Remove "Build Type" option from albany input file. #6286
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Update MALI-Dev/develop with latest code in E3SM-Project/master after having recently merged the other way. Needed to keep MALI-Dev/develop from getting out of date with non-MALI code. * e3sm/master: (2435 commits) Provenance: do not crash if config_src does not exist SMP_PRESENT is gone, use BUILD_THREADED Disable gpu env-vars in cpu runs; append cuda path to link flags CPL: Fix uninitialized variable. update fates external for API31 refactor refactor elmfates api to avoid calling hlm-side procudures Set E3SMinput project paths and 1:1 MPI-to-GPU affinity Array initialization Re-added FATES specific initialization for fire calculation. Update gnugpu, nvidia and nvidiagpu cmake and modules Add `-fopenmp` for OpenMP support with cray-cray Adjust gnugpu_frontier.cmake macro Update paths and modules, add gnu config Update paths and queues Update Polaris config add polaris machine persist-credentials: false Bump actions/setup-python from 4.7.0 to 5.0.0 Fix Homme Cmake, EKAT subm update changed the EkatBuildKokkos module Fix MOAB compile issues ...
After further inspection, I realized that the mass conservation equation used to evolve sheet thickness is missing a source term from the production of meltwater within the channel. This is the 2nd term on the RHS of Hewitt (2013) Eq. 7, and it is missing from Eq. 54 in the Hoffman et al. 2018 GMD paper. Melt production in the channel is generally a small contribution to the total water balance, which may explain why the model matched GlaDS well for the SHMIP experiments. It's possible this correction will help with channel blowup events, or it could make them worse - it's hard to think through how this affects those events without better characterizing what's causing them. In any case, this is a clear omission and should be used going forward.
This reverts commit ae98181. It seems like we would want the channel model to see the elevation head produced by a lake's water thickness to ensure that channel flow is consistent with the presence of the lake. This commit reverts the commit that dropped the water thickness elevation head from the hydropotential gradients used by the channel. In the sheet model, the elevation head is removed from the advection equation, but it is handled by the diffusion equation, so it's still represented in the evolution. As the channel model currently exists, the elevation head is almost completely absent (other than its effect on the magnitude of gradMagPhiEdge). It seems like this could lead to channels draining *into* lakes even when the full hydropotential gradient of the lake indicates water flow should be draining the lake.
The water thickness in the ocean is undefined, and if anything we'd have diffusion from the ocean (thick water layer) to the grounded ice, and we don't want any flux from the ocean inward.
Previously the call to mpas_calculate_barycentric_weights_for_points to set up barycentric weights for interpolation from cells to vertices occurred in the mpas_li_sia.F module. However, these weights may also be needed by the subglacial hydro code, which can be called with any (or none) velocity solver. This commit moves that initialization to mpas_li_setup and adds logic to call the weight generation routine if either the SIA or SGH code requires it.
Without explicitly doing this, the gradients are likely using neighboring values of 0, which would yield unexpectedly large gradients.
This commit explicitly treats the calculation of hydropotentialBaseSlopeTangent near mesh boundaries. For the two vertex based methods, edges with vertices on the boundary of the mesh will have invalid values. For the normal slope method, any edges on cells that are at the edge of the mesh will have contaminated values. In those cases, set tangent slope to zero rather than leaving the possibility of garbage values.
Without this change, the CFL condition might be overly restrictive because values from cells that aren't part of advection get included in the calculation and they may happen to have values that are more restrictive than the valid edges.
Simplify calculation disallowing inflow and move it *before* extra logic limiting outflow gradient.
Channels now use updated version of gradMagPhiEdge, which is dependent on full hydropotential. This is done to keep channels from filling up lakes even when lake is full and hydropotential is in opposite direction. Distributed system is now dependent on new variable gradMagPhiBaseEdge, which is dependent only on hydropotentialBase and is equivalent to gradMagPhiEdge in previous commits.
Creates hydroTerrestrialMarginMask variable, which designates the terrestrial margins of the active subglacial hydrology model
Edits description of config_SGH_tangent_slope_calculation in Registry_subglacial_hydro.xml based on detailed testing of 'from_normal_slope' and 'from_vertex_barycentric' options.
Explicitly sets diffusivity to zero at boundary edges, instead of just waterFluxDiff
Redfines hydroTerrestiralMarineMargin to AT or above sea level, whereas hydroMarineMarginMask is exclusively below sea level.
Debugs calculations of channelAreaChangeCell and channelMeltInputCell. Loop where these were being calculated was overwritting cell values each iteration through nEdgesOnCell. Now each cell is the sum of all of its edges as intended.
