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.

June 4, 2021

MINUTES 002/2021 - Revision 004

Meeting of the Scientific Committee of the Unified Earth System Community Model.

At 3:00 pm on June 4, 2021, 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), ITA (Technological Institute of Aeronautics), INPA (National Institute for Amazon Research), UFSM (Federal University of Santa Maria), USP (University of São Paulo), LNCC (National Laboratory for Scientific Computing), UFMS (Federal University of Mato Grosso do Sul) and MB (Brazilian Navy) met virtually, with the aim of discussing and continuing the work of the Scientific Committee of the Community Model of the Unified Earth System (MCSTU). These minutes record the meeting and gather the information entered in the chat, such as links and information relevant to the discussions held. Following the opening of the meeting, conducted by Saulo Freitas, this document is oriented according to the guidelines established by Saulo Freitas during his presentation.

Agenda 1 - Opening

Saulo Freitas announced the new members of the MCSTU Scientific Committee, nominations from the Brazilian Navy (MB), Walid Maia Pinto Silva and Seba (absent) and Flávia Rodrigues Pinheiro, who will be the focal points of the MB in the context of the MCSTU and will bring the needs, demands and contributions of the MB. From UFPA, Júlia Clarinda Paiva Cohen (absent), from USP, Marcia Akemi Yamasoe and from INPE, Karla Longo. Then, the floor was given for the presentation of the new members.

Flávia Pinheiro began her presentation to the MCSTU members and informed them that she had learned about the Scientific Committee the week before the date of this meeting. She explained that, due to the absence of the MB in the first meetings, she would take this opportunity to learn about the issues and discussions within the MCSTU, as this would allow her to identify the points of collaboration and needs of the MB with the MCSTU.

Saulo Freitas agreed and pointed out the possibility of scheduling a specific meeting with MB members to explain in detail how collaboration could occur with the armed forces.

Pedro Dias added that it is important to have feedback from the armed forces, as they have specific needs. Those who work in the operational area know the difficulties that models have in terms of accuracy and that the specificities of this group ( DHN - Hydrography and Navigation Directorate) must be better understood. He adds that there is a lack of manpower and it is necessary to count on the help and experience of specialized people.

Flavia Pinheiro agreed with the idea of scheduling a meeting for next week.

Saulo Freitas commented that all proposals for the MCSTU will come from the Scientific Committee meetings, but that the MCSTU is still under construction. He then gave the floor to Karla Longo, who made her presentation. In her presentation to the MCSTU Scientific Committee, Karla Longo commented on her work on the chemical composition of the atmosphere, with models that describe its chemical composition and its interaction with other systems. She added that she expects to contribute to the MCSTU, which represents the future of modeling in Brazil.

Saulo Freitas announced that during the current meeting, Renato Henrique Ferreira and Cristiane Mariano Zavati, INPE employees, will talk about management techniques for highly complex projects.

Saulo Freitas reported that the Minutes of the 2nd MCSTU Meeting are available. He set aside a few moments for the members of the Scientific Committee to comment on the corrections to the minutes of the meeting. As there were no comments, he declared the Minutes of the 2nd MCSTU Meeting approved.

Saulo Freiras returned to the subject of scientific events and the collaboration program with ECMWF, within the scope of the MCSTU. He highlighted an important issue, about the need to deepen the discussion of the focal members of the MCSTU Scientific Committee, since there are several instances of participation in the MCSTU and the Scientific Committee is a higher instance.

Agenda 2 - Definition of Members of the Committee to Collect and Document the Requirements, Demands and Counterparts of the MCSTU

Saulo Freitas commented on the current members of the MCSTU Scientific Committee and the topics they are associated with. He reported on a meeting between Pedro Dias, himself and an expanded group from the Oceanography area. He added that Júlia Cohen will be responsible for the universities in the North of Brazil and that Flávia Pinheiro and Walid Seba will be the representatives of the MB.

Ronald Buss commented on the participation of the MB, and since there are two representatives, he assumed that one was associated with the ocean and the other, with the atmosphere. He reported that the day before the date of this meeting, a live broadcast was held (live broadcast on the internet) related to the Oceanography Society, where there was a lot of talk about the work and involvement that the MB has with the REMO network (Oceanographic Modeling and Observation Network).

Flávia Pinheiro reported that she is in charge of the Operational Numerical Forecasting division, which has two sections - atmospheric and ocean modeling, which are integrated. She commented that Walid Seba is from the DGN (General Directorate of Navigation) of the MB, and that the DHN is subordinate to the DGN. Therefore, Walid Seba's participation will have an organizational focus on the part of the MB.

Otávio Acevedo commented that he initiated contacts in the southern region of Brazil, where there are two meteorology courses and others in related areas. He added that his impression is that an initial contact should be made and then the distribution by areas of Meteorology. He questioned whether this position is correct or whether it is necessary to find out what the demands of the southern region of Brazil are for the MCSTU.

