AI-Translated Page

This page was translated using artificial intelligence (AI). The content may contain inaccuracies or misinterpretations. For critical or official use, please consult the site administrator to confirm the information.

March 28, 2022

MINUTES 001/2022 - Revision 002

Meeting of the Scientific Committee of the Ocean, Land Surface, and Atmosphere Forecasting Model. At 9:00 a.m. on March 28, 2022, representatives of INPE (National Institute for Space Research), INMET (National Institute of Meteorology), UFCG (Federal University of Campina Grande), CENSIPAM (Management and Operational Center of the Amazon Protection System), USP (University of São Paulo), LNCC (National Laboratory for Scientific Computing), UFMS (Federal University of Mato Grosso do Sul), FAB (Brazilian Air Force), and MB (Brazilian Navy) met virtually with the aim of continuing the work and discussions of the Scientific Committee of the Ocean, Land Surface, and Atmosphere Forecasting Model. These minutes record the meeting's minutes and gather the information posted in the chat, such as links and other information relevant to the discussions held. Following the opening of the meeting, led by institutional coordinators Saulo Freitas/INPE and Pedro Dias/USP, this document is oriented according to the guidelines established by Saulo Freitas during his presentation.

Opening

Saulo Freitas welcomes participants to the 7th meeting of the Scientific Committee of MONAN - Model for Forecasting the Oceans, Land Surfaces and Atmosphere. He lists the meeting's agenda, including the announcement of researcher Gilberto Bonatti /INMET as a new member representing INMET on the Scientific Committee (CC) of MONAN, the communication by researcher Gilvan Sampaio of the General Coordination of Earth Sciences ( CGCT / INPE ), a retrospective of the actions carried out to date and those that are in progress, the presentation of the report of the High Performance Processing (PAD) subcommittee, the report of the oceans and cryosphere subcommittee and the seminar by researcher Ariane Frassoni /INPE on the activities of the Working Group on Numerical Experimentation ( WGNE ) of the World Meteorological Organization ( WMO ).

Communication - Gilvan Sampaio/INPE

Saulo Freitas invites Gilvan Sampaio to update the members of the MONAN CC on the latest developments related to the prospecting of financial resources for the implementation of the MONAN project.

Gilvan Sampaio thanks Saulo Freitas and greets the members of the CC. He begins his presentation by announcing the prospecting of a new resource for the renewal of INPE's supercomputing infrastructure and for the development of MONAN, based on a project approved by FINEP , entitled "Renewal of INPE's Supercomputing Infrastructure and its application in the advancement of Forecasts and monitoring of weather, climate and environment in the country". He mentions some of the project's goals, such as the expansion of supercomputing capacity, updating of the data storage system, support infrastructure, and an associated photovoltaic power generation plant. For MONAN, he mentions the construction of the modeling coupling system, involving all components of the Earth system, the development of the model's operationalization system at INPE and partner institutions in production mode for the time scale of hours, days, subseasonal and seasonal with global and regionalized spatial domain over South America and adjacent oceans. It adds that the amount of resources is R$200 million, of which R$60 million will be for 2022, R$50 million for 2023, and R$45 million for 2024 and 2025. These resources come from the National Fund for Scientific and Technological Development ( FNDCT ), which is not subject to contingency and must be executed as it has already been approved. It reports that these resources are already being awaited. It adds that a committee has already been established to specify the new supercomputer, among other actions. It also adds that for next year, a specific line will be opened for MONAN within the INPE budget. He explains that the idea is to have a specific MONAN program within the National Council for Scientific and Technological Development ( CNPq ), which should be negotiated so that INPE's financial resources (the expected amount is R$5 million) can be transferred to MONAN for granting scholarships at any level and in any place in the country where the model can be developed (universities and research institutes in Brazil). Another action taken refers to a meeting with the representative of the São Paulo State Research Support Foundation ( FAPESP ), within the scope of global climate change research. As a result of this meeting, he emphasizes that MONAN can and should be submitted as a thematic project (in the range of R$30 million). He informs that FAPESP has offered to establish partnerships with other Research Support Foundations in other states (FAPES), also depending on the partnerships of the members of the MONAN CC, as representatives in their respective states. This is important as it will allow bilateral calls to be opened (eg, FAPESP and FAPEAM). In addition, there is the possibility of coordination with other international organizations, such as the Newton Fund , among others. These actions and coordination can make it possible to increase the budget for the development of MONAN. It emphasizes that there are other possibilities for prospecting resources and that community coordination is very important. It also provides information about the National Meteorological Network, whose decree is being reviewed by the Civil House and should be published soon. This decree is also important because MONAN is one of the central elements of development and in the generation of operational forecasts.

