2019 ESIP Winter Meeting

The Council of Data Facilities (CDF) is committed to working with relevant agencies, professional associations, initiatives, and other complementary efforts to enable transformational science, innovative education, and informed public policy through increased coordination, collaboration, and innovation in the acquisition, curation, preservation, and dissemination of geoscience data, tools, models, and services. Existing and emerging geoscience data facilities – through the Council – are committed to serving as an effective foundation for EarthCube. The General Assembly meeting is open to the official representatives from all member data facilities, additional member organization personnel as desired by the members, as well as observers.

A quick primer on all things ESIP to help you navigate the meeting and the community! Come meet other new and returning ESIP members and ESIP leadership. 

View live-stream here: ESIP 2019 Winter Meeting - Day 1 Plenaries

View live-stream here: ESIP 2019 Winter Meeting - Day 1 Plenaries

Bio: Lesley Wyborn is an Adjunct Fellow at the National Computational Infrastructure and the Research School of Earth Sciences at the Australian National University. She also works part time for the Australian Research Data Commons. She previously had 42 years’ experience in scientific research (geochemistry and mineral systems research)  and in geoscientific data management in Geoscience Australia from 1972 to 2014. In geoinformatics her main interests are developing international standards that support the integration of Earth science datasets into transdisciplinary research projects and in developing seamless high-performance data sets (HPD) that can be used in high performance computing (HPC) and cloud environments. She is currently Chair of the Australian Academy of Science ‘Data in Science Committee’ and is on the American Geophysical Union Data Management Board. She was awarded the Australian Government Public Service Medal in 2014 and the Geological Society of America Career Achievement Award in Geoinformatics in 2015.
 
Abstract:
Just as a Philharmonic Orchestra comprises four sections (strings, woodwind, brass and percussion) so too does the discipline of Earth and environmental sciences comprise four main subdisciplines of ‘players’: geochemistry, geophysics, Earth Observation and geology.  Each section in an orchestra has its own distinctive sets of instruments with their own sound, the sheet music and musical notation they read are common across all four. Likewise, in Earth and environmental science research, data and information are fundamental to all four subdisciplines, although each collects different types of data and uses different suites of ‘instruments’ to collect data then store and process it. The real grand challenge is how do we enable full integration of data across all these subdisciplines and beyond, where we reach out to those beyond the traditional sciences. If we can orchestrate and ‘tune’ our Earth and environmental data systems into harmonious ‘symphonies’ that can address the new transdisciplinary drivers, particularly the needs of low to middle income countries, then we will automatically increase the reach, use and value of our data.

Christina Bandaragoda joined the University of Washington in 2013. She received her PhD in Civil & Environmental Engineering, Master’s of Business Administration, and Master’s in Biological & Agricultural Engineering from Utah State University, and a BS from Wheaton College. Prior to obtaining her graduate degrees, she worked in the National Park Service and studied International Development with extensive travel in Asia and the Caribbean. She provides hydrologic modeling services to multi-institutional watershed groups, and maintains professional relationships through sponsored projects with agricultural and tribal science communities in the Pacific Northwest.

Christina’s research specializations are in the linkage between water resource management and theoretical physical hydrology – using numerical modeling and software development to communicate about flood, drought, and future water scenarios. Her current projects include flood model integration for Prediction and Resilience Against Extreme Events (PREEVENTS), the Landlab open-source Python modeling toolkit, and HydroShare, an online collaborative platform for sharing water data and models.

Christina is currently leading an NSF RAPID team that has recently proposed to build software infrastructure to prevent disasters, with a pilot to address the current humanitarian crisis in Puerto Rico after Hurricane Maria. To learn more about her area of work and collaborators, visit freshwater.uw.edu. You can also discover data and models she has shared on www.hydroshare.org, including HydroPoetry.

View live-stream here: ESIP 2019 Winter Meeting - Day 1 Plenaries

Dan Pilone is CEO/CTO of Element 84 and oversees the architecture, design, and development of Element 84's projects including supporting NASA, the USGS, Stanford University School of Medicine, and commercial clients. He has supported NASA's Earth Observing System for nearly 13 years; currently acting as Chief Technologist for the NASA EOSDIS Evolution and Development 2 contract. He has supported transitioning NASA’s PB scale archive to the cloud, contributed to metadata standards, led multiple working groups on data services and cloud architectures, authored studies on architecture and transition plans for cloud-native data management solutions, and helped shape software development processes for both government and commercial clients.

View live-stream here: ESIP 2019 Winter Meeting - Day 1 Plenaries

Session Abstract:
Sustaining our data resources is part of the bigger issue of sustaining all research products. As we contemplate technologies to enable this, we must also consider organizational and financial enablements. The Science Gateway Community Institute recently conducted a session to work on characterizing 11 ESIP related projects in terms of concepts borrowed from entrepreneurship. This panel will focus on discussion of some of those results and their applicability in a broader sense to sustainable data infrastructures and repositories.

Session Takeaways (post-meeting):
1) The Science Gateways Community Institute (SGCI) aims to connect resources and data to science authors by providing a gateway catalog of gateways related to teaching, research, and learning.
2) SGCI hosts a Science Gateways Bootcamp which is a 5 day bootcamp to help participants develop the framework for: generating pitches, developing communication skills, bringing an idea to fruition, selling the idea, setting goals, and producing deliverables.
3) Bootcamp pitches aim to service the needs of a variety of communities (ex. teachers, crop map stakeholders, ESIP, repositories, and data providers), improving data accessibility and usability within these communities.

Session Abstract:
This session will discuss cross-disciplinary data discovery and introduce the EarthCube Data Discovery Studio project, which currently indexes over 1.6 mil datasets and other geoscience resources from 40+ repositories. The DDS system relies on a scalable metadata augmentation pipeline designed to improve and re-index metadata content using text analytics and an integrated geoscience ontology. The addition of automatically-generated ontology-anchored keywords enables faceted browsing and lets users navigate to datasets related by additional characteristics, such as measured variables, equipment, science domains, or geospatial features. The system also publishes the metadata using schema.org markup, and lets users validate or invalidate automatic metadata enhancements using a custom metadata editor. In addition, we will demonstrate how DDS portal users can invoke Jupyter notebooks residing on one of several Jupyterhubs, and pass discovered document metadata to the notebooks for additional visualization, analysis or modeling, thus bridging cross-domain resource discovery with more in-depth data exploration. We will also show how users can contribute their own notebooks to process additional types of data indexed in DDS.

Session Takeaways (post-meeting):
1) Data discovery studio does not host any data and does not intend to replace repositories. They intend to ‘enhance’ metadata quality.
2) This is a valuable resource for multi-disciplinary projects because it acts as a central area to locate available datasets for a given topic even if they cover a diverse range of datasets.
3) Changes made in Data Discovery Studio do not go back to the original repository, they only exist in the studio. Additional work would need to be done to make those changes at the repository level.