This is necessary to get BFB results on different decompositions. This addition was made necessary by c9c13aa. There may be a way to adjust halo updates to avoid needing to add these both, as they are both local calculations. But it is likely complex, so this solution is adequate in that it works.
As previously ordered, the operations in evolving waterThickness and pressure variables were out of sync, resulting in the full hydropotential variable using the old waterThickness in the term calculating the water layer elevation head. This was causing restarts with the channel model to fail and channel hydropotential gradients to be inaccurate (unknown how significant that was). This commit moves the pressure calculation to eliminate this issue. Code comments explain in more detail.
This is needed for BFB restarts with the till model. We never use the till model, but I identified this issue while debugging channel restarts.
…ev/develop This PR introduces a number of fixes and improvements to the subglacial hydrology model to get it to run more stably, mostly related to handling of boundaries of the subglacial hydrology domain: * Adds a missing channel melt source term to the mass cons eqn. This term is typically very small but was missing * Make channels see the elevation head of lake thickness in the hydropotential to avoid channels flowing in to lakes * Disable diffusive water flux at grounding line where water thickness is undefined * Explicitly handle gradients at boundaries of the mesh where they are undefined * Move the ordering of the pressure calculation to enable consistency in the time levels used for key model variables (affects channel model only) * Update channel CFL calculation to only consider valid edges * zero the diffusivity at domain boundaries to prevent those edges from affecting diffusive CFL * creation of new hydroTerrestrialMarginMask to be used for handling boundaries of the SGH domain * other minor cleanup/adjustments * origin/matthewhoffman/mali/hydro_boundary_cleanup: Initialize deltatSGH only on a cold start & make it a restart variable Move pressure calc so it uses consistent time levels of variables Add halo updates on two channel variables fix typo channelAreaChangeCell/channelMeltInputCell debug hydroTerrestrialMarginMask >= sea level Zero diffusivity at boundary edges Edit SGH_tangent_slope_calculation description Create hydroTerrestrialMarginMask Channel gradMagPhiEdge dependent on full water hydropotential Update gradient calulations at boundaries of hydro domain Adjust hydro CFL calculation to ignore invalid edges Treat tangent slope calculation near boundary of mesh Zero gradients at edges of the mesh Move call to mpas_calculate_barycentric_weights_for_points Disable diffusive water flux at grounding line Revert "Ignore the water thickness head in the channel model" Add missing channel melt source term to mass cons eqn
- Add "totalSubglacialWaterVolume" - Add "totalLakeVolume" - Add "totalBasalMeltInput" - Add "totalExternalWaterInput" - Add "totalChannelMelt" - Add "totalGLMeltFlux" - Add "totalTerrestrialMeltFlux" - Add "totalChannelGLMeltFlux" - Add "totalChannelTerrestrialMeltFlux" - Add "totalFlotationFraction" - Add "avgFlotationFraction"
Adds conditional statements in mpas_li_global_stats.F to prevent calculating SGH global stats when SGH model is deactivated.
The basalMeltInput term was including non-grounded ice in the calculation of the totalBasalMeltInput and giving wrong answers. The mask ensures that only basal melt occurring under grounded ice is considered.
This PR request includes updates to the global stats to include new analysis members for subglacial hydrology. The new analysis members are: * totalSubglacialWaterVolume * totalSubglacialLakeVolume * totalSubglacialLakeArea * totalBasalMeltInput * totalExternalWaterInput * totalChannelMelt * totalDistWaterFluxMarineMargin * totalDistWaterFluxTerrestrialMargin * totalChnlWaterFluxMarineMargin * totalChnlWaterFluxTerrestrialMargin * avgFlotationFraction * origin/cshafer/hydro_stats: Correct flotation fraction globalStat to use ice density Update variable names and other code review edits Add grounded ice mask to BasalMeltInput term Fix hydro globalStats bug - protect global stats calcs if SGH disabled Add SGH analysis members to global stats
This PR updates handling of timesteps between MALI and the SLM in a few ways: * switch config_slm_coupling_interval to be an integer in years because we only allow integer year values * On init, check that config_adaptive_timestep_force_interval divides evenly into config_slm_coupling_interval * On init, check that restart interval is an even multiple of config_slm_coupling_interval * On a restart, calculate which SLM time level to use based on the elapsed time from the start of the original simulation and config_slm_coupling_interval and make sure these divide cleanly
Also add missing =>next pointer assignment to keep code from hanging
Trying to cast intervals into dateTimeStrings did not work.
Following the recent addition of subglacial hydro quantities to the global stats analysis member, we've used those additions as template to add the quantitites to the regional stats analysis member.
The reduction of `fluxAcrossGroundingLine` was accidently moved within a `config_SGH` condition, so that `groundingLineFlux` would only be calculated if `config_SGH` was turned on. Moved the reduction back to where it was to ensure it's calculated in all situations it's needed.
