Periodic Reporting for period 2 - IMMERSE (Improving Models for Marine EnviRonment SErvices)
Période du rapport: 2019-12-01 au 2021-05-31
The Copernicus Marine Monitoring Environment Service (CMEMS) routinely provides data products on the state of the oceans. These data products describe of the present state of the ocean, its evolution over the last decades and predictions of the situation a few days ahead.
In practice, marine monitoring services rely on both observational data and numerical ocean models. Observational data are collected by in-situ sensors and observation satellites as for instance the Copernicus Sentinel series. Numerical ocean models are then used to extrapolate available information at locations and at times where no observational data is available.
In the near future, observation satellites will provide improved observations of the oceans. Upcoming observations will cover a wider range of oceanic properties and will be available with an improved accuracy at finer resolution, which should open many new applications at decision-relevant scales. But a key challenge will be to integrate this variety of new data sources in a physically consistent description of oceanic properties at finer scales (typically ~1km and smaller).
IMMERSE project adresses the question of preparing numerical ocean models to this challenge of providing consistent marine data products at kilometric resolutions for the next generation Copernicus marine services. In practice, IMMERSE project intends to improve the accuracy of NEMO numerical ocean model, which is widely used in the current generation of marine monitoring services.
This is achieved by using the most recent mathematical concepts and high performance computing technologies for solving the equations governing ocean dynamics on a much finer discrete grid. IMMERSE project also prepares NEMO to finer spatial and temporal resolutions by improving the physical representation of the oceanic motions happening at kilometric scale, which include intense fronts, surface waves, tidal currents and strong interactions with the underlying bathymetry.
IMMERSE project then prepares the integration of this improved model code base in Copernicus marine services, in particular by studying how to optimally combine the new high resolution observations with numerical ocean models. IMMERSE project also puts much effort into demonstrating how its new developments will impact Copernicus marine services, its products and its downstream applications.
All IMMERSE developments are delivered in an open source, well-tested and well-documented, easy-to-use and easy-to-interface, improved software code base and in a series of prototype demonstrators for the Copernicus marine services. These demonstrators include a model describing the ocean at 1/36° resolution (2-3km) over the whole globe, which will be deployed by Copernicus marine services by 2025.
In practice, marine monitoring services rely on both observational data and numerical ocean models. Observational data are collected by in-situ sensors and observation satellites as for instance the Copernicus Sentinel series. Numerical ocean models are then used to extrapolate available information at locations and at times where no observational data is available.
In the near future, observation satellites will provide improved observations of the oceans. Upcoming observations will cover a wider range of oceanic properties and will be available with an improved accuracy at finer resolution, which should open many new applications at decision-relevant scales. But a key challenge will be to integrate this variety of new data sources in a physically consistent description of oceanic properties at finer scales (typically ~1km and smaller).
IMMERSE project adresses the question of preparing numerical ocean models to this challenge of providing consistent marine data products at kilometric resolutions for the next generation Copernicus marine services. In practice, IMMERSE project intends to improve the accuracy of NEMO numerical ocean model, which is widely used in the current generation of marine monitoring services.
This is achieved by using the most recent mathematical concepts and high performance computing technologies for solving the equations governing ocean dynamics on a much finer discrete grid. IMMERSE project also prepares NEMO to finer spatial and temporal resolutions by improving the physical representation of the oceanic motions happening at kilometric scale, which include intense fronts, surface waves, tidal currents and strong interactions with the underlying bathymetry.
IMMERSE project then prepares the integration of this improved model code base in Copernicus marine services, in particular by studying how to optimally combine the new high resolution observations with numerical ocean models. IMMERSE project also puts much effort into demonstrating how its new developments will impact Copernicus marine services, its products and its downstream applications.
All IMMERSE developments are delivered in an open source, well-tested and well-documented, easy-to-use and easy-to-interface, improved software code base and in a series of prototype demonstrators for the Copernicus marine services. These demonstrators include a model describing the ocean at 1/36° resolution (2-3km) over the whole globe, which will be deployed by Copernicus marine services by 2025.
The project activities over RP2 have contributed to meeting the project objectives as follows :
- Objective 1: "Develop a new, efficient, stable and scalable NEMO reference code with improved performances adapted to exploit future HPC technologies in the context of CMEMS systems."
The project activities have contributed to reduce the time-to-solution in NEMO applications. The efficiency, stability and scalability of the NEMO code has been improved through : (a) the new RK3 time-step scheme (WP3), (b) tiling of 3D calculations (WP4), (c) loop fusion (WP4), (d) implementation of coarse-grained OpenMP (related to WP4), (e) use of a CFL limiter in vertical advection (WP3). Most of these developments have been delivered in a beta-release of the NEMO code in June 2021.
- Objective 2: "Develop NEMO for the challenges of delivering ocean state estimates and forecasts describing ocean dynamics and biogeochemistry at kilometric scale with improved accuracy”
The project activities have focused on improving the physical consistency of the simulated dynamics in CMEMS systems. We have prepared the NEMO model to better represent high frequency dynamics (including internal waves, fast barotropic motion) and processes at the air-sea interface. This has been delivered in a beta-release of the NEMO code in June 2021. In combination with the improved time discretisation described above, these developments should allow future CMEMS systems to run at higher spatial resolution. The impact of these developments will be tested in a series of prototype systems, that have been defined and implemented as part of the project.