Session Abstract:
This session will cover and invite discussions on challenges in implementing and managing a Cloud environment. Topics will include costs, collaboration, data stewardship, usability, and governance in Cloud environment.

Presentations:
NASA Archives and Data Stewardship in the Cloud – Lauren Frederick, Cumulus Scrum Master, NASA EED-2

Increasing Security through a Collaborative Model, Kevin Mentzer, Sr DevOps Engineer, Element84

Cloud costing and governance, Dede Dascalu, CEO Stratus Solutions

Panel
· Dan Pilone, Chief Technologist EED-2
· Dede Dascalu, CEO Stratus Solutions
· Jeff de La Beaujardiere, PhD, Director, NCAR/CISL Information Systems Division

Session Takeaways (post-meeting)
1) New open-source systems building on Amazon Web Services (leveraging AWS Lambda) allow huge earth-science datasets to be stored in the cloud. Example: NASA Cumulus software to support EOSDIS data archive.
2) Considering how organizations control access to cloud components is crucial for building a productive data-management workflow with AWS, while protecting from errors and data loss.
3) Cloud environments can create lots of complexity in governance, data security, and financial management, but also create an easier environment for managing and using petabyte-scale data archives across the Earth Sciences.

Session Abstract:
Researchers, data repositories, publishers, funders, and other stakeholders are increasingly obligated to ensure that data are not only open, but also FAIR - findable, accessible, interoperable, and re-usable. Various initiatives, projects, and working groups are working to advance the implementation of the FAIR principles, among them the AGU project "Enabling FAIR Data", the European GOFAIR initiative, and Working Groups within the RDA and WDS. This efforts are making it increasingly clear that many aspects of FAIRness, specifically reusability, are highly context and domain specific. The original FAIR guiding principles explicitly point to “domain-relevant community standards (R1.3.)”, but there are many open questions regarding such domain-relevant community standards: Who develops and maintains them? Who has the authority to approve them and govern them? How granular do domain-specific standard definitions need to be?
This session is intended to foster a dialog between data repositories, researchers, and other stakeholders to clarify and address the above listed questions for the Earth, Space, and Environmental informatics community.


Agenda:

1. Welcome & Introduction (Kerstin Lehnert)
2. The role of the International Science Unions in endorsing standards for interoperability (Lesley Wyborn)
3. Domain metadata (Ted Habermann)
4. Discussion in breakout groups
5. Synthesis and explore next steps

Session Takeaways (post-meeting):
1) Interoperability and Reusability are driven by community, and conventions are an important part of communities.
2) Interesting analogies with power adapters and rail gauges. As long as I’m within one country, I have a standard and everything works together. It’s only when I cross into other countries that interoperability enters in.

Session Abstract:
This is the first session of our new Community Resilience ESIP cluster. We aim to introduce a broad landscape of community resilience from an earth science data perspective, as well as gain an understanding of the session participants’ interests/experiences with community resilience. The session includes a series of short presentations and small break out groups to gather feedback from participants. Overall, the session examines how data and information can be a catalyst for overcoming social barriers in communities to help them realize and work collaboratively towards broadly appreciated, overarching goals, like the enhanced sustainability of their food and/or energy system. Place-based community resilience examples include the assessment and mitigation of community vulnerability across coastal communities and tribal groups. Institutional community resilience examples from governmental organizations, NGOs, and Private Sector entities demonstrate how efforts are best aligned in the context of their long established missions. We explore how the ESIP Federation can step into a leading role toward more effective and efficient use of the Nation’s earth science data and information within a community resilience context.

Speakers:

• Arika Virapongse, Middle Path EcoSolutions: “Introduction”: https://doi.org/10.6084/m9.figshare.7622009.v1
• Jonathan Blythe, Bureau of Ocean Energy Management (BOEM): “When are Data Management Best Practices Not Sufficient to Support Agency Processes? An Institutional Perspective on Community Resilience and the Role of Data in Federal Decision making”: https://doi.org/10.6084/m9.figshare.8009594.v1
• Lauren Showalter, The National Academies of Science, Engineering, and Medicine: “Large Scale Data Management and its Role in Improving Community Resilience in the Gulf of Mexico”: https://doi.org/10.6084/m9.figshare.8009582.v1
• Rupu Gupta, New Knowledge: “Balancing Data and Community Needs in Collaborative Resilience Efforts”: https://doi.org/10.6084/m9.figshare.7688804.v1
• Natasha Udu-gama, Thriving Earth Exchange, AGU: “Data Sharing and Management for Community Resilience: Insights from Community Science”: https://doi.org/10.6084/m9.figshare.8009600.v1
• Ranalda Tsosie, University of Montana: “Using Modern Technologies to Inform a Diné Community About Water Quality Issues” (NO SHOW)

Session Takeaways (post-meeting):
1) Data can help overcome social barriers to resilience measures and also drive decision making. GIS is a common and useful format.
2) There are challenges to creating complete and robust data sets to drive resilience. This can be due to proprietary constraints, siloing issues, and social barriers. More work is needed to overcome these challenges.
3) At-risk and front line communities are especially in need of good hazards data and need help to ensure it meets their needs.

Session Abstract:
Round-table discussion about the UMM models, answer metadata questions from users, and gather any feedback for future improvements to the models based on user needs. Briefly discuss what's new with the models.

The UMM model documentation can be found at https://wiki.earthdata.nasa.gov/display/CMR/CMR+Documents

Meeting Notes can be found at https://tinyurl.com/yau45sqq

Session Takeaways (post-meeting):
1) The UMM models are continuing to evolve based on user needs and mission requirements.
2) There should be a regular schedule for UMM releases.
3) As UMM-S records mature, there needs to be UMM-S equivalent records go to the CMR.

Session Abstract:
Members of ESIP's Education Committee invite attendees who have a working knowledge of Jupyter technologies to share information and examples that can be used in an introductory workshop for educators.

Session Takeaways (post-meeting):
1) Working in a “hub” environment instead of a local environment can be beneficial. It enables users to engage with resources without spending time and space downloading programs.
2) The ESIPhub is a Jupyterhub that has been created as a workspace for the ESIP community. It’s possible to use many different programming languages within the hub, but Python is one of the most common.
3) The Education Committee is actively developing Jupyter notebooks within the ESIPhub for use in classroom educational environments as programming learning tools for students.

Session Description:
Are you providing data services? Do you know what your users would say about using your services? If you answer yes to the first question but no to the second, you will want to join this session!

At the National Center for Atmospheric Research (NCAR), a new suite of data system tools is being developed and collectively called the NCAR Digital Asset Services Hub (DASH). It is crucial that the DASH services help the users in achieving their data goals regardless of whether or not they are familiar with data repositories. Additionally, it is vital that the user interactions are friendly and intuitive, so that the users will not mind coming back!

