Changes between Version 16 and Version 17 of 2017UpdateNEMOStrategy/FirstSteps
- Timestamp:
- 2016-12-08T01:32:46+01:00 (8 years ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
2017UpdateNEMOStrategy/FirstSteps
v16 v17 60 60 == Chapter 3 - HPC performance constraints - HPC Working Group == 61 61 Please see the attached word document [https://forge.ipsl.jussieu.fr/nemo/attachment/wiki/2017UpdateNEMOStrategy/FirstSteps/NEMO_Development_Strategy_HPC_questions.docx] 62 62 63 == Chapter 4 - Future evolutions of NEMO ocean kernel. - Mike Bell == 63 64 == Chapter 5 – Ocean dynamics component of NEMO (including state of the art on subgridscale parametrisations and on tides, see above) - Julien Le Sommer == 64 65 == Chapter 6 – Toward locally higher effective resolution: AGRIF - Jérôme Chanut == 65 66 == Chapter 7 - The assimilation and ensemble component of NEMO - Pierre-Antoine Bouttier and Dan Lea == 66 * Establish an identity for the NEMO DA community67 -Create a list of NEMO users working on DA or using the assimilation interface for other purposes (for example model validation compared to observations). This would be held on WG wiki or the NEMO website.68 -Keep the wiki regularly updated69 -Create a NEMO DA mailing list70 * The future of NEMO-TAM and whether and how it is kept up to date with model updates is a question the DA WG will consider. No recommended actions currently. This will be reviewed at future WG meetings.71 * Ensemble mode. There is an increasing interest in the data assimilation community for using ensemble information in data assimilation.72 -Many group members would find a single executable approach useful for running all the ensemble members. This requires an MPI communicator for the sub-domains of each ensemble member and another MPI communicator for passing information between ensemble members.73 * Higher resolution modelling74 -Improve the efficiency of the assimilation interface code (OBS and ASM)75 -Deal correctly with observations which are spread over several model grid cells (i.e. have a "footprint")76 * NEMO-OBS package77 -If the above ensemble mode is introduce the NEMO-OBS package should be updated to produce ensemble diagnostics.78 -Update the central version of the OBS plotting tool (dataplot) to a new python version79 -Develop some community tools for assessing ensembles which can either go into the TOOLS directory or be made available on the WG wiki or NEMO website.67 * Establish an identity for the NEMO DA community 68 * Create a list of NEMO users working on DA or using the assimilation interface for other purposes (for example model validation compared to observations). This would be held on WG wiki or the NEMO website. 69 * Keep the wiki regularly updated 70 * Create a NEMO DA mailing list 71 * The future of NEMO-TAM and whether and how it is kept up to date with model updates is a question the DA WG will consider. No recommended actions currently. This will be reviewed at future WG meetings. 72 * Ensemble mode. There is an increasing interest in the data assimilation community for using ensemble information in data assimilation. 73 * Many group members would find a single executable approach useful for running all the ensemble members. This requires an MPI communicator for the sub-domains of each ensemble member and another MPI communicator for passing information between ensemble members. 74 * Higher resolution modelling 75 * Improve the efficiency of the assimilation interface code (OBS and ASM) 76 * Deal correctly with observations which are spread over several model grid cells (i.e. have a "footprint") 77 * NEMO-OBS package 78 * If the above ensemble mode is introduce the NEMO-OBS package should be updated to produce ensemble diagnostics. 79 * Update the central version of the OBS plotting tool (dataplot) to a new python version 80 * Develop some community tools for assessing ensembles which can either go into the TOOLS directory or be made available on the WG wiki or NEMO website. 80 81 81 == Chapter 8 - The ice components of NEMO - Ed Blockley and Martin Vancoppennolle == 82 == Chapter 8 - The ice components of NEMO - Ed Blockley and Martin Vancoppenolle == 83 There are numerous directions towards which the representation of sea ice could be improved. 84 85 A non-exhaustive and perfectible list would include 86 87 * sea ice rheology: Towards improving the representation of deformation scaling laws, landfast ice & ice arches. Available approaches to be discussed include 88 89 * VP-likes vs anisotropic vs elasto-britlle rheologies. 90 91 * Eulerian vs lagrangian approaches 92 93 * ice atmosphere coupling: Towards improving the representation of ice-atmosphere interactions, notably in polynya regions or near the ice edge. This would clearly benefit from new possibilities offered by the interactive boundary layer physics. A new approach of coupling via conduction flux in the ice has been proposed at UKMO. 94 95 * ice thermodynamics: Ice thermodynamics are central because they control the energetic transfer in the sea ice zone. Improvements include better account of 96 97 * mushy layer thermodynamics (salt & heat conservation, frazil ice, platelet ice) 98 99 * snow & surface meltwater (melt ponds) physics 100 101 * radiative transfer (beer law -> delta-eddington schemes) 102 103 * ice floe size reperesentation. Ice floe size is absent from current sea ice representations in large-scale models. Yet floe size controls lateral melting and wave-ice coupling. Recent works have explored possible ways of improvements. 104 105 * sea ice biogeochemistry. Sea ice BGC is more of an observational than a modelling problem at the moment. Yet recent studies suggest simple first-order representations of organic matter synthesis in sea ice, carbonate chemistry and carbon fluxes in the sea ice zone. 106 107 Experts suggested: D. Feltham (Reading), T. Fichefet (UCL), D. Salas (Meteo France), 82 108 83 109 == Chapter 9 - Air sea interface and surface boundary layer physics - Laurent Brodeau, Gurvan Madec == 84 110 Please see [https://docs.google.com/document/d/14tj6jZEgqXhVp1n1_IcZJn4XMIoT5K76a6Mi8sWFuh4/edit?usp=sharing here!!!] Can't cope with this wiki syntax... 