Periodic Reporting for period 1 - ACCIBERG (Arctic Cross-Copernicus forecast products for sea Ice and iceBERGs)
Berichtszeitraum: 2023-01-01 bis 2024-06-30
A new way to detect and forecast sea ice and icebergs in the Arctic
The Arctic ocean will always be treacherous in spite of its warming. Sea ice and icebergs remain major risks for navigation in Arctic waters; therefore, monitoring and forecasting them is crucial. While the Copernicus Marine & Climate Change Services can provide the necessary information, users may not benefit from it as intended due to both a lack of consistency and missing uncertainty estimates. The EU-funded ACCIBERG project will address these issues. It will develop a new iceberg forecast service using remote sensing algorithms, data assimilation and cloud computing to consistently offer probabilistic sea ice and iceberg forecasts based on Copernicus data. The new forecasts will be automated and used by those navigating in the Arctic – from fisheries to cruise tourism.
The Arctic ocean will always be treacherous in spite of its warming. Sea ice and icebergs remain major risks for navigation in Arctic waters; therefore, monitoring and forecasting them is crucial. While the Copernicus Marine & Climate Change Services can provide the necessary information, users may not benefit from it as intended due to both a lack of consistency and missing uncertainty estimates. The EU-funded ACCIBERG project will address these issues. It will develop a new iceberg forecast service using remote sensing algorithms, data assimilation and cloud computing to consistently offer probabilistic sea ice and iceberg forecasts based on Copernicus data. The new forecasts will be automated and used by those navigating in the Arctic – from fisheries to cruise tourism.
WP2 Sea Ice Forecasts
Task 1 has progressed as planned. Most features of the ICECAP software are now implemented, and a final version will be handed over to the project partners at the end of 2024 (Deliverable 2.1). It currently handles medium-range sea ice forecasts from the CMEMS ARC-MFC system and ECMWF, extended-range forecasts from ECMWF, and seasonal forecasts from the C3S seasonal forecasts, with more systems to be added. Investigations to assess the quality of ensemble forecasts of the different systems have already demonstrated the utility of the software to forecast system developers. Work on end-user products is progressing well and has been showcased to stakeholders. Current work will refine useability and portability of the software, decide on calibration methods and first version of end-user product prototypes, finalize the documentation and end-user interfaces.
Task 2 should start in 2025, however, some preparatory activity has taken place over the reporting period. Initial discussions on technical implementation details and management of code repositories have taken place, and the software has been run successfully on the WeKEO platform.
In Task 3, a new version of the ARC-MFC ensemble forecasting system based on TOPAZ5 has been developed. Since November 2023, daily forecasts have been made available to the project partners for early testing and evaluation using the ICECAP software. Retrospective forecasts for the period August - December 2021 will be run in late 2024. These will feed into WP3. For ARC-MFC sea-ice forecasts based on the neXtSIM model, developments have taken place to add the capacity to run ensemble forecasts, and experimentation with that will start in 2025.
Task 4 was planned for 2025 but has started early to exploit synergies with activities at ECMWF and other C3S projects. The code for stochastic sea-ice parameter perturbations provided by CNR has been implemented in the ECMWF seasonal forecasting system. A first set of seasonal forecasts has been run, with good results, but an unexpected change in the mean state of the sea ice requires further investigation. Future work will focus on adjusting the strength of the perturbations to obtain better forecast quality, understanding atmospheric impact, and providing recommendations for implementation of the perturbations in future C3S seasonal forecasting systems.
WP3: Data Assimilation
Task 3.1 on the further development of the microwave satellite simulator, has advanced without blockers. First, the original RTM (Radiative Transfer Model) was retrieved and tested using AMSR2 and ERA5 data. Results indicate significant errors over the sea-ice areas, with an over-estimation of Tbs over multi-year ice and under-estimation over first-year ice areas. Then the RTM was modified where the ice emissivity is a function of two variables: Distance Along the Line and surface temperature (2 m). Providing a more realistic ice emissivity field to the RTM has allowed a significant improvement of the simulated Tbs. The updated RTM has led to the submission on time of M3.1 and paves the way to the completion of D3.1.