Saulo Freitas agreed and said that it is important to seek out demands and identify the most important processes in the region that the MCSTU should be able to resolve. He added that the Universities will be important for promoting the MCSTU.

Otávio Acevedo commented that the South region can contribute more to the representation and modeling of the Planetary Boundary Layer.

Saulo Freitas stated that it is important to publicize the information at Universities and identify the regional demands.

Jairo Panetta added to the discussion by telling Otávio Acevedo that it is not enough to consider only the meteorological aspect. He commented that in the South of Brazil there is a PAD (High Performance Processing) group represented by Philippe Navaux/UFRGS , which works in this area and can collaborate with MCSTU. The important thing is the "whole work", which goes beyond the meteorological aspect.

Otávio Acevedo welcomed the information and stated that help will be needed with aspects of the MCSTU that are not his specialty.

Joaquin Costa commented that his participation in the Scientific Committee is in the contribution between the neutral and ionized atmospheres, in the common use of data (even in Data Assimilation) and in the experiments that can be related to the other components of the MCSTU.

Antonio Manzi contributed to the discussion by saying that it is necessary to organize the issue of the local representative with the focal point representing the country (this person must be a facilitator), as PAD is a specific theme, whose collaborators can be in any region.

Saulo Freitas agreed with Antonio Manzi and added that this matter should be discussed.

Haroldo Fraga stated that the issues are complementary. He added that information should be disseminated at universities indicating where the MCSTU needs are, but in general terms (in the areas of Meteorology, Oceanography, PAD, Software Development, etc.). The teams should be multidisciplinary. The issue is broad and everyone should speak the same language.

Luiz Flávio added that some Universities have a clear vocation (e.g., software quality) and that they can be fundamental for the development of MCSTU. He added that no one should be excluded and that everyone should be involved for the best development of MCSTU, as it must be solid.

Jairo Panetta intervened, commenting that the point raised by Antonio Manzi has not been resolved, as there appears to be duplication of duties.

Antonio Manzi added that good coordination is needed to avoid duplication of information.

Pedro Leite added to the discussion by saying that it is important to bring this issue to the coordination team so that a decision can be made on how to handle each case of conflict. He said that contact with some Scientific Societies is very important. He commented that a few years ago, SBMAC (Brazilian Society of Applied and Computational Mathematics) showed interest in climate modeling programs. There was a lack of evidence of a program like this for SBMAC to participate, as it has the potential to contribute.

Joaquim Costa commented that one way to integrate these people is through the role of a liaison. This person participates in meetings with other groups and brings this information to the MCSTU Scientific Committee.

Enio Bueno added that it is necessary to provide people who want to participate in the MCSTU initiative with the information they need. There could be a homepage where people can find information about the MCSTU and find out where they can participate. The minimum that the MCSTU should have and be is already known. People can join certain aspects based on affinity. This helps to organize the groups and people should know that this initiative exists.

Saulo Freitas argued that the role of universities, in the context of MCSTU, is precisely to disseminate information and act on specific topics and that he is concerned about the issue raised by Antonio Manzi. As an example, he commented that in the southern region of Brazil there are people who work with Data Assimilation and that João Gerd is the one who should organize the groups on this topic, even though he is not in this region.

Pedro Dias added that he has started contacting some groups, but that there are still many people to talk to. The first contacts made were with the Oceanography group. Enio Bueno argued that there is no duplication of duties. The problem is when an expert does not know about the MCSTU and cannot contribute.

Saulo Freitas pointed out that each member of the MCSTU Scientific Committee should contact their networks of collaborators and organize collaborations.

João Gerd said that he should talk to Otávio Acevedo so that he can put him in contact with people from the southern region of Brazil, so that they can talk about specific topics.

Haroldo Fraga added a suggestion, and says that Otávio Acevedo is responsible for presenting the project in the Southern region of Brazil and that he will be able to disclose what the focal points are on the themes.

João Gerd pointed out that it would be interesting to formulate a document with the idea of the MCSTU Scientific Committee, informing what is sought in each of the areas so that partners can understand what is being done. In a standardized way, the members of the Scientific Committee can pass on the same information about the MCSTU.

Jorge Gomes agreed with the statements made by João Gerd and Enio Bueno, and commented that the homepage and the document can help to equalize everyone's discourse. Locally, partners must be multipliers. The network will grow locally when MCSTU partners have the necessary information.

Saulo Freitas asked João Gerd to organize a standardized note that could be distributed to the members of the Scientific Committee.

Caio Coelho commented on the topic "weather and climate pre- and post-processing" and said that he may be the focal point, but that his expertise is not in Numerical Weather Prediction. He indicated that Carlos Bastarz and Paulo Kubota are people who have experience in this part of pre- and post-processing for Weather Forecasts.