During this session, the attendees will have the opportunity to learn about the techniques that can be applied to a data repository, see evaluation examples and results from DASH, and practice how to evaluate the usefulness of a data repository by using the newly created DASH Repository as the test case.

Session Takeaways (post meeting):
1) Usability is generally important to making useful software, especially in the sciences, and it is extensively oriented around testing with individuals.
2) Key technique for usability testing: Heuristic Evaluation (evaluating software against standard design principles). This technique can be very helpful when testing interfaces mid-development. Demonstratively effective with NCAR/UCAR DASH as use case.
3) Usability is a unique problem from a developer standpoint. Developers are balancing a lot of disparate interests but we can get developers working on the side of usability by talking to developers.

This panel will focus on scalable architectures for implementing machine learning and AI models across Earth Sciences focus areas. The specific details would include both HPC and cloud-centric operational pipelines for automated data processing, parallel multi-GPU model implementations, cluster scheduling, resource optimization and time to delivery - a few tools of discussion can focus around DC/OS, Terraform/Nomad, and Kubernetes. In addition to compute constraints, the panel also encourages discussion on ML framework optimizations and operational complexity, including available frameworks and workflow tools like Tensorflow, Mxnet, Pytoch, Google ML pipeline, AWS Sagemaker and others. GPU hardware optimization is another topic of interest where the panel will focus on how new generation GPU’s (with more CUDA cores) are able to perform faster training as opposed to old generation GPUs.

Session Abstracts:
This session encourages use of semantic technologies which further the use and value of earth science data through linkage, interoperability and standardization. Sessions content will be organized as follows

Speaker: Dr. Dalia E. Varanka, Research Geographer. U.S. Geological Survey.

Title: Advances in Geospatial Semantics and Ontology of the National Hydrography Dataset
Abstract: The study of surface water is a primary global priority, leading to the development of many perspectives about the way that representational data is collected, and how surface water interacts with other geographic phenomena. It is sometimes unclear how these perspectives reconcile to enable collaborative work and data sharing. A study of the semantics of the National Hydrography Dataset (NHD) of the U.S. Geological Survey has resulted in a formal specification of NHD as Web Ontology Language (OWL) files, a form of Resource Description Framework (RDF). The motivation for publishing NHD as RDF is for its topographic mapping context; RDF offers the opportunity to easily integrate hydro data with other geospatial features, such as transportation, structures, or administrative units for base map data applications.  The semantics of these large, national databases were originally standardized in field survey observations of the nineteenth and twentieth centuries and converted to feature-based digital databases. Updated data model designs inscribed NHD semantics as geographic information systems (GIS) products. An automated mass-conversion of GIS attribute table data to RDF of sample NHD datasets resulted in RDF linked data. The ‘bottom-up’ approach of defining NHD semantics as prototype ontologies began with those converted data with manually re-aligned categories along upper ontology principles using Basic Formal Ontology (BFO). This first version was called the Surface Water Ontology (SWO) and was developed for comparative research, including hydro ontology alignment.  Other USGS-funded research has contributed to solutions to efficiently handle coordinate geometry objects as linked open data. Ontology development is in progress for the National Hydrography Dataset Plus Elevation – High Resolution (NHDPLus HR) that combines NHD integrated with LiDAR elevation and watershed boundary data for applications at multiple, including local, geographic scales (USGS). This set of ontologies involves a surface water hydrology module that can be aligned along domain reference ontologies such as Hydro Foundational Ontology (HyFO); a geospatial science module for spatial concepts, and the specifications of the GIS-based NHDPlus HR data model for feature instances.
Speaker(s): Dr. Chris Lynnes, NASA EOSDIS System Architect & Mr. Doug Newman, NASA EOSDIS - Raytheon  
Title: Smart Hand Offs & Earthdata Search
Abstract: We will describe the concept of the ‘smart handoff’. This idea leverages the schema.org’s searchAction concept using simple markup to facilitate the transfer of user context between tools and services to provide an efficient workflow from discovery right through to acquisition. Finally, we will talk about how this concept could be expanded to other tools and services.

Speaker: Dr. Chris Lynnes, NASA EOSDIS System Architect & Mr. Doug Newman, NASA EOSDIS - Raytheon 
Title: Google Dataset Search & CMR
Abstract: Describing the work NASA Earthdata CMR has done to leverage Google Dataset Search capabilities with respect to CMR HTML landing pages leveraging schema.org markup. We will demonstrate the markup we exploit and the way it is used to discovery and visualize results in Google Dataset Search. Finally, we will touch on areas where we think our interaction with Google Dataset Search can improve and grow.

Speaker: Dr. Lewis J. McGibbney, Data Scientist, Jet Propulsion Laboratory/California Institute of Technology
Title: Publishing Geospatial Data as Linked Data: Graph Processing Techniques for Automated Feature Detection and Resolution within Hydrography GIS Products
Abstract: Interesting, largely unexplored data analysis and information retrieval opportunities exist for GIS data. In their current form, traditional data usage patterns for data persisted in shapefiles or spatially-enabled relational databases are limited. Opportunities exist to achieve ESIP’s Winter 2019 theme of ‘increasing the use and value of Earth science data and information’ by transforming geospatial data from their original formats into their Resource Description Framework (RDF) manifestation. This work establishes an innovative workflow enabling the publication for Geospatial data persisted in geospatially enabled databases (PostGIS and MonetDB), ESRI shapefiles and XML, GML, KML, JSON, GeoJSON and CSV documents as graphs of linked open geospatial data. This affords the capability to identify implicit connections between related data that wasn't previously linked e.g. automating the detection of features present within large hydrography datasets as well as smaller regional examples and resolving features in a consistent fashion. This previously unavailable capability is achieved through the use of a semantic technology stack which leverages well matured standards within the Semantic Web space such as RDF as the data model, GeoSPARQL as the data access language and International Resource Identifier’s (IRI) for uniquely identifying and referencing entities such as rivers, streams and other water bodies. In anticipation of NASA’s forthcoming Surface Water Ocean Topography (SWOT – https://swot.jpl.nasa.gov) mission, which once launched in 2021 will make NASA’s first-ever global survey of Earth’s surface water, this work uses Hydrography data products (USGS’s National Hydrography Dataset and other topically relevant examples) as the topic matter. The compelling result is a new, innovative data analysis and information retrieval capability which will increases the use and value of Earth science data (GIS) and information.

Session Takeaways (post-meeting):
1) Adding geospatial metadata could allow for spatial queries among datasets, and is currently underutilized. Leveraging implementation of new large datasets may help rectify this.
2) End to End user-ship within geospatial data can integrate both the ESIP’s COS as well as google's non-spatial parsing.

Session Abstract:
As data, software and services have citations there are other objects that do not fall easily into these categories. These “Research Objects” can include ontologies, workflows, physical samples, instruments, etc. As these are used as elements in one’s research they should be properly attributed to and referenced, thus citable. This session will discuss and try to quantify what research objects are, how do they effect the research community and how can the current citation formats be used or adjusted to suit a wider breadth of objects.