Task 3.2 on the Preparation of satellite and auxiliary data, interface of the satellite simulator with EnKF, AMSR2 data has first been projected into the TOPAZ5 spatial grid. For now, the RTM provides the model Tbs at the background ensemble preparation stage before the EnKF analysis. Currently, work on the interface of RTM and EnKF is ongoing, in order to provide a Python package that combines both softwares (the RTM and the EnKF), which will lead to the submission of M3.2 at the end of August 2024.
Initial work has been performed in Task 3.3 on the Experiments and tuning of the DA update. We have performed the first assimilation analysis using the EnKF software with the objective of correcting TOPAZ5 10-day forecast ensemble, by assimilating AMSR2 SIC and Tbs in parallel experiments. First, the updated RTM is used to simulate TOPAZ5 Tbs, and a background ensemble of TOPAZ5 SIC and Tbs is corrected through the EnKF analysis. Initial results indicate that the assimilation of Tbs can potentially provide better corrections than the assimilation of SIC. We will explore the assimilation of AMSR2 individual swaths of TB which should provide further improvements.
WP4: Iceberg Forecasts
Satellite based iceberg detection: DMI has modified the satellite detection algorithm to remove ships based on their AIS data. DMI has also improved the false detection of sea ice as icebergs thanks to an automated new sea ice concentration product. Further developments are ongoing using additional sensors and thermal noise removal. Progress is according to plan.
Validation data has been collected by DMI and sent to NERSC for modeling. Since the ocean models did not reproduce the 2021 trajectories, too close to the coast, older data from 2011 and 2012 in the East Greenland and 1990 in the Barents Sea were processed. Deliverable D4.32 Due June 2024 (Delayed)
The iceberg module has been developed in two steps, first a V0 module with the simplest functionalities (ocean currents and wind effects, D4.1 delivered Dec. 2023) and a V1 module accounting for all processes (waves, sea ice, melting and roll-over) has been tested on past experiments in the Barents Sea and Baffin Bay. There was no real-time demonstration in Summer 2024 but a demonstration is planned for in Summer 2025.
The validation of the iceberg module has proven more difficult than anticipated.
The implementation of OpenDrift on WEkEO has gone smoothly and has been highlighted by the EDITO Model Lab project.
WP5: Dissemination and Communication activities
The project website was set up in advance of the KO meeting. There were some problems in updating the site due to the Drupal client used but these have now been resolved. A LinkedIn page was also set up.
The first ACCIBERG workshop is a combined stakeholder & training workshop (originally planned as separate workshops), on 8th November 2024, Frascati, Italy. It will run directly after the IICWG meeting, therefore taking advantage of the fact that IICWG attendees are key ACCIBERG stakeholders.
The PEDR was delivered 6 months after kick off.
Since the project kick-off, four new members have been added to the Advisory Board. Two newsletters have been sent out so far, with a third planned in July 2024. These are sent to the Advisory Board and a wider stakeholder group, who have a more general interest in the ACCIBERG project.
Task 1 has progressed as planned. Most features of the ICECAP software are now implemented, and a final version will be handed over to the project partners at the end of 2024 (Deliverable 2.1). It currently handles medium-range sea ice forecasts from the CMEMS ARC-MFC system and ECMWF, extended-range forecasts from ECMWF, and seasonal forecasts from the C3S seasonal forecasts, with more systems to be added. Investigations to assess the quality of ensemble forecasts of the different systems have already demonstrated the utility of the software to forecast system developers. Work on end-user products is progressing well and has been showcased to stakeholders. Current work will refine useability and portability of the software, decide on calibration methods and first version of end-user product prototypes, finalize the documentation and end-user interfaces.
Task 2 should start in 2025, however, some preparatory activity has taken place over the reporting period. Initial discussions on technical implementation details and management of code repositories have taken place, and the software has been run successfully on the WeKEO platform.
In Task 3, a new version of the ARC-MFC ensemble forecasting system based on TOPAZ5 has been developed. Since November 2023, daily forecasts have been made available to the project partners for early testing and evaluation using the ICECAP software. Retrospective forecasts for the period August - December 2021 will be run in late 2024. These will feed into WP3. For ARC-MFC sea-ice forecasts based on the neXtSIM model, developments have taken place to add the capacity to run ensemble forecasts, and experimentation with that will start in 2025.