Saulo Freiras reiterated that Caio Coelho should be the focal point and that he should look for people to collaborate on specific topics.

Ronald Buss added that it is becoming clearer how the topics related to the MCSTU components will be addressed, but that there is still a lack of representation in the area of Hydrology. There is a huge deficiency in Hydrological modeling, from a national perspective (e.g., forecasting river levels, river discharges, etc.). He said that in Brazil there is a network for monitoring hydrological parameters, but that representation of this component is missing in the MCSTU.

Pedro Dias commented that there are some studies in this line, e.g., studies by Walter Collichonn/UFRGS and others at USP as well. He added that the important point in this area is the flooded areas and the coupling between the hydrological and atmospheric models. There needs to be a focal point in this area.

Haroldo Fraga suggested that the Scientific Committee focal points list their points of contact and who are experts in specific areas. The Scientific Committee members need to know who these people are in order to direct the work of the focal points.

Saulo Freitas pointed out that FUNCEME (whose point of contact would be Francisco Hoilton Rios, Technical Director of FUNCEME) is an important entity that can help in this area.

Agenda 3 - Community Model Development: Partnership with PAD Companies - The Case of NVIDIA

Saulo Freitas informed that Luiz Flávio is well advanced in contacting private companies and that he will make a communication about the NVIDIA case.

Luiz Flávio began his talk by commenting on the importance of software development and the contact he has with private companies (e.g., NVIDIA , NEC , BULL , IBM ). These companies are interested in participating in the development of MCSTU. Computer architectures evolve very quickly, and if we don't keep up, we will no longer be able to port our codes to new architectures. He asked what is good for our dynamic core. The private sector has a lot of experience in porting codes (they have a practical interest in optimizing codes) and in adapting models to specific architectures. They also have other commercial interests. Gains with NVIDIA — they have knowledge of the different dynamic cores, they know the difficulties of each dynamic core, and they have expertise in using current and future GPUs. NVIDIA's success story — IBM's GRAF (Global High-Resolution Atmospheric Forecasting System) model, completely ported to GPU (with data assimilation and the dynamic core of MPAS — Model for Prediction Across Scales). This product is a competitor from the private sector. They have experience in heterogeneous partnerships — Research Centers, Operational Centers and University (eg, ESCAPE project ). NVIDIA proposed 30 minutes of lectures for the members of the MCSTU Scientific Committee.

Saulo Freitas asked about the advantages of using GPUs over CPUs, from a cost and coding point of view.

Luiz Flávio responded by saying that it depends on what is going to be coded for the GPU. He said that in CUDA language it is like porting Fortran to C language. He added that numerical models, when ported to GPU, have very expressive results in performance.

Saulo Freitas asked again about the level of intervention in the code. He asked Luiz Flávio if he knows how many additional lines of code are needed to port code written for CPU and then for GPU.

Luiz Flávio responded with the example of OpenACC where the modifications are few. Starting from a dynamic core ported to GPU, it is simpler because it will be necessary to port the physics and do the same work that was done with the dynamic core.

Jairo Panetta added to the discussion by arguing that a decision must be made on this aspect: will MCSTU's development efforts be based on OpenACC? He said that it is debatable whether the future will maintain OpenACC. It may be that the OpenMP target will take the place of OpenACC. The effort of this conversion is certainly not small.

Saulo Freitas expressed concern about the rapid developments in computer architecture, in view of important decisions being made for the development of MCSTU.

Jairo Panetta commented on the example given by Luiz Flávio saying that it is not just a mapping between OpenACC and OpenMP and that it is necessary to discuss the subject very well. The biggest concern is about which software structure should be adopted. It is necessary to know what centers like ECMWF , NCAR , NCEP , etc. will do.

Luiz Flávio agreed with Jairo Panetta's statement about OpenACC and stated that it is essential to connect with companies that are at the forefront of these developments. Companies are interested in contributing because they want to sell their own machines and the models need to have good performance on them.

Pedro Peixoto commented on his experience at MetOffice and ECMWF. They develop software in layers, one of which is the scientific part of the model and the other is the coupling (separation of concepts). As examples of projects that use separation of concepts, he mentioned the projects LFRic (from MetOffice) and PSyclone (component of the GungHo project ).

Haroldo Fraga added to the discussion and said that interventions and team organization will depend on the choice of computing structure that will require fewer interventions. It is not yet clear what this structure will be, but we have to count on the one that will require fewer interferences in the code.

Pedro Peixoto responded to the last question asked by Saulo Freitas (about the rapid evolution of computer architectures and the MCSTU), saying that it is necessary to develop software structures that allow flexible adaptations, using the separation of concepts.

Saulo Freitas commented that this is why a specific community training program is needed.