Agenda:

1. Intro and explanation about history of citing research Objects (Mark Parsons, Rensselaer Polytechnic Institute)
2. Work done by Force11 on software citationand what's next (Dan Katz, Univ. Illinois at Urbana-Champaign -- remote)
3. Identifiers for physical samples -- lessons learned and plans for IGSN2 (Kerstin Lehnert)
4. The Digitial Object Interface Protocol (DOIP) (Larry Lannom, CNRI--presented by Mark Parsons)
5. Discussion. What is a research object? What should be cited? What are the general use cases to consider.

Notes at https://docs.google.com/document/d/18ooEixbchKp-qgAG7qtnebKrWDYsutt4d2eX3HsaWls/edit?usp=sharing

Session Takeaways (post-meeting):
1) A grand challenge of citation is determining the purpose of citation: Credit? Access? Reusability? Interoperability?

Session Abstract:
There are many efforts underway to map the landscape of projects in earth science, including Research Data Alliance, EarthCube, ESIP and the RISCAPE project in Europe. This session will provide the current status of many of these efforts. We would also like to use this session to describe how ESIP meeting participants can contribute to the mapping effort during the Winter Meeting.
Ideally this session would be scheduled on Tuesday to give meeting participants to add to the landscape that be available during the meeting for additions and corrections.

Agenda:

• Welcome & Introduction (Rebecca Koskela)
• RDA Mapping the Landscape Interest Group (Steve Diggs)
• ESIP/RDA Earth, Space, and Environmental Sciences Interest Group (Lesley Wyborn)
• ESIP - The Market Landscape (Erin Robinson)
• Mapping RDA, EarthCube (and other) Landscapes with SuAVE (Ilya Zaslavsky)
• Group Discussion

Session Takeaways (post-meeting):
1) There are a TON of communities within ESIP and related organizations. Mapping the working relationships between these groups can help researchers and the public understand the work that is being done.
2) Tools for visualizing both Graph Data (Kumu maps) and survey data (SuAVE) can be valuable for interacting with metadata on the Earth science data community.
3) Mapping the links between ESIP communities, projects, working groups, collaborators, etc. can make the program more understandable and tractable to new members, funding organizations, and the public.

Session Abstract:
This session will introduce the Marine Data Cluster with brief presentations from chairs and founding members, followed by an open discussion on what the Marine Data Cluster should focus on in the first year. What problems will we tackle? What are the most important Marine Data topics on which this group can focus its efforts?

A few examples:

• NetCDF-CF compliance
• Preserving calibration information
• Software portability and reusability
• Best practices for managing 24 Hz CTD data

The main goal of this session will be to determine the most useful 2 or 3 topics to begin addressing in the first calls. We will also brainstorm the infrastructure and support needs and opportunities of the cluster.

Marine Data Cluster wiki: http://wiki.esipfed.org/index.php/MarineData

Session Takeaways (post-meeting):
1) Winter 2019 ESIP launch of the Marine Data Cluster, geared towards addressing advancements, challenges, and collaborations as it pertains to data management in the marine geosciences.
2) The interests of Marine Data Cluster in terms of actionable short-term goals are broad yet interrelated, spanning standardization of vocabulary, parameter reconstruction, and data harmonization across sub-disciplines.
3) The topics to begin addressing in 2019 are: NetCDF-CF compliance, best practices for managing software implementation, and best practices for managing specific marine data types.

Session Abstract:
This session presents the latest results from the OGC innovation program Testbed-14, which addressed several topics that are directly inline with this year's ESIP theme of "increasing the use and value of Earth science data and information". The OGC innovation program is the experimentation lab of OGC, where new technologies are explored that enhance earth observation research and science. Testbed-14 addressed among others topics new standardized approaches to access and exploit satellite swath data (level 1 and 2); explored enhanced, resource-oriented, discovery and access mechanisms for geospatial data and information that reduces the burden on clients, as data can be browsed in addition to being queried; developed an architecture to execute any type of application physically close to Big data; explored quoting and billing mechanisms for geospatial products; developed an architecture to integrate machine learning and artificial intelligence into earth science workflows.
Complementing the discussion on satellite data, the session will address considerations and best practices for organizing data and metadata in HDF5 files when serving earth science data products in Amazon S3 via OPeNDAP. The two talks will be devoted to the best practices for organizing data and user-defined metadata including CF conventions in HDF, interoperability with netCDF-4 and its role in accessing data in HDF5, OPeNDAP considerations for S3 access to HDF5 data.

Presentations

• Ingo Simonis, Luis Bermudez: OGC Innovation Program Testbed-14 Results for Earth Science Community
• Elena Pourmal, James Gallagher: OPeNDAP and HDF5 in the Cloud: Techniques and best practices for serving HDF5 data 
• Aleksandar Jelenak, Peter J. T. Leonard, Charlie Zender: NASA Dataset Interoperability Recommendations for Earth Science.

Session Takeaways (post-meeting):
1) OGC innovation program Testbed-14’s has completed a series of activities to standardize handling of earth observations, to promote the marketplace for EO applications across clouds, and integrate machine learning/AI into OGC web services; Testbed-15 call for participation is out now (bit.ly/Testbed-15).
2) It is important to produce earth science datasets in HDF5 with the optimized setting and version to ensure the interoperability with netCDF4.
3) NASA Data Interoperability Working Group has generated practical recommendations for increasing the usability of earth science datasets.

Session Abstract:
NASA's Cloud Analytics Reference Architecture (CARA) Working Group was established at the 2018 Earth Science Data Systems Working Group (ESDSWG) annual meeting to define a reference architecture enabling discovery and analysis of big data in cloud environments. Current key activities include gathering use cases for cloud analytics, and development of the reference architecture document. The goal of this working session is to inform the ESIP community of the CARA WG's efforts to date, and solicit review and contributions to the use cases and reference document.

Session Takeaways (post-meeting):
1) Many use cases exist as to how to best use cloud services to do big data analysis. Using these the ESDSWG is building a common architecture based on these use cases.
2) Building this architecture, requires community members to integrate the works that have been done previously.
3) Identifying and assembling the building blocks for technology components that enable the evolution of data-driven, science-informed resilience plans that come with full traceability from decisions back to the underlying data and models.