111 85 112 == Chapter 10 - The biogeochemical component of NEMO: TOP and its interface - Olivier Aumont and Tomas Lovato == 86 * Establish the production of scientifically validated simulations (e.g., validation suite and data repository) to enable the long term assessment of model developments and improvements.113 * Establish the production of scientifically validated simulations (e.g., validation suite and data repository) to enable the long term assessment of model developments and improvements. 87 114 88 * The medium term development of TOP should lead to a general interface for ocean tracers 115 * The medium term development of TOP should lead to a general interface for ocean tracers 116 89 117 capable to solve all key processes (BC treatment, light in the water column, sinking etc), independently from a specific biogeochemical model. This will translate into a seamless interface where only the core equations of biogeochemical models will be needed to couple with NEMO. 90 118 91 * The uprising complexity of biogeochemical models will require the revision of current HPC approach and the adoption of novel techniques to cope with the increasing computational efforts. In particular, the exploitation of accelerators and GPUs might become a key resource to deal with computational constraints.119 * The uprising complexity of biogeochemical models will require the revision of current HPC approach and the adoption of novel techniques to cope with the increasing computational efforts. In particular, the exploitation of accelerators and GPUs might become a key resource to deal with computational constraints. 92 120 93 121 == Chapter 11 - NEMO validation and range of user support (including user interface to build configurations, and Configuration Manager) - Andrew Coward == 122 * A clean-up of the configuration set-up routines in NEMO is seen as an essential simplification. This has been started in 2016 and will continue. This process puts greater responsibility on a configuration developer to provide either external input files or user-defined code modules. The system team will endeavour to provide sufficient examples and tools to ease this burden. The benefits are a simpler code kernel which is more suitable for adaptation to future architectures and, ultimately, a less error-prone route to setting up new configurations. Examples and methods need to be well documented and these guides need to be readily available to ease the transition. 94 123 95 * A clean-up of the configuration set-up routines in NEMO is seen as an essential simplification. This has been started in 2016 and will continue. This process puts greater responsibility on a configuration developer to provide either external input files or user-defined code modules. The system team will endeavour to provide sufficient examples and tools to ease this burden. The benefits are a simpler code kernel which is more suitable for adaptation to future architectures and, ultimately, a less error-prone route to setting up new configurations. Examples and methods need to be well documented and these guides need to be readily available to ease the transition.124 * The new interface eases the task of providing simple test cases and the range of such options is expected to grow rapidly. Test cases should be viewed separately to reference configurations but they are useful for demonstrating and comparing numerical schemes; especially when analytical solutions exist for comparison. 96 125 97 * The new interface eases the task of providing simple test cases and the range of such options is expected to grow rapidly. Test cases should be viewed separately to reference configurations but they are useful for demonstrating and comparing numerical schemes; especially when analytical solutions exist for comparison.126 * XIOS-2 should be adopted as default but this still lacks documentation targeted towards NEMO users. The system team should consider compiling an 'XIOS in NEMO' guide detailing relevant settings and tricks for obtaining some of the more complex outputs. Mostly this is a case of documenting the work done to obtain CMIP6 diagnostics. 98 127 99 * XIOS-2 should be adopted as default but this still lacks documentation targeted towards NEMO users. The system team should consider compiling an 'XIOS in NEMO' guide detailing relevant settings and tricks for obtaining some of the more complex outputs. Mostly this is a case of documenting the work done to obtain CMIP6 diagnostics. 100 101 * A basic but well-tested Configuration Manager (SIREN) is now part of the system. These tools allow a regional sub-model of any ORCA grid to be set up easily. Users will expect such tools for setting up regional models with open boundaries (BDY) and for AGRIF nests; ideally the same tool should be used for both. The UK has made some progress with Python-based tools for constructing regional sub-models from any gridded input. This effort has currently stalled but is expected to restart next year. The aim should remain to provide a single set of tools for regional model configuration. 128 * A basic but well-tested Configuration Manager (SIREN) is now part of the system. These tools allow a regional sub-model of any ORCA grid to be set up easily. Users will expect such tools for setting up regional models with open boundaries (BDY) and for AGRIF nests; ideally the same tool should be used for both. The UK has made some progress with Python-based tools for constructing regional sub-models from any gridded input. This effort has currently stalled but is expected to restart next year. The aim should remain to provide a single set of tools for regional model configuration.