Agenda 1 - Retrospective and Actions

Saulo Freitas thanks Gilvan Sampaio for his communication. Starting the presentation of Agenda 1, he gives a retrospective of the meetings held in 2021. He recalls that the CC was established by the Director of INPE on April 8, 2021, with community participation as one of its fundamental premises. He mentions some of the participating institutions such as INPE, INMET, LNCC , INPA , CENSIPAM , Department of Airspace Control ( DECEA ), Brazilian Army EB , Brazilian Navy ( MB ) and several universities, among its current 35 members. He recalls that 6 CC meetings have already been held, as well as 4 subcommittee meetings (surface, atmosphere, PAD and oceans and cryosphere). He comments on the MONAN Project Opening Term (TAP) within the organizational structure of the Ministry of Science, Technology and Innovation ( MCTI ), which was approved by INPE with an initial duration of 10 years. It also mentions the development and application of MONAN as one of the pillars of INPE's strategic planning, the agreement with the European Centre for Medium-Range Weather Forecasts ( ECMWF ) from which it will be possible to install a version of the Open Integrated Forecasting System ( OpenIFS ) to serve as a benchmark for MONAN, in addition to access to the data series produced by the centre, seminars and training. In addition, the agreement between INPE and ECMWF also foresees the joint development of a system for visualising unstructured grids.

Saulo Freitas mentions the efforts made to prospect for financial resources. In addition to the news reported by Gilvan Sampaio, Saulo Freitas reports that INPE will allocate R$ 1.5 million from its budget to purchase a system based on a Graphics Processing Unit (GPU) for the development of MONAN. He reports that this machine will be acquired before the new supercomputer, which should help in prospecting the dynamic core and physics for the development of the MONAN component models, not only for INPE but also for the regional centers. He also reports that there was a meeting with the Central Bank and that, although there has not yet been a decision, there is the possibility of financing personnel through the Central Bank.

Saulo Freitas mentions, during the retrospective of the actions carried out in 2021, the definition of the name of the model MONAN - Model for Ocean-laNd-Atmosphere predictioN**** , whose meaning, in short, is "land without evils". He adds that Luiz Flávio developed a logo for MONAN , a Tupi-Guarani symbol whose meaning is the word union. He mentions that the logo features the main colors that represent the components of the Earth system (e.g., the green of the forests and the brown of the land and sand on the Earth's surface, the blue on the ocean surface and the sky in its various shades).

Saulo Freitas presents details about the goals of the CT-INFRA 2021 Action , in reference to Gilvan Sampaio's communication (title: "Renewal of INPE's supercomputing infrastructure and its application in meeting the growing demands of Brazilian society for better forecasts and monitoring of weather, climate and environment"). He mentions that among the 7 goals, 3 are related to MONAN, namely:

1. Construction and coupling of the MONAN modeling system and testing of stability, efficiency and correctness of results;
2. Availability of the developed code in a public area with free access for the national and international community;
3. Operationalization of the community model of the MONAN Earth system at INPE on a global and regional scale for weather forecasting.

He adds that the total budget for the CT-INFRA Action is R$200 million, with approximately R$18 million distributed among these three goals, which must be achieved through the hiring of legal entities, which is a condition for the allocation of personnel for the development of the model. A public notice must be issued by FUNCATE , which will include the qualification requirements, level of education and experience of human resources.