Agenda: 
State of Federation

1. President’s Report
2. Staff Report
3. Looking ahead to 2019
4. Q&A
5. Updates from other community groups: EarthCube, OGC, Core Trust Seal, Metadata 2020 and RDA 

Assembly Business

1. Governance Update 
2. Modification to Partnership Policy 1.2 (http://bit.ly/ESIP_NewPartner)
3. Software Citation Guideline Adoption (http://bit.ly/ESIPSoftwareCitation)
4. Caucus for Type rep positions to administrative committees - Finance, Partnership, Governance 

Session Abstract:
This session will convey the process for GCMD keywords - fast track and yearly reviews through the ESDIS Standards Office, and how Earth science users can influence keyword additions and modifications. The session will also highlight how keywords facilitate the discovery of EOSDIS data and services.
There will also be an opportunity for attendees to ask questions about the keywords and discuss how you would like the keywords and the supported tools to evolve.
Please use the Google document below to post your keyword questions and feedback. We will review them during the meeting.
https://tinyurl.com/y9w7suym 

Session Takeaways (post-meeting):
1) GCMD keywords are used by numerous national and international organizations.
2) There is an interest in normalizing measurement terms in the GCMD KMS for the UMM-Var model.
3) There are some gaps in the keywords, especially within Biology that need to be addressed. The GCMD team will be getting input from SME's in that area.

Session Abstract:
The Machine Learning Cluster was formed following the ESIP 2018 Summer Meeting. Since then, members have been meeting via telecon to establish areas of focus from across a broad range of possibilities. Overall, we seek to serve the ESIP Community in its efforts to apply machine learning. We continue to seek community input on achieving that, and hope to hear more as a result of this session.

This session will start with an introduction to the cluster, a brief discussion regarding options for areas of focus, and also the results to date of a community survey. People interested in the topic will be invited to take the survey and/or join the cluster.

Following this, in the spirit of learning what ESIP members are doing in this arena, we hear from three cluster members involved in machine learning. The first presentation describes the application of machine learning techniques to a specific science problem, and the following two presentations describe tools for machine learning. The following are abbreviated presentation abstracts:

Estimating Daily Surface Air Temperature Using Satellite Land Surface Temperature and Top-of-Atmosphere Radiation Products over the Tibetan Plateau
Yuhan Rao, Department of Geographical Sciences, University of Maryland
In this study, we developed a new method based on the rule-based Cubist regression model to estimate daily surface air temperature under all-sky conditions with temporally gap-filled LST, incident solar radiation at the surface, top-of-atmosphere (TOA) albedo and outgoing longwave radiation products as the input. Two different model strategies are compared in this study: 1) building a unified all-sky model or 2) building individual models for clear-sky and cloudy-sky separately.

Marvin: An Automated Machine Learning as a Service for Primitive Annotation and Execution
Brian Wilson, Sujen Shah, Chris Mattmann, Ryan McGranaghan, Giuseppe Totaro, Mark Hoffmann, Vishal Lall, Alice Yepremyan, Srinidhi Nandakumar, Wayne Burke, Paul Ramirez, Jet Propulsion Laboratory
A NASA team at JPL has been participating in the DARPA Data Driven Discovery of Models (D3M) program, whose goal is to create a system that can automate the composition of ML Pipelines to simultaneously solve a 1000 ML problems, ranging from classification or regression on a simple “feature matrix” to image recognition, audio segmentation, EKG/EEG classification, video processing, graph clustering, time-series forecasting, etc.

Raster Vision: Deep Learning for Aerial and Satellite Imagery
Rob Emanuele, Azavea
Raster Vision is an open source framework for doing deep learning on satellite, aerial and other large raster data. It allows engineers performing deep learning on raster data a way to quickly and repeatably prepare training data, train models, make predictions, and evaluate models. It works with local GPU machines as well as AWS services.

Session Takeaways (post-meeting):
1) Bias can easily be built into these ML models. Therefore, you need to have domain experts or somebody who knows something about the ingoing data to deal with bias or problems with input data.
2) There are a lot of questions about what the ML cluster should focus on and how much depth vs breadth the cluster should address in such a broad field
3) Its easy to get lost in the proverbial ML weeds so either you need expertise in the ML side of things (data/computer science) and the earth science applications or bring in a few people with different expertise.

Session Abstract:
FAIR principles need to be applied to all outputs of scientific research, including physical materials such as natural and synthetic specimens that are collected as part of the scientific process. Making physical samples and collection FAIR requires virtual representations of the physical objects that can be located on the web via a persistent identifier and that are documented with sufficient metadata that are persistently accessible in trustworthy metadata catalogs to be findable and reusable, and that follow standard protocols for accessibility and interoperability. While best practices for sample identification, documentation, and citation are emerging, there are many open questions and challenges that need to be solved, including vocabularies and ontologies for describing and classifying samples; effective and consistent ways to link samples with the literature and digital data; as well as policies and infrastructure for trustworthy curation of the sample metadata and of the actual physical objects. In this session we will look at ongoing and new initiatives related to physical samples as part of digital research data infrastructure and explore how ESIP can facilitate collaborative solutions to some of the most urgent problems.

Session Takeaways (post-meeting):
1) Physical sample behind the data needs to be findable. Need licenses on physical samples.
2) Need to work on the guidelines of what metadata is necessary
3) Priority topics: Metadata, Policies for citations, Trustworthiness of metadata catalogs and sample repositories, Open access policies for physical objects & digital representations, Easily usable system for uploading metadata

Session Abstract:
Across industries and government, there are many groups working with virtual, augmented, and mixed reality technologies inside and outside of the realm of Earth science. In this dynamic panel, we will hear from scientists and researchers working with VR/AR/MR from various scientific disciplines about their approach to data visualization techniques. Interdisciplinary work and complex data sets will provide interesting perspective to the ongoing exploration of how immersive technologies can be used for Earth science and beyond.

Session Takeaways (post-meeting):
1) Presently, we can see the opportunities for AR/VR. The term “magic moments” is expressed to describe the peaks in using technology like this, where we are fully immersed in the activity we are doing and forget the layer of technology connecting us to the activity.
2) There are many key limitations with this technology, worthy of consideration. Data ingestion pipelines, standardization of data for VR visualization pipelines, hardware limitations (Field of View, physical strain on users), Costs, evaluating how this visualization technique is beneficial
3) Simplicity is a wholly necessary element of UI with VR/AR because input modalities with this technology are very simple as well, like gestures or voice commands. It’s a rich area of research at present and we need to find the basic elements which can be standardized.

Session Takeaways (post meeting):
1) ESIP Lab provides the support and framework for a variety of projects that span the realm of earth science data sciences.
2) Products/projects developed through the ESIP lab help to improve data management, organization, quality control, and/or accessibility for a number of different audiences/communities.
3) The next steps for these projects include securing further funding via proposal opportunities, advancing participants’ PhD research, and increase user count.