The grounded ice mask is used in the calculation of the `basalMeltInput` term but was excluded from `externalWaterInput`, which could cause problems in closing budgets. The grounded ice mask was not added to any of the calculations dependent on `waterThickness` b/c SGH model explicitly sets the `waterThickness` to zero outside of the grounded ice area.
Masking by `regionCellMasks` in SGH regional stats terms was missing, which meant all cell centered SGH values would have been uniform across the regions and match the global stats value. Still not determined how the edge centered SGH value should be mask by region (e.g. using upwind cells region or by the `regionEdgeMasks`), but once that is decided the matching region masking needs to be done for the SGH edge quantities.
Was missing the deallocation of `regionalSumFlotationFraction`, which is a local variable needed for calculating the numerator of `regionalAvgFlotationFraction`.
This commit changes how restarts are handled when the SLM is active to allow MALI to be restarted at any arbitrary restart interval and have the SLM restart correctly. This is done by changing the SLM coupling alarm to be based off of the original simulationStartTime (instead of the start time of the current execution). This required moving the creation of the coupling alarm to later in initialization so that the variable simulationStartTime is available. With this change, it was also necessary to change the way the SLM time level is calculated on a restart to take the floor of the elapsed time divided by the coupling interval, rather than requiring that there be no remainder. This adds a little fragility because there is no way to double check that is the correct SLM time level, but if this is set up correctly, it should be handled properly. Finally, as part of these changes, I also removed the check on init that the coupling interval divides evenly into the restart interval, because that's no longer a requirement. That's sort of too bad, because it was a lot of work to figure out how to make that check! But it's nicer to not have that restriction.
This doesn't serve any internal purpose, but it could help a user detect an error in their configuration.
It's not needed, and if the restart time is not a coupling interval, it will make the SLM get out of sync.
…Dev/develop Previously, when the regional sea-level prediction capability was added to MALI (#21), the restart config option for the sea-level model was not added. This led the sea-level model to get initialized to Timestep zero when coupled MALI-SLM simulations are being restarted, forgetting about the ice loading changes and associated viscoelastic solid earth deformation that happened in the timesteps prior to current model time. This PR fixes the problem by allowing the sea-level model to resume where it was left off. Note in parallel to this PR, the version of the SLM needs to incorporate the changes made in the following accompanying PR (MALI-Dev/1DSeaLevelModel_FWTW#9) * hollyhan/add_restart_functionality_slm: Don't call SLM on init of a restart Add addl info on restart about the calculated time since last SLM call Allow restarts at any interval when using SLM Add missing error flag so model actually dies when error occurs Add missing arguments to log write statement Update restart check to also use time interval division Adjust check if adaptive dt is on or not Update checks using interval division Improve error handling, correct other usage of config_uplift_method Improve synchronization of timesteps between MALI and SLM Add restart option when the SLM is coupled to MALI
Addes `regionEdgeMask` and `regionVertexMask` to the registry so the additional mask variables can be included in the `regionsInput` stream.
Co-authored-by: Matt Hoffman <wyeast@gmail.com>
`regionVertexMasks` can be added to the registry later when it's needed. I've also added `regionEdgeMasks` to the restart stream and fixed and xml type I made in the previous commit.
…evelop This PR follows #82, which added subglacial hydrology (SGH) quantities to the global stats analysis member, by adding SGH quantities to the regional stats analysis member. This PR also address a bug introduced in #82 where the calculation of groundingLineFlux and groundingLineMigrationFlux were moved within an if config_SGH then condition. The bug prevents groundingLineFlux and groundingLineMigrationFlux from being calculated by global stats unless the SGH model is turned on, despite these quantities applying to simulations where SGH is not used. * MALI-Dev/andrewdnolan/mali/hydro_regional_stats: Fix registry typo Matt caught in review. Remove `regionVertexMasks` from registry to minimize file size. Apply suggestions from code review Use `regionEdgeMask` to calculate SGH regional stats defined on edges. Deallocate `regionalSumFlotationFraction`. Add missing `regionCellMasks` call from SGH regional stats terms Add grounded ice mask to `externalWaterInput` term Move reduction of `fluxAcrossGroundingLine` outside SGH condition. Add support for subglacial hydro quantities in regional stats.
Epetra solver stack is no longer available, and this change will be needed with newer versions of Albany. See sandialabs/Albany#1028
|
@mperego, we should merge this into |
|
Wrong repo, sorry. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Remove "Build Type" option from albany input file.
Epetra solver stack is no longer available through Albany and this change will be needed with newer versions of Albany.
See sandialabs/Albany#1028
This will work with older versions of Albnay if Albany is built with no Epetra support.