Task 4 was planned for 2025 but has started early to exploit synergies with activities at ECMWF and other C3S projects. The code for stochastic sea-ice parameter perturbations provided by CNR has been implemented in the ECMWF seasonal forecasting system. A first set of seasonal forecasts has been run, with good results, but an unexpected change in the mean state of the sea ice requires further investigation. Future work will focus on adjusting the strength of the perturbations to obtain better forecast quality, understanding atmospheric impact, and providing recommendations for implementation of the perturbations in future C3S seasonal forecasting systems.
WP3: Data Assimilation
Task 3.1 on the further development of the microwave satellite simulator, has advanced without blockers. First, the original RTM (Radiative Transfer Model) was retrieved and tested using AMSR2 and ERA5 data. Results indicate significant errors over the sea-ice areas, with an over-estimation of Tbs over multi-year ice and under-estimation over first-year ice areas. Then the RTM was modified where the ice emissivity is a function of two variables: Distance Along the Line and surface temperature (2 m). Providing a more realistic ice emissivity field to the RTM has allowed a significant improvement of the simulated Tbs. The updated RTM has led to the submission on time of M3.1 and paves the way to the completion of D3.1.
Task 3.2 on the Preparation of satellite and auxiliary data, interface of the satellite simulator with EnKF, AMSR2 data has first been projected into the TOPAZ5 spatial grid. For now, the RTM provides the model Tbs at the background ensemble preparation stage before the EnKF analysis. Currently, work on the interface of RTM and EnKF is ongoing, in order to provide a Python package that combines both softwares (the RTM and the EnKF), which will lead to the submission of M3.2 at the end of August 2024.
Initial work has been performed in Task 3.3 on the Experiments and tuning of the DA update. We have performed the first assimilation analysis using the EnKF software with the objective of correcting TOPAZ5 10-day forecast ensemble, by assimilating AMSR2 SIC and Tbs in parallel experiments. First, the updated RTM is used to simulate TOPAZ5 Tbs, and a background ensemble of TOPAZ5 SIC and Tbs is corrected through the EnKF analysis. Initial results indicate that the assimilation of Tbs can potentially provide better corrections than the assimilation of SIC. We will explore the assimilation of AMSR2 individual swaths of TB which should provide further improvements.
WP4: Iceberg Forecasts
Satellite based iceberg detection: DMI has modified the satellite detection algorithm to remove ships based on their AIS data. DMI has also improved the false detection of sea ice as icebergs thanks to an automated new sea ice concentration product. Further developments are ongoing using additional sensors and thermal noise removal. Progress is according to plan.
Validation data has been collected by DMI and sent to NERSC for modeling. Since the ocean models did not reproduce the 2021 trajectories, too close to the coast, older data from 2011 and 2012 in the East Greenland and 1990 in the Barents Sea were processed. Deliverable D4.32 Due June 2024 (Delayed)
The iceberg module has been developed in two steps, first a V0 module with the simplest functionalities (ocean currents and wind effects, D4.1 delivered Dec. 2023) and a V1 module accounting for all processes (waves, sea ice, melting and roll-over) has been tested on past experiments in the Barents Sea and Baffin Bay. There was no real-time demonstration in Summer 2024 but a demonstration is planned for in Summer 2025.
The validation of the iceberg module has proven more difficult than anticipated.
The implementation of OpenDrift on WEkEO has gone smoothly and has been highlighted by the EDITO Model Lab project.
WP5: Dissemination and Communication activities
The project website was set up in advance of the KO meeting. There were some problems in updating the site due to the Drupal client used but these have now been resolved. A LinkedIn page was also set up.
The first ACCIBERG workshop is a combined stakeholder & training workshop (originally planned as separate workshops), on 8th November 2024, Frascati, Italy. It will run directly after the IICWG meeting, therefore taking advantage of the fact that IICWG attendees are key ACCIBERG stakeholders.
The PEDR was delivered 6 months after kick off.
Since the project kick-off, four new members have been added to the Advisory Board. Two newsletters have been sent out so far, with a third planned in July 2024. These are sent to the Advisory Board and a wider stakeholder group, who have a more general interest in the ACCIBERG project.