Brian Wee: Enabling the encoding and visualization of provenance metadata for better discovery and understanding of climate resilience strategies for agriculture-related decision-making
Keith Maull: ESIPhub: Developing and promoting an ESIP community resource for sharing and running scientific workflows via JupyterHub
Ziheng Sun: Geoweaver: a web-based prototype system for managing compound geospatial workflows of large-scale distributed deep networks
Mike Daniels: SensorDat: Real-time sensor testbed for improved provenance and data quality
Yunsoo Choi: A deep-learning driven improved ensemble approach for hurricane forecasting
Jessica Fayne: Building an Operational Network to Validate Novel Inland Water Swath Altimetry

Session Abstract:
Data rescue requires resources that are often scarce, and can also hinge of the timeliness of rescue efforts. Efforts by ESIP, RDA, EarthCube, and the IMLS-funded Data Rescue Toolkit have all focused on the development of elements to help focus and motivate data rescue efforts and improve the efficiency of efforts in the future. A clear need from within the community is to help synthesize findings from each of these groups, and to obtain input from the broader geosciences community, those actively involved in ad hoc data rescue efforts, and other issues around data rescue that may be particular to that community.

This session will begin with a brief overview of the work accomplished by the various groups mentioned above, and provide an introduction to the goals of the newly-approved CODATA Task Group, Improving Data Access and Reusability (IDAR-TG). The IDAR-TG highlights the importance and timeliness of not only data rescue, but also the need to collect and/or develop tools and frameworks to assist with assessing what data is in most need of rescuing.

The majority of the session will be a working session where we will discuss what characteristics of the data and its environment (repository, funding stream, etc.) increase data’s risk level, what considerations need to be made when deciding whether the data needs to be rescued, what resources are already available for this task, and how the Data Nomination Tool and other activities in the Data Rescue Toolkit project might be able to help.
Agenda: Set-up of problem (20-30 minutes):

• Intro (Hills)
• RDA Data Rescue (Diggs
• EarthCube/GeoDeepDive (Goring)
• ESIP Data Stewardship (Mayernick) slides
• Data Nomination Toolkit (Boehm)
• CODATA IDAR TG (Diggs)

Working on defining the path to a solution (45-60 minutes) Topics might include:

• How to define risk
• categories of risk
• ways to quantify (or weight) various risk factors
• other geo-disciplines that we might need to reach out to
• other scientific disciplines that have addressed data at risk evaluation
• ??

Summary/wrap up (5-10 minutes)
Please help with collaborative note-taking HERE!

Session Takeaways (post-meeting):
1) Sharing/telling data rescue stories (the process; what was enabled by the rescue; financial benefit) is necessary to get the essential support for data rescue efforts.
2) Coordination with aligned groups is key to be sure that effort is not duplicated and that efforts are aligned.
3) Harnessing Code for America style-infrastructure to pair motivated people/students with relevant skills to the opportunities to participate in data rescue efforts could be a good model to identify data heroes / data minions within the data nomination tool.

A review of results from data collected among ESIP members about their perspective and vision of the changing state of making Earth Science data matter. Please join us if you'd like to hear about the results, contribute feedback, and/or make an impact on our strategic direction as an Earth Science data community.  

Session Abstract:
The ISO metadata standards include the capability to identify metadata components and resources in several ways and this capability is being used in many different ways by many data providers. We will have presentations by Uwe Schindler from PANGAEA and Adam Shepard from BCO-DMO on how they are using identifiers and links in their metadata and a summary of uses in other collections by Ted Habermann.

Session Takeaways:
1) There exist many different ways to express linked metadata and no one way is necessarily best for all practices.
2) Linking metadata/data is the natural evolution of database technologies, first implemented for the world wide web, but is now expanded by semantic technologies.
3) Link early and link often with identifiers that can be resolved.

Session Abstract:
The use of soils data within the Earth sciences often requires the researcher to harmonize and reconcile data from multiple sources. This process can take up to 80% of a researcher’s time (OGC, 2016). Multiple groups are discussing standards for data collection, archival, and metadata to increase the ease-of-use and portability of data. These groups include (1) the International Union of Soil Sciences Working Group on Soil Information Standards, which has conducted an Open Geospatial Consortium (OGC) Soil Data Interoperability Experiment (OGC, 2016); (2) the UN Food and Agriculture Organization (FAO) Global Soil Partnership, which has developed the “Five Pillars of Action” -- three of which relate directly to the research data lifecycle (UNFAO, 2018); (3) the USDA National Institute of Food and Agriculture (NIFA), which has co-convened a Data Science in Agriculture Summit (NIFA, 2017); and (4) the USDA Natural Resources Conservation Service (NRCS), which recently released a technical note on common protocols to improve data quality and facilitate data sharing (NRCS, 2018).

In this session, we will hear from speakers in the Earth sciences community who are utilizing soils data in their research. We will focus on integration challenges they've had using these data through the research data lifecycle, from data creation to archival in a repository for re-use. The session will also provide perspectives on how these problems were (or may be) addressed and the effect these problems -- and solutions -- have on data analysis and the re-use of data. We will have presentations from researchers conducting both primary and secondary research, with the goal of highlighting solutions and next steps for the use and curation of soils data in the Earth sciences. By the end of the session, we aim to have identified the major challenges in using and re-using soils data, particularly as they relate to the use of soils standards.

Speakers
1) Ben Bond-Lamberty, “The global soil respiration database: goals, impacts, lessons.”
2) David Klinges, “A network for coastal carbon research: soil data archival as a community resource and to reduce uncertainties in modeling and mapping.”
3) Greg Wilson, “Agricultural Collaborative Research Outcomes System (AgCROS)”

References
NIFA, 2017, Data Science in Agriculture Summit (https://nifa.usda.gov/data-science-agriculture-summit)
NRCS, 2018, USDA Releases Standard Indicators and Laboratory Procedures to Assess Soil Health for Public Comment (https://www.nrcs.usda.gov/wps/portal/nrcs/detail/national/newsroom/releases/?cid=NRCSEPRD1420430)
OGC, 2016, OGC Soil Data Interoperability Experiment (https://portal.opengeospatial.org/files/?artifact_id=69891)
UNFAO, 2018, The 5 pillars of action
http://www.fao.org/global-soil-partnership/pillars-action/en/

Session Takeaways (post-meeting):
1) The more difficult the metadata guidelines, the fewer submissions you will have to your database.
2) To get buy in from the science community, they have to care and feel like it supports them.
3) Targeting a small community and having dedicated data management personnel...

How can we make it possible to coordinate management, replication, and governance of data on decentralized infrastructure? What efficiencies can we gain? What practices and lessons from centralized data governance should we take care to protect or learn from? ...and what does 'decentralized infrastructure' even mean!? Let's talk, together.

This is a working session -- there will be post-its for you to play with! Perfect for the last session of the day. It will be light on slides, heavy on small group conversation. The outputs of our time together will be digitized and share back to the ESIP community.

Session Abstract:
What information is hidden in your data files? Can scientific communities come together to create, review, and publish tools that automatically extract information?
We will talk about some examples

• Delineating trees from LIDAR images,
• Determining canopy cover of crops from drones,
• Extracting information from weather sensors.

We will discuss a global system for deploying and running these tools as well as proposing a process for community collaboration. 