- Objective 3: "Prepare the exploitation of the next generation of high resolution observing networks within CMEMS systems and in detailed, downstream modelling systems."
The project activities carried out in WP7 have prepared the exploitation of high resolution observations in the next generation CMEMS systems. A task has focused on the short term predictability, and developed methods to quantify to what extent high resolution modelling systems are able to propagate information acquired through data-assimilation. Another task has focused on the forecast errors and tested different methods for characterizing the short-range forecast error in assimilative systems. The activities initiated in WP6-8 will also help better exploiting high resolution observations for assessing kilometric scale resolution ocean models.
- Objective 4: "Develop a flexible and generic software tools series for interfacing CMEMS observation and model-based products and detailed, downstream modelling systems"
A specific task in WP7 has been dedicated to prepare a software toolbox for simplifying the use of CMEMS products in downstream systems. A project deliverable (D7.4) describes this toolbox, and demonstrates how it can be deployed on DIAS portals. Objective 4 will also be addressed through WP6 with the IBI-1km demonstrator, which will illustrate how CMEMS systems may use local grid refinement with the AGRIF software in order to meet the users needs for downstream applications.
- Objective 5: "Provide proven model code and software tools with assessments suitable for rapid deployment in CMEMS"
A large fraction of the project deliverables consist in software codes. A key deliverable in this respect is the NEMO code beta-release (v4.2_RC D2.4) which integrates most of the developments to NEMO carried out in WP3-5. The project has also delivered non-NEMO software code. This includes tools for sharing NEMO experiments (D2.2) tools for using CMEMS products in downstream systems (D7.4) and other software related to specific project deliverables (as for instance D7.1). Each of these software tools is distributed with documentation through a dedicated repository on GitHub, in order to allow their rapid deployment into CMEMS and downstream systems.
- Objective 1: "Develop a new, efficient, stable and scalable NEMO reference code with improved performances adapted to exploit future HPC technologies in the context of CMEMS systems."
The project activities have contributed to reduce the time-to-solution in NEMO applications. The efficiency, stability and scalability of the NEMO code has been improved through : (a) the new RK3 time-step scheme (WP3), (b) tiling of 3D calculations (WP4), (c) loop fusion (WP4), (d) implementation of coarse-grained OpenMP (related to WP4), (e) use of a CFL limiter in vertical advection (WP3). Most of these developments have been delivered in a beta-release of the NEMO code in June 2021.
- Objective 2: "Develop NEMO for the challenges of delivering ocean state estimates and forecasts describing ocean dynamics and biogeochemistry at kilometric scale with improved accuracy”
The project activities have focused on improving the physical consistency of the simulated dynamics in CMEMS systems. We have prepared the NEMO model to better represent high frequency dynamics (including internal waves, fast barotropic motion) and processes at the air-sea interface. This has been delivered in a beta-release of the NEMO code in June 2021. In combination with the improved time discretisation described above, these developments should allow future CMEMS systems to run at higher spatial resolution. The impact of these developments will be tested in a series of prototype systems, that have been defined and implemented as part of the project.
- Objective 3: "Prepare the exploitation of the next generation of high resolution observing networks within CMEMS systems and in detailed, downstream modelling systems."
The project activities carried out in WP7 have prepared the exploitation of high resolution observations in the next generation CMEMS systems. A task has focused on the short term predictability, and developed methods to quantify to what extent high resolution modelling systems are able to propagate information acquired through data-assimilation. Another task has focused on the forecast errors and tested different methods for characterizing the short-range forecast error in assimilative systems. The activities initiated in WP6-8 will also help better exploiting high resolution observations for assessing kilometric scale resolution ocean models.
- Objective 4: "Develop a flexible and generic software tools series for interfacing CMEMS observation and model-based products and detailed, downstream modelling systems"
A specific task in WP7 has been dedicated to prepare a software toolbox for simplifying the use of CMEMS products in downstream systems. A project deliverable (D7.4) describes this toolbox, and demonstrates how it can be deployed on DIAS portals. Objective 4 will also be addressed through WP6 with the IBI-1km demonstrator, which will illustrate how CMEMS systems may use local grid refinement with the AGRIF software in order to meet the users needs for downstream applications.
- Objective 5: "Provide proven model code and software tools with assessments suitable for rapid deployment in CMEMS"
A large fraction of the project deliverables consist in software codes. A key deliverable in this respect is the NEMO code beta-release (v4.2_RC D2.4) which integrates most of the developments to NEMO carried out in WP3-5. The project has also delivered non-NEMO software code. This includes tools for sharing NEMO experiments (D2.2) tools for using CMEMS products in downstream systems (D7.4) and other software related to specific project deliverables (as for instance D7.1). Each of these software tools is distributed with documentation through a dedicated repository on GitHub, in order to allow their rapid deployment into CMEMS and downstream systems.
The activities over M13-M30 have mostly consisted in new developments to NEMO code-base based on innovative mathematical concepts for discretising the equations of fluid motions, new insight on the physics of ocean fine scale processes and innovative approaches for leveraging complex computer architectures. All these developments are beyond the state of the art in the field. Activities focusing on the integration of model and observational data (inc. study of predictability and the quantification of forecast accuracy) are beyond the state of the art in terms of targeted resolution. The demonstrator prototypes deployed in the project (inc. the global 1/36° model) are also beyond the state of the art in existing operational systems.