented Jun 11, 2026 •

edited

Loading

Uh oh!

xylar commented Jun 11, 2026

Uh oh!

xylar commented Jun 11, 2026

Uh oh!

xylar commented Jun 11, 2026

Uh oh!

Uh oh!

Uh oh!

xylar commented Jun 12, 2026

Uh oh!

xylar commented Jun 13, 2026

Uh oh!

xylar commented Jun 13, 2026

Uh oh!

xylar commented Jun 13, 2026

Uh oh!

cbegeman left a comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

xylar commented Jun 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changes

Move SurfacePressure into OceanState

Define all pressure fields as relative (gauge) pressure

Uh oh!

xylar commented Jun 11, 2026

Uh oh!

xylar commented Jun 11, 2026

Polaris omega_pr suite

Uh oh!

xylar commented Jun 11, 2026

Uh oh!

Uh oh!

Uh oh!

xylar commented Jun 12, 2026

Uh oh!

xylar commented Jun 13, 2026

Uh oh!

xylar commented Jun 13, 2026

Updated Testing

Polaris omega_pr suite

Uh oh!

xylar commented Jun 13, 2026

Uh oh!

cbegeman left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

xylar commented Jun 11, 2026 •

edited

Loading

Move `SurfacePressure` into `OceanState`

Polaris `omega_pr` suite

Polaris `omega_pr` suite