Plenty of time will be set aside to allow open discussion about how this idea can be advanced.

Session Takeaways (post-meeting):
​​​​

Session Abstract:
The Disasters Lifecycle cluster in collaboration with the All Hazards Consortium developed initial ORLs with the electric utilities, and is now “operationalizing” ORLs for data-driven decision-making support to improve situational awareness. The criteria for the ORLs and a flowchart assessment tool, exercised by the AHC team at Duke Energy for response to the 2018 Hurricane Florence, proved very useful. The initial ORL criteria were defined for the electric sector’s use to transport work crews and restore power after destructive storms. However the usability criteria is expected to change for different disaster scenarios. Some factors may go across many use cases for many applications, such as security criteria that is key to establishing trust. However other factors are driven by the use case to address latency or resolution criteria. Work continues on refining strategies and criteria for assessing candidate datasets for specific operational use cases, and maturing the ORL concept as a Framework for different applications.

During this session we plan to address the issue of terminology to seek a common vocabulary relevant to various disasters application, leveraging what we’ve learned from the electric utility sector. We also would like to examine the ORL assessment tool and how the current ORL criteria are applied and look for gaps and lessons learned. The goal is to refine a framework strategy enabling the ORL concept to be applied to other disasters applications.

Speakers & Presentations

• Karen Moe & Dave Jones, Disasters Lifecycle cluster Co-Chairs: Introduction and highlights from 2018 
• Tom Moran, Executive Director All Hazards Consortium: Emergency management perspectives on the ORL approach - https://doi.org/10.6084/m9.figshare.7599911
• Jason Fleming, Seahorse Coastal Consulting: Real-time ADCIRC storm surge model, mapper, and ORLs - https://doi.org/10.6084/m9.figshare.7599908
• Sophie Hou, National Center for Atmospheric Research NCAR / UCAR: The importance of end user terminology from a usability standpoint - https://doi.org/10.6084/m9.figshare.7599914
• Maggi Glasscoe, NASA Jet Propulsion Laboratory / Cal Tech: Earthquake Clearinghouse sensor data request form - https://doi.org/10.6084/m9.figshare.7599917
• Ge Peng, NOAA Affiliate / NCSU: Data Maturity Matrix concept and the User Service - https://doi.org/10.6084/m9.figshare.7599920

Session Takeaways (post-meeting):
1) Speed of impact of the data
2) Need to have training and socialize the product
3) Need to look at dynamic data vs. forecasts vs. real time

Session Abstract:
The Metadata Game brings teams together to collaborate to create complete metadata repositories.

Session Takeaways (post-meeting):
1) Making complete repositories is difficult.
2) Meaningful metrics that work for communities are helpful

Session Abstract:
Our ESIP Lab funded project asked the question: how effectively can we capture the provenance between data products and decisions? We still do not have all the answers, but we hope you’ll find this session thought provoking, regardless of whether you’re a provenance expert asking the same question, or a member of a climate resilience project looking for ways to find and adapt existing data-driven solutions, or a policy analyst estimating the socio-economic value of data by tracing decisions back to data.

Among the topics covered will be: ontologies for capturing decision making, machine parsing of human readable documents to extract concepts, capturing decisions using process frameworks like the US Climate Resilience Toolkit planning guidelines and the National Park Service’s climate change scenario planning guidelines, text mining legally mandated public records for resilience infrastructure, and others. Audience feedback to our findings will be used to guide follow-on activities.

Session Takeaways (post-meeting):

View live-stream here: ESIP 2019 Winter Meeting - Day 3 Plenaries

View live-stream here: ESIP 2019 Winter Meeting - Day 3 Plenaries

Bio: Since joining the National Academies in 1990, Arrison has directed a range of studies and other projects on international science and technology relations, innovation, information technology, and strengthening the U.S. research enterprise. Recent projects include the studies Open Science by Design: Realizing a Vision of 21st Century Research, and Fostering Integrity in Research. He was also staff director for the InterAcademy Partnership reports Responsible Conduct in the Global Research Enterprise and Doing Global Science. He served as executive director of the InterAcademy Partnership for Research from 2013 to 2017. Arrison earned a master of public policy and a master of arts in Asian studies from the University of Michigan.

Mark A. Parsons, Chelle Gentemann, and National Academies of Sciences, Engineering, and Medicine Committee

MARK A. PARSONS is a senior research scientist at Rensselaer Polytechnic Institute (RPI). He is also director of data science operations for the Tetherless World Constellation at RPI. Previously he was Secretary General of the Research Data Alliance and an associate director of the Rensselaer Institute for Data Exploration and Applications. Prior to that, he was lead project manager at the National Snow and Ice Data Center at the University of Colorado, Boulder. He has been involved in data management for more than 20 years, during which he defined and implemented comprehensive data management processes for many projects and organizations. He is active in multiple international informatics efforts and led the data management effort for the International Polar Year (IPY). Mr. Parsons is a member of the Foundation for Earth Science Information Partners Board of Directors and a member of the Coordinating Committee for the Transparency and Openness Promotion (TOP) Guidelines. He received the American Geophysical Union/Earth Science Information Partners Charles S. Falkenberg Award. He earned his M.A. in geography at University of Colorado Boulder. Mr. Parsons has served on the Committee on the Development of a Strategic Vision and Implementation Plan for the U.S. Antarctic Program, as an ex officio member of the Board on Research Data and Information, and a member of the Committee on Archiving and Accessing Environmental and Geospatial Data at NOAA.

View live-stream here: ESIP 2019 Winter Meeting - Day 3 Plenaries

View live-stream here: ESIP 2019 Winter Meeting - Day 3 Plenaries

Bio: Heather Joseph serves as the Executive Director of SPARC (the Scholarly Publishing and Academic Resources Coalition, and international coalition of libraries promoting the open sharing of scholarship. As SPARC’s Director since 2005, she has focused on supporting emerging publishing models, enabling  digital archives, and promoting open access policies, nationally and internationally. Prior to joining SPARC, Heather spent 15 years as a journal publisher, primarily with scholarly societies. She continues to be active in this community, serving on the Board of Directors for organizations ranging from the Public Library of Science to the Arcadia Fund. She is a frequent speaker and writer on scholarly communications in general, and on open access in particular.

Session Abstract:
Centralized or common repositories expose data to new users by bringing together metadata from various archives into a single repository. Centralized repositories also serve as a one stop for data discovery and help increase the likelihood that data will be reused for new research and applications.

Since these repositories expose new users to new data, it is essential that the information provided in these repositories be trustworthy and reliable. Ensuring trustworthiness is an ongoing challenge for centralized repositories.

This session will focus on approaches to ensuring quality and trustworthiness in centralized repositories and the challenges faced by centralized repositories in meeting those goals.

Session Takeaways (post-meeting):

Session Abstract:
Since 2016, selection criteria have guided decisions made about the range, type and depth of educational resources published in the ESIP-hosted Data Management Training Clearinghouse (DMTC). The selection criteria were chosen with input from the first three organizational collaborators: ESIP, DataONE, and the U.S. Geological Survey. Once accepted for publication, the educational resources have been described using a subset of the Schema.org endorsed metadata schema for educational resources called the Learning Resource Metadata Initiative (LRMI) that is maintained by the Dublin Core Metadata Initiative (http://dublincore.org/dcx/lrmi-terms/1.1/ ). Specific elements of the LRMI metadata scheme were then chosen as featured search facets and descriptive elements in the search results display of the DMTC.
In July of this year, the DMTC received an Institute of Museum & Library Services (IMLS) National Leadership Grant focused on the expansion of the disciplines in the Clearinghouse, and enhancing the discovery and evaluation of the resources in the DMTC. As a result, it is good time to solicit feedback from the science and social science researcher and trainer communities on the DMTC editorial policies described above. In this working session, participants serving as DMTC resource submitters, reviewers and editors will be invited to test current submission and editorial practice by choosing educational resources not yet included in the publishing queue, deciding whether to submit them for publication using the DMTC selection criteria, and describing them using the LRMI metadata elements present in the full DMTC submission form. A discussion will follow these individual or small group activities in which participants will be asked to provide feedback on what was challenging and/or positive about their experience. The purpose of the discussion will be to reflect upon how the editorial policies and practices affect the scope, range and potential for sustainable growth of the DMTC inventory. Information to provide context and background data for the activities will be distributed to participants ahead of time based on the Sched participant list.

Use the Selection Criteria to guide your evaluation of educational resources to add to the DMTC at: https://github.com/imls-dmt/resources-workflow/blob/master/Nov17_2018_SelectionCriteria.pdf .

Take the brief tutorial to aid you in creating and submitting the metadata for educational resources at: https://github.com/imls-dmt/resources-workflow/blob/master/Tutorial%20for%20Submitting%20Educational%20Resource%20Metadata.pdf .

Use Working Session spreadsheet to evalute, submit and review educational resources in the DMTC: https://docs.google.com/spreadsheets/d/1BtZOsftJWGGZsYnbImz7TnnkX5dlbDoVjCc3010hMgA/edit?usp=sharing https://drive.google.com/file/d/1VfUCRaGMuCevHgwFwjIgjT1j1TcS5qR9/view?usp=sharing

Session Takeaways (post-meeting):
1) Need to make sure that we’re considering different learning styles in our selection criteria, so that we accommodate both visual & aural learners as well as more formal readers, for example. Finding experts to advise us in this area would be very helpful.
2) Would be helpful to take advantage of what others in ES education / web portals are doing with respect to evaluating & noting when educational resources are made accessible to those who are visually or otherwise challenged.
3) “Uniqueness” of coverage of a topic can be an interesting or memorable way to present a concept that is often discussed elsewhere in addition to discussing a new or unique topics per se.

Session Abstract:
Deep learning (DL) is the hottest method to realize artificial intelligence in many applied science domains. Our ESIPLab project Geoweaver has started to use Deep Learning method in producing crop maps with higher resolution and accuracy than conventional crop map production methods.
The success of DL relies on massive training datasets and powerful compute nodes like Graphics Processing Units (GPU). A good neural network requires careful engineering and considerable domain expertise in network training. It is never easy to fit DL on any Earth dataset. This session will carry out discussion on the research areas, technical details, data sources, and performances of DL in agriculture. We will work on harmonizing and generating a common strategy to connect and prepare Earth datasets for the training/testing of customized deep neural networks to help advance agricultural researches into next level: intelligent agriculture.

Session Takeaways (post meeting):
1) The workflow of deep learning: gather data > choose network type > choose DL library > find powerful hardware > data preprocessing > training > predicting > validation. This workflow can be used in all three aspects of agriculture: monitoring, predicting, and decision making.
2) The quality of crop information in the training datasets is the key to a successful model. The deep learning models must be fed with very accurate information so they can learn those pattern features. Less accurate training datasets will lead to a waste of training time.
3) The trained models have restrictions in time and location. The accuracy is largely related to the quality of training dataset, the chosen deep neural network type, the training process control, and the expertise of the practitioner. Normally, this is a collaboration work which need a stable group to work on it. ESIP machine learning cluster and ESIP Github repos are great platforms and tools for these efforts happen.

Session Abstract:
The ESIP Semantic Technologies committee has been working to harmonize ontologies in the Earth sciences with ontologies in other fields (environmental science, biology, chemistry, etc.). We will be working on harmonizing SWEET with ENVO, with particular concentration on cryospheric terminology.

Session Takeaways (post-meeting):
1) Creating harmony between ontologies requires, expert knowledge, iterative testing, and considerable time.
2) Few concepts are as simple as you think, the relations between objects build the complexity. This complexity gives value to data they represents. Ice-cores and ice-fields as an example. These are made of ice, have similar base properties but are fundamental differ in what they can demonstrate in science. This is an example the value of a thorough ontology.
3) A monthly meeting will be formed starting 1/31, a cluster may also be formed.

Session Abstract:
The majority of scientific programming workflows are developed in isolation by graduate students and postdoctoral researchers. While packages and libraries in R and Python help support the advancement of scientific discovery, researchers are often challenged with combining and analysing data in new ways. Regardless, code use and re-use in the Earth Sciences is often complicated by the fact that few well-developed workflows exist as templates. Most code examples in R packages for example, use well-worn datasets that are not well suited to extrapolation for Earth Science applications. For this reason, the discovery and analysis of existing code resources, such as those undertaken by the FUNding Friday grant, become critical to providing resources to scientific programmers in the Earth Sciences.
This Session will introduce early-career researchers to the principle workflows for sharing code publicly, including discussion of some of the pros and cons of sharing code before it is “good enough”. The session will then provide an overview of work that has been undertaken to analyse a large number of Jupyter notebooks on GitHub, and then provide session members with an opportunity to help build the web of examples for coding resources, discussing what makes code useful as a “cookbook recipe” for Earth Sciences, what particular libraries or data resources are of interest, and how further automation might be undertaken.

Session Notes:
https://docs.google.com/document/d/1S5p4v77B3kCdWSKbxKw3gkPFTotMeS9_pm5ic9rYhak/edit?usp=sharing

Session Takeaways (post-meeting):
1) Cultural knowledge around data use and storage can impact usage and keep data and use in ‘silos.’
2) There are a lot of people now that are putting notebooks on GitHub that are associated with a specific publication. The people that are doing this well are associating the DOI with the original publication and the people doing really well are setting this up so that the repository also has its own DOI.
3) Earth science data cookbook has an easy form to fill out information for earth science datasets and resources. This is intended to make these resources more accessible and clearly labeled. This thing is not live yet, but users can input keywords etc. available at https://bitly.com/esip2019-cookbook