Saulo Freitas also comments on the actions being taken to reorganize INPE around the MONAN program. He mentions that it is a tradition at INPE to have several groups, each with its own model, and that in this type of organization, one difficulty is communication. He mentions that the Numerical Modeling Division of the Earth System ( DIMNT /INPE) is being organized around the MONAN model and that the groups are being reorganized around their components, instead of being organized by models. Within the scope of CGCT/INPE, he lists the divisions and their various activities and attributions (e.g., DIMNT, Satellite and Meteorological Sensors Division - DISSM , Earth Observation and Geoinformatics Division - DIOTG , Impacts, Adaptation and Vulnerabilities Division - DIIAV and Weather and Climate Forecasting Division - DIPTC ) in relation to the development of MONAN. DIMNT should lead the modeling effort; DISSM should lead the data production part, mainly to feed the MONAN data assimilation system; DIOTG should provide the necessary data for MONAN to be the "Digital Twin of the Earth" (e.g., Working towards a Digital Twin of Earth ), in the sense that MONAN's forecasts are a realistic representation of the Earth system; DIIAV should take care of the representation of anthropogenic changes on Earth; DIPTC should produce the numerical forecasts of MONAN and be the communication channel with the CC through the model performance reports in relation to the references used in the development of the model. Within the scope of CGCT, the groups will be organized into macrogroups (e.g., atmosphere, oceans, PAD, etc.) and each of these macrogroups corresponds to a MONAN subprogram, with specific attributions and TAP.

Saulo Freitas concludes the first agenda of the meeting, considering the country's economic situation as a relative success for the beginning of the development of MONAN, which is important for the country's strategic planning. He invites Pedro Dias to comment.

Pedro Dias emphasizes the importance of seeking financial support from FAPES. He mentions that the meeting with FAPESP shows that without a thematic project it will be more difficult to have a specific funding program for MONAN. He also mentions the difficulties of seeking funding from traditional agencies such as CNPq, whose contribution is currently very limited to national level programs. He concludes by saying that the resources are available from FAPES and that it is essential to establish connections between them.

Saulo Freitas adds that the surface subcommittee led by Antônio Manzi already has some coordination with the Santa Maria/RS group to create a thematic project.

Enio Souza asks Pedro Dias and Saulo Freitas about the documentation needed to prepare a proposal. He says that there is already some experience between FAPE in his region and FAPESP. Pedro Dias responds that the document sent to FINEP could be used because it contains a lot of information about MONAN. Enio Souza asks about the project page where all the information about the MONAN project can be found. Saulo Freitas responds that it has not yet been possible to create a page due to the various articulations that are being made in the project, in addition to the coordination of DIMNT at INPE. He adds that there is a DIMNT page, but that there is not yet a page for MONAN. Enio Souza suggests that the documentation could be gathered by Fabielle Alves and sent by email to the leaders of the subcommittees.

Agenda 2 - Report of the PAD Subcommittee

Luiz Flávio begins his presentation by discussing the initial tests performed with the dynamic cores that are being verified for use in MONAN. Luiz Flávio informs that there are two groups within the PAD subcommittee. The first group, formed by INPE's native members - Luiz Flávio, Roberto Souto, Denis Eiras , Eduardo Khamis and Carlos Renato and the second group, formed by transversal members. He informs that the group has held weekly meetings with the purpose of organizing all the aspects related to the software infrastructure for the development of MONAN. He informs that a page has been created on GitHub and that this platform was chosen because it is suitable for the development needs of the model (an open platform that allows all members who want to develop or verify the code). He mentions that the page contains the standards that are being developed and applied in the developments, e.g., criteria for choosing the dynamic core, development standard, etc., in addition to the description of the tasks of each group of the PAD subcommittee (e.g., coding and testing). He adds that, at this moment, these standards are being defined and that some are already well defined, with the collaboration of native and transversal members.

