Integrated, Coordinated, Open, and Networked (ICON) Science to Advance the Geosciences: Introduction and Synthesis of a Special Collection of Commentary Articles

The sciences struggle to integrate across disciplines, coordinate across data generation and modeling activities, produce connected open data, and build strong networks to engage stakeholders within and beyond the scientific community. The American Geophysical Union (AGU) is divided into 25 sections intended to encompass the breadth of the geosciences. Here, we introduce a special collection of commentary articles spanning 19 AGU sections on challenges and opportunities associated with the use of ICON science principles. These principles focus on research intentionally designed to be Integrated, Coordinated, Open, and Networked (ICON) with the goal of maximizing mutual benefit (among stakeholders) and cross‐system transferability of science outcomes. This article (a) summarizes the ICON principles; (b) discusses the crowdsourced approach to creating the collection; (c) explores insights from across the articles; and (d) proposes steps forward. There were common themes among the commentary articles, including broad agreement that the benefits of using ICON principles outweigh the costs, but that using ICON principles has important risks that need to be understood and mitigated. It was also clear that the ICON principles are not monolithic or static, but should instead be considered a heuristic tool that can and should be modified to meet changing needs. As a whole, the collection is intended as a resource for scientists pursuing ICON science and represents an important inflection point in which the geosciences community has come together to offer insights into ICON principles as a unified approach for improving how science is done across the geosciences and beyond.


INTRODUCTION
ICON-enabled approach to creating the crowdsourced collection; (c) summarizes insights from across the articles and the authors' experiences; and (d) explores lessons learned and next steps for ICON science.

What Is ICON?
ICON represents four principles (defined below) that together form a framework to guide the intentional design of any research project or scientific endeavor that is motivated by the pursuit of (a) mutual benefit and (b) transferable knowledge. ICON science is an approach to designing and carrying out research activities that has existed in many forms throughout scientific disciplines but coalesced into a framework in a 2019 U.S. Department of Energy (DOE) Biological and Environmental Research (BER) workshop report (U.S. DOE, 2019). Goldman et al. (2021) advertised involvement in the ICON Collection and provided definitions for each ICON principle. Here, based on the commentary articles, we have slightly modified the definitions in an attempt to reflect geoscience-wide perspectives. ICON is intended as a tool or heuristic to help researchers intentionally bring these principles into their projects by design: 1. Integrates across physical, chemical, biological, and/or social attributes and across spatial and/or temporal scales; 2. Coordinates use of consistent protocols and methods across systems to enable transferability across systems and researchers; 3. Openly exchanges ideas, data, software, and models throughout the research lifecycle that are findable, accessible, interoperable, and reusable (FAIR) such that all researchers are enabled to contribute and leverage resources; and 4. Networks efforts, whereby research is designed and/or implemented across the research lifecycle with a broad range of stakeholders to ensure mutual benefit.
These definitions are not static. The ICON Collection was approached with an awareness that the different AGU sections would have a spectrum of perspectives on what each piece of ICON meant within their discipline. Each assembly of writing teams elaborated upon definitions and expanded them as needed. Each ICON principle is described in more detail in the following paragraphs, including examples from articles within the Collection, recognizing that these definitions may differ from others. Best practices associated with ICON principles will differ across research disciplines that vary in technical details and across research settings that vary in terms of culture, resource access, and stakeholder needs. For example, to achieve mutually beneficial outcomes via a "Networked" research effort, different considerations/approaches may be required depending on variation in social, economic, and cultural details across research sites. It is important to emphasize that ICON science is about the intentional use of all four principles, not any one of them. For example, ICON science includes "Open" science based on FAIR principles, but also complements this approach with three additional principles that go beyond "Open" science via intentional integration of disciplines, coordination of methods, and development of mutually beneficial networks.

Integrated
There was agreement across all of the articles on the importance of integration to scientific impact and advancement. Some of the AGU sections even have integration across disciplines built into their names (e.g., Biogeosciences). However, the complexity of integration can make it challenging to achieve. In the ICON Collection's Natural Hazards article, Sharma et al. (2022) describe that addressing the need to assess multihazard multisector risk requires the "integrated assessment of complex interactions between hazard probabilities, exposure, and the vulnerability of the affected human or ecological system." Because multihazard risks are dependent on many factors such as environment, demographics, and socioeconomic conditions, the integrated understanding of these risk drivers is essential to a comprehensive view of natural hazard systems (Sharma et al., 2022).

Coordinated
A common driver behind geoscience research questions is to discover explanations and causality to phenomena regardless of location and time. To accomplish this, data and findings must be comparable across space and time to allow hypotheses to be investigated across diverse settings and scales. The "Coordinated" principle addresses the need to share protocols and methods that allow for improved quality and utility of the data generated resulting from consistency in its collection. In the ICON Collection's Cryosphere Sciences article, Brügger et al. (2021) highlight that different ice core laboratories may establish chronologies or proxies in ice cores using different methods, leading to challenges comparing within and across ice core records. The importance of the "Coordination" principle extends beyond physical sample collection. In the Earth and Space Science Informatics article, Hills et al. (2022) describe the importance of coordinated efforts "to implement standards for effective interdisciplinary data discovery and exchange…", yet point out that there are limitations in data reuse and discovery due to the lack of consistent and transparent protocols, for example, in data and code production, and processing methods across interdisciplinary teams.

Open
The "Open" principle of ICON refers most closely to the "Open Science by Design" framework laid out by the National Academies of Science, Engineering, and Math and elaborated upon in the "Open Watershed Science by Design" report from the U.S. Department of Energy. Open access in data repositories and research publications is one component, but the "Open" principle encompasses achieving openness in the whole lifecycle of research: provocation, ideation, knowledge generation, validation, dissemination, and preservation (National Academies of Sciences, Engineering, and Medicine, 2018; U.S. DOE, 2019). The "Open" principle of ICON is also intentionally defined to include the FAIR (findable, accessible, interoperable, reusable) data principles (Wilkinson et al., 2016). ICON is often used interchangeably with ICON-FAIR to make this more explicit, because as a general concept openness does not require being FAIR and vice versa, as highlighted in the ICON Collection's Earth and Space Science Informatics article (Hills et al., 2022). Some barriers to achieving the "Open" principle are consistent across fields and some are discipline-specific. In the Collection's Paleoclimatology and Paleoceanography article (Belem et al., 2022), describe one of the open science challenges as accessing "dark data," data collected before online and digitized data collection tools. Another challenge described by Belem and colleagues is in knowing where to look for data that a researcher needs because of the lack of a centralized and organized catalog of the databases and their contents. In the Biogeosciences article, Dwivedi et al. (2021) also describe that openness measured in publications does not translate to openness for the average citizen anywhere in the world. They call for a need to incentivize the dissemination of findings beyond the professional scientific community (Dwivedi et al., 2021).

Networked
Most science ultimately is pursued as a benefit to society. "Networked" goes beyond the casual, conference-style networking that happens, before, during, and after the workday, and instead focuses on the benefits of mutualism in