Pedro Dias added that if the right choices are not made, the chances of failure are high. He commented on Brazil's competence in this area and that we can contribute with international partnerships and the expertise that the country has.

Enio Bueno agreed with Pedro Peixoto. In the area of Physics, he stated that there are many people who are specialized, but who are not experts in Scientific Computing. Therefore, they will deliver good physics for the computing people to optimize.

Jairo Panetta added that the problem goes deeper, e.g., if you write a radiative transfer model like RRTM-G (Rapid Radiative Transfer Model for GCMs), with a single column, there is no way to make OpenACC or OpenMP optimize that code to run on GPUs. All the physics of the model must work independently, across multiple columns.

Agenda 4 - Organizational Project Management at INPE - Example from Monitora-BR

Saulo Freitas introduced Renato Branco and Cristiane Zavati, from the Project Management and Technological Innovation Coordination — COGPI of INPE. He invited Renato Branco to his presentation.

Renato Branco presented examples of project management at the Institute and how INPE is responding to the demand for restructuring of the MCTI (Ministry of Science, Technology and Innovation). Much of what has been done within INPE helps in the organization of internal projects, in addition to meeting the MCTI's own demand. In INPE's organizational structure, the COGPI (intellectual property) monitors the entire Institute and organizes a technology development roadmap, in line with what MCTI expects. With this, INPE unfolds this roadmap according to its strategic planning (in terms of products, services and results). As an example for MCSTU, the structure of the "Monitora-BR" project was presented, a component of the "Biomas-BR" project (alongside the "BIG" (INPE Georeferenced Information Base) and MCSTU projects; Biomas-BR should be one of INPE's structuring projects). Within "Monitora-BR", there are four components that can be financed with resources from the GCF (Green Climate Fund).

Saulo Freitas explained that, in this context, the GCF is a funding program from the Ministry of Economy and could be a possible source of resources for the MCSTU. He asked Renato Brando about how to monitor the execution of the project and what kind of support COGPI can provide for its effective implementation.

Antonio Manzi intervened and highlighted that the MCSTU project aims at software development, and that we need specific management of this platform, that everything has to work in harmony and this aspect is very important.

Renato Branco responded to Saulo Freitas and stated that there is no way for COGPI to manage MCSTU in any of these aspects. What COGPI can do is present the methods and the way in which the MCSTU Scientific Committee can develop its project. It is up to the Scientific Committee to seek out the systems and partners that will help manage the project. We can create a routine to periodically monitor the project, but the management should not be done by COGPI.

Saulo Freitas stated that MCSTU needs a guide on how to apply these management tools in practice, so that it is possible to see how the project is progressing, what the weak points are and what needs to be done to improve.

Pedro Dias added that the issue of software management is something that needs to be addressed. He stated that he believed that INPE's COGPI would also manage software.

Renato Branco highlighted that it is possible that in a few years this management could be carried out by BIG, once it is established. At INPE, there is no entity or coordination that carries out this management or that has established this standard.

Roberto Souto added that in his experience, software such as Trello and Redmine can be used for project management and software development. For communication between members, Slack and Discord can be used .

Renato Branco added that Trello is used to manage project actions. Trello is used to manage requirements and actions, Redmine is used to organize delivery packages, and Subversion and Tortoise are used to version Microsoft Word files.

João Gerd added that CPTEC is already used to using Subversion and Redmine for software versioning and project management, respectively.

Luiz Flávio argued that he has been concerned about these aspects, and has already requested the inclusion of Git in Redmine at CPTEC . The reason for this is that the community around MCSTU should be large and Git allows local commits, allows better control of revisions and avoids problems. Agile methods such as Kanban should come before these tools. He stated that he is writing about this and that we have to worry about software development principles. It is a community model, that there will be groups and they need to distribute their tasks, otherwise we do not need a community. This community needs software development principles and everything needs to be very well controlled.

Renato Branco added that the legacy is important, but that the MCTI tools are an imposition, and outside of it, there is nothing to be done.

Joaquim Costa stated that what Renato is saying is that there is a template to include information, but this does not solve the problem that Saulo Freitas exposed, which is the need for program management.

Saulo Freitas concluded and thanked everyone for their contribution and presence.

Actions for the Next Meeting

Presentation of a standardized document for communication between focal points in their regions;
Request the Scientific Committee focal points to provide their expert contacts in the various components of the MCSTU;
Provide for the construction of an MCSTU homepage to aggregate information about the model and, where possible, identify possible areas of collaboration for participants;
Invite FUNCEME to participate in MCSTU in the area of Hydrology;
Discuss the software structure to be adopted, given the development options aimed at new architectures (separation of concepts, OpenACC, OpenMP Target);
Seek means for managing MCSTU projects and software, in compliance with MCTI requirements.

Attachments