Luiz Flávio presents the MONAN development flow . He informs that this flow is already being applied and that the development methodology is based on the Agile Method (eg, Scrum(not yet defined). He adds that training in this area should be carried out and that these courses are included in the model's budget. Everyone involved in the development of MONAN should participate in the training, not just the PAD subcommittee. He emphasizes that everyone's participation is important because this should improve the necessary articulations for the technical development of the model. Regarding the model's development flow, he explains that there is a main repository and that each developer can create a branch (i.e., a code branch) to carry out their own developments. From this branch, with their developments completed, the developer makes a pull-request (i.e., a request for delivery of the code) which will be submitted to a code review subgroup. This subgroup, in accordance with the normative document pertinent to code standardization, carries out the review stage. If problems related to the writing of the code are found, the developer who made the pull-request is contacted and is guided by the subgroup depending on the complexity of the changes. In the next stage, the software testing subgroup, in accordance with the normative document relevant to the testing standard, performs the necessary software operation tests. Luiz Flávio explains that in this stage, the software is subjected to tests that indicate whether it produces the expected results. Once approved in the testing stage, the next step is to activate the PAD subgroup which, in accordance with the normative document relevant to high-performance processing, performs the necessary tests and assessments. In this stage, the developer may also be called upon to make interventions in his code, depending on the complexity of his developments. With the approval of the proposed developments, the PAD subgroup activates the code versioning subgroup which, in turn, consolidates the modifications in the model's main repository and creates a release candidate. The release candidate version is then delivered to the center's operations that will perform the operational tests of the code. Luiz Flávio emphasizes that these tests are important, because it is in this stage that the code will be tested in production mode and operating environment, which differs from the tests performed in the previous stages. Once the release candidate version has been tested in an operational environment, the version is consolidated as a release of the model. This is the stable version that can be used by the MONAN community for applications and other developments. In addition to the stable version of the code, there may also be development versions that can also be used and tested by the community. Luiz Flávio emphasizes that, in addition to the subgroups involved in the MONAN code development flow, there is also a central management that will monitor all stages of the development process. In this first stage of the work,The PAD team is working with codes from other projects, but the interest is to have the dynamic core of these projects but with the physics that have already been developed in the country.

Luiz Flávio comments on the process of choosing the dynamic core of the model. He says that an objective methodology was established based on criteria of quality, functionality, reliability, usability, efficiency and maintainability. He highlights that, in relation to functionality, the Model Evaluation Group, led by researcher Ariane Frassoni , should establish which phenomena and cases will be reproduced by MONAN. The other criteria are being applied by the PAD group. He adds that the evaluation of the code's maintainability is more advanced, and that this evaluation indicates the degree of ease with which it is possible to modify the code. He argues that these criteria are important to be evaluated due to the community characteristic of the model, which must involve many people and that the code must be understandable for everyone. The methodology and results of the evaluations can be found at https://github.com/monanadmin/monan/wiki . Other aspects highlighted by Luiz Flávio involve portability and reliability. Regarding these aspects, Luiz Flávio highlights the need for the developed code to be able to be executed on CPU ( AMD , Intel ), Graphics Processing Unit (GPU), hybrid machines (which combine the use of CPUs and GPUs) and vector machines (e.g., TSUBASA ). In addition, the software requirements are also being verified, i.e., whether the code works with MPI , MPICH , oneAPI , OpenMP , OpenACC , etc. Still on the software issue, there are also compilers that can be used with the codes, i.e., NEC , Intel , NVIDIA / PGI , GNU and highlights that Intel, NVIDIA and GNU are open and that the priority is to evaluate the codes with these compilers. For the efficiency and reliability issues, he reports that the Global Eta Framework ( GEF ), Finite-Volume Cubed-Sphere Dynamical Core ( FV3 ) and Model for Prediction Across Scales ( MPAS ) models were evaluated. All codes were compiled and executed for the evaluations. Among the models mentioned, the MPAS model has the highest score.

Luiz Flávio invites Roberto Souto to his part of the presentation. Before starting his presentation, Roberto Souto answers Haroldo Fraga's question about whether there is any intention to test the use of processors with the Advanced RISC Machine ( ARM ) architecture. Luiz Flávio responds that the idea is to test with all processors available on the market and mentions that the ARM architecture is a major competitor. He adds that there is not yet a machine with this architecture available for testing, but that some tests may be performed on machines from Fujitsu and NEC.

Roberto Souto begins his presentation by mentioning the tests performed with the MPAS and FV3 models on a cluster provided by Dell and on the Santos Dumont supercomputer at LNCC . Roberto Souto mentions that the MPAS code used is available on GitHub and that this code was used in the tests. He adds that the most recent stable release ( 7.3 ) was not used because the previous stable version ( 6.3 ) is the one compatible with GPUs. He mentions that there is a branch of version 7.x under development for GPUs and that it is not yet stable. The tests were performed to simulate the atmosphere using GPU processing and OpenACC instructions. The data used for the tests varied in horizontal resolution, between 120 and 10 km. He adds that in the tests, the memory consumption was measured for each spatial resolution. In the tests, he mentions that there was a limitation of using only one node and that this limitation is being verified. Regarding the tests performed, considering the atmospheric part of the models with 30 km of horizontal resolution and 56 vertical levels (which requires 92 GB of memory), it presents comparisons between the execution of the code with only CPUs and with CPUs and GPUs combined. Regarding the configurations of the test environments, it mentions that the Dell Rattler cluster and the Santos Dumont supercomputer were used. The Rattler cluster has two partitions, the first configured with 2 Intel Xeon Gold 6334 CPUs, with 16 cores and 250 GB of RAM each and 4 NVIDIA A100 PCIE GPUs (PCIE is a communication protocol) with 40 GB of memory; the second configured with 2 AMD EPYC 7543 CPUs, with 32 cores and 250 GB of memory each and 4 NVIDIA A100 SXM4 NVLINK GPUs (NVLINK is a faster communication protocol than PCIE) with 40 GB of memory. The Santos Dumont supercomputer was used from the Sequana partition, configured with 2 Intel Xeon Gold 6252 CPUs with 24 cores and 188 GB of RAM each and 4 NVIDIA V100 SXM2 GPUs with 32 GB of memory. Roberto Souto mentions that the amount of RAM of the GPUs is an important characteristic for the execution of the models. Regarding the results (Table 1), he comments that the execution time of the tested model was higher in the tests that considered only the use of CPUs, and that the best performance was obtained with the Rattler cluster using AMD processors. Regarding the tests with GPUs, which performed better (i.e., shorter execution time) compared to CPUs, he mentions that NVIDIA processors were the best, also in the Rattler cluster (the test results can be viewed from minute 50'50" of the meeting recording video).

Table 1 - Benchmark of the MPAS-A model (30KM_L56, atmosphere) using CPUs and GPUs on the Rattler cluster and Santos Dumont supercomputer.

Partition	Architecture	MPI ranks	Time (s)
Rattler gpuq	Intel Xeon 6334 CPU	CPU:16	12.00
Rattler gpuq	GPU A100-PCIE	GPU:4	1.48
Rattler xe8545	AMD EPYC 7543 CPU	CPU:16	9.14
Rattler xe8545	A100-SXM4 GPU	GPU:4	1.10
SdumontSequana	Intel Xeon 6252 CPU	CPU:16	15.28
SdumontSequana	GPU V100-SXM2	GPU:4	2.14

Roberto Souto explains that the initial tests were performed for short runs and the time step of the model dynamics was measured, without the radiation part. He explains that these initial tests were done this way because the radiation call does not occur in the same time step as the model dynamics call when processing is done by CPUs and GPUs. He adds that only the dynamics are executed on GPUs. Saulo Freitas asks Roberto Souto what the time step of the MPAS-A dynamics tested is. Roberto Souto informs him that it is 180 seconds.

Roberto Souto, regarding the FV3 code, reports that similar tests were performed by Eduardo Khamis/INPE, based on the code made available in the GitHub repository . He reports that the tests were conducted using the systems (models) System for High-resolution prediction on Earth-to-Local-Domains ( SHiELD ) and the UFS Short-Range Weather Application ( UFS-SRWA ). In the initial tests, he reports that during the use of the SHiELD container on the Dell clusters, there were execution problems and that this is being investigated.

Saulo Freitas thanks the presentation and congratulates the scientific computing group for the work carried out and opens the floor for the other members to ask their questions.

Luciano Pezzi comments on the initial efforts with the dynamic core of the atmospheric model and asks about the plans to deal with the oceanic component of MONAN. He comments that the choice of the model for the oceanic component must already be more advanced, but that he has doubts about the scalability and adaptation of this component to the computational tests that are being carried out with the atmospheric component. He adds that he has doubts about whether the codes for the oceanic component are already at the same level of development and testing as the atmospheric component. He concludes by saying that if MONAN is a model of the Earth system, the development of its components must be in sync and that a strategy is needed for this. Another question is about what type of coupler will be used and that it is also necessary to have a computer professional dedicated to this task, who should also help direct the development efforts of the oceanic component of MONAN.

Luiz Flávio comments that he has been in contact with NVIDIA, who have agreed to make their development team available for whatever is necessary in the development of MONAN. He mentions that a former researcher from the National Center for Atmospheric Research ( NCAR ) was hired by NVIDIA to join the company's development team, which is focused on applications similar to MONAN. He adds that this strategy of the company is important to assist the MONAN development team and that it should also help in the development of the oceanic part of the model. On the other hand, he comments that he still does not know which coupler will be used.

Haroldo Fraga contributes to the discussion by saying that proximity to other developer teams, whether from other models or from NVIDIA, should help the MONAN CC to identify the parts that need to be developed and should also help to identify the people who should work on certain developments.

Pedro Dias comments that the Modular Ocean Model ( MOM ) has incorporated some features of the HYbrid Coordinate Ocean Model ( HYCOM ), and that this should be a candidate model to represent the oceanic component of MONAN. He adds that, because the MOM model is developed by NCAR, this should also facilitate the choice of the model for the oceanic component.

Saulo Freitas argues that there should be a list of priorities to help the MONAN CC in its choices. In this order of priorities, he mentions that first, MONAN must meet the physics requirements that are necessary and at the scales of interest to the project; second, the choice of computational architecture; and third, the way in which the coupling between the components of the Earth system will be performed. He comments that, following the discussions of the oceans and cryosphere subcommittee, the group must decide on the modeling system that will be used according to the needs of representing the physics that it wants to solve and at the scales of interest. He adds that, in his view, the coupling system is a technological issue and that it must be discussed so that the components can communicate harmoniously within MONAN.

Agenda 3 - Report of the Subcommittee on Oceans and Cryosphere

Saulo Freitas invites Ronald Buss to give a presentation on the discussions within the subcommittee on oceans and cryosphere. Ronald Buss begins his speech by mentioning the cryosphere and the seasonality of the Arctic and Antarctic ice sheets, which are important processes that need to be understood. He mentions that, currently, there is a consensus that it is better to work with ice area rather than its extent, since it is a variable that does not depend on the size of the model grid and is estimated by satellites. In ocean circulation models, Brazil has several experienced groups spread throughout the country and that these are focused on specific processes (e.g., estuarine processes, shelf processes, transport and dispersion of hydrocarbons or sediments). Regarding modeling, from a climate perspective, he points out that the number of groups dedicated to this subject is even smaller. He reports that, for the first meeting of his group, he tried to bring these groups together and adds that he also invited some researchers from Argentina, but they have not yet been able to participate. He mentions that, among the Brazilian researchers, there was the participation of around 40 researchers. He adds that there was a delay in holding this first meeting because there was a need to clarify what the delivery priorities for this component are for MONAN.

Ronald Buss reports on the main topics discussed at the first meeting of the oceans and cryosphere subcommittee. Among the participants of this first meeting, he highlights the participation of groups from INPE, Federal University of Rio de Janeiro ( UFRJ ), Oceanographic Observation and Modeling Network ( REMO ), Oceanographic Institute of USP ( IOUSP ), Federal University of Maranhão ( UFMA ), Petrobrás , Federal University of Santa Catarina ( UFSC ), Federal University of Rio Grande ( FURG ) and MB. Other representatives from universities whose professors were unable to attend also participated in the meeting (Federal University of Pernambuco ( UFPE ), Federal University of Espírito Santo ( UFES ), Federal University of Paraná ( UFPR ) and other universities). Regarding the topics discussed during the subcommittee meeting, it mentions the need to work with a focus on processes and variables at multiple time and space scales, the importance of wave models, the consensus on the concept of the Community Earth System Model ( CESM ) from NCAR that executes the MOM6 model (it mentions the experience of the group led by researcher Paulo Nobre /INPE), on the integration of the efforts of the Decade of the Oceans , on the need to have a coupled data assimilation system, on the multi-scale aspect in a global and regional model, on the training of human resources, on who will be the model's clients and on the development of a questionnaire to assess priorities.

Saulo Freitas, after Ronald Buss' report, invites the other members of the MONAN CC to the discussion.

Afonso Paiva comments on the possibility of making a presentation on the conception, experience, achievements and challenges faced by REMO. Ronald Buss appreciates Afonso Paiva's comment and mentions that there is an intention that the meetings of the oceans and cryosphere subcommittee always begin with a seminar on the various subjects discussed within the subcommittee. He comments that REMO's experience is very important for the development of the activities within the subcommittee. Saulo Freitas adds that it is important that this REMO presentation be made within the MONAN CC meeting and that it is necessary for the atmospheric and oceanic modeling groups to have closer communication.

Luciano Pezzi thanks Ronald Buss for the report on the meeting held by the oceans and cryosphere subcommittee. He takes the opportunity to reinforce the representation of some groups within the subcommittee and mentions the participation of the groups represented by researcher Clemente Tanajura /UFBA and researchers Carlos Lentini /UFBA and Luiz Felipe Mendonça /UFBA. Luciano Pezzi, in addition to Ronald Buss' report, lists some of the challenges of the oceans and cryosphere subcommittee, namely: waves and feedback with the atmospheric boundary layer and the role of waves in mixing the oceanic boundary layer and mentions researcher Alexander Babanin from the Natural Ocean Engineering Laboratory ( NOEL ) as a collaborator and his role in developing this topic together with the MOM model. He adds that data assimilation is also one of the challenges to be addressed in the subcommittee's activities and that interaction with the data assimilation subcommittee team will be necessary. He comments that one of the issues to be addressed is how assimilation should be done in the coupling between the ocean and the atmosphere, whether separately, i.e., in each separate component, or in a coupled manner, i.e., in both coupled components.

Pedro Peixoto, via chat, asks whether the MPAS-Ocean model was considered in the discussions of the oceans and cryosphere subcommittee. Saulo Freitas, who also participated in the subcommittee meeting, responds that this version of the model was not considered in the subcommittee's initial discussions. Saulo Freitas takes advantage of Pedro Peixoto's question to suggest that the oceans and cryosphere subcommittee also consider this model in its discussions.

João Gerd reports that the data assimilation subcommittee, which he heads, is also considering the oceanic component in the discussions. He adds that, although there is the impression that the atmosphere is the main component of the discussions, the oceanic component is being considered in the development strategy outlined by the data assimilation subcommittee. He mentions that there is an alignment with Clemente Tanajura regarding the discussions on the topic and that there are even plans to bring their work closer to CPTEC, in order to promote integration between the atmosphere and ocean working groups on the topic of data assimilation. Ronald Buss comments that he also spoke with researchers Natália Ruddorf /INPE and Luiz Sapucci /INPE about the assimilation of satellite data over the ocean. He also comments on the use of observational data from institutional agreements in model validation studies, which is a routine activity, and the use of these data within the scope of operational activities, which is not well established within INPE. He adds that Luiz Sapucci has already been in talks with the MB so that data from institutional agreements can also be used in this context. Luciano Pezzi also comments on the information provided by João Gerd, saying that the exchange of experiences is important and that he is willing to discuss the different aspects of data assimilation with this subcommittee.

Luciano Pezzi comments on his experience in data assimilation with the Local Ensemble Transform Kalman Filter ( LETKF ) and the computational cost associated with it. He mentions past experiences in the use of LETKF and the uncertainties and possibilities regarding the use of data assimilation algorithms for the oceanic component of MONAN. João Gerd comments on the work developed in data assimilation by researcher Clemente Tanajura and how the data assimilation subcommittee can take advantage of it. He also mentions the integration of the MOM6 model with the Joint Effort for Data assimilation Integration ( JEDI ) system and how these experiences can be integrated and used for the development of MONAN.

Joaquim Costa comments on his work in the area of ​​space weather and how this area can be related to the verticalization of meteorological models and the interactions between the neutral and ionized atmosphere. Saulo Freitas adds that the discussions of the MONAN CC need to include aspects of upper space and that it is necessary to bring in more representatives from this area to improve the discussions. Joaquim Costa adds that there are ongoing discussions to hold a workshop with researchers who work with the verticalization of models.

Saulo Freitas invites Flávia Pinheiro to comment on Ronald Buss's presentation. Flávia Pinheiro argues that the oceanic component may not require a very complex data assimilation scheme, as reported by Luciano Pezzi's experience. She mentions that the HYCON model uses the simple Optimal Interpolation scheme and that this is a scheme that has shown good results. More complex schemes that require greater computational power, such as ensemble-based schemes, are necessary for atmospheric models due to the rapid changes in the atmosphere. This requires updates of the models' covariance matrices, which can be calculated from the ensemble members. Other schemes such as 4DVar also require an adjoint model and its update, which also increases the complexity of data assimilation. Flávia Pinheiro adds that coupled data assimilation in ocean-atmosphere models is a major challenge and that this topic is still a research topic and that it is necessary to improve the interaction between research and development groups in the areas of atmospheric and oceanic modeling.

Afonso Paiva comments on the topics of ocean modeling and data assimilation. He says that the data assimilation system he uses in his work group is even simpler than the one used by REMO. He emphasizes that when one says that the results obtained with the data assimilation scheme obtained by REMO are good, it means that they are good for the objectives that were set by REMO, i.e., short-term forecasting. He emphasizes that the ocean model should be considered on the same level of importance as the atmospheric model, not only in terms of complexity, but also in terms of objectives. He adds that, in the same way that one is looking for the best model, one should also look for the best data assimilation system. He adds that the MOM and HYCOM models also have data assimilation systems and that they can be considered in this process, aiming not only at the same objectives as REMO, but also at the new objectives to be set by MONAN.

João Gerd argues that the data assimilation system to be chosen depends on the types of data to be used in the assimilation process. Flávia Pinheiro comments that one possibility is ensemble-based systems and that variational systems may be more complicated in the future. Luciano Pezzi comments that his experience with LETKF was not bad, but that it simply did not achieve the expected performance. Carlos Bastarz, via chat, adds that the decision on the type of data assimilation system to be adopted depends on the types of observations to be assimilated and that simpler systems are not necessarily worse than more robust ones, such as variational ones, but that they meet specific applications.

Pedro Dias comments that the time scale to be considered in the ocean and atmosphere is an important factor and that the coupling between ocean and atmosphere is dependent on this factor. He argues that MONAN predictions are for the Earth system, i.e., not just for the ocean or just for the atmosphere. He adds that it is very difficult to point out which is the best or worst system due to these factors. He adds that the system to be chosen should be the one that is most viable for certain applications and that ensemble-based systems are more attractive.

Haroldo Fraga comments that an operational forecasting system, as intended for MONAN, must have a data assimilation system and that it must be mastered by the developers. He adds that when a new type of observation is made available, the team must be able to include it in the data assimilation process. He cites the example of the Sheffield University Plasmasphere-Ionosphere Model SUPIM for predicting ionosphere dynamics that works with the nudging method (Newtonian Relaxation), which was one of the first data assimilation schemes to be established. Regarding the results presented by Roberto Souto, he questions whether the tests were performed without the model's radiation. Luiz Flávio responds that the radiation used in the MPAS model is the Rapid Radiative Transfer Model for GCMs ( RRTMG ) and that it has not yet been ported to GPUs and that NVIDIA has been working on this.

Other Matters

Saulo Freitas comments with Gilberto Bonatti about the need to discuss MONAN within the scope of INMET and bring INMET's needs and requirements for discussion at the CC.

Ariane Frassoni briefly presents a survey of the data assimilation groups reported to the WMO regarding the types of operational ocean data assimilation systems in use (valid for the year 2017). She mentions that there is a project led by ECMWF that aims to understand the systematic errors of the models associated with ocean initialization. She adds that more information about this project can be found on the WGNE website.

Pedro Dias comments on the quality of the discussions that have been taking place during the CC meetings. He says that the number of people participating in the CC is still small in relation to the requirements that have already been listed for the development of MONAN, but that there is an expectation that, as the discussions and developments are made, more people will be involved. He mentions that the topic of coupled data assimilation is still a new topic and that it should be possible to include more people, especially during the training of staff in postgraduate programs. He adds that the topic of data assimilation is important and that undergraduate programs are not yet fully prepared to train students in this area. In this regard, he comments that it is also the role of the CC members to prepare their students so that they can work in the area.

Haroldo Campos announces a mini-course on data assimilation to be offered during the second week of July 2022.

Actions for the Next Meeting

1. Request Fabielle Alves to send the MONAN documentation (project sent to FINEP) so that the subcommittee leaders can gather the necessary information for their discussions with FAPES;
2. Identify the development needs of the model components and the possibilities of collaboration with other development teams (i.e., other models and companies such as NVIDIA);
3. Presentation by researcher Afonso Paiva on REMO's history, activities, strategies and experience in modeling for the MONAN CC;
4. Propose the inclusion of other space weather representatives to collaborate in discussions between the upper and lower atmosphere;
5. Presentation by researcher Ariane Frassoni on the activities at WGNE/WMO.

Attachments

• Slides Saulo Freitas
• Recording of the MONAN Scientific Committee Meeting - March 28, 2022
• Presentation by Luiz Flavio
• Presentation by Roberto Souto
• Presentation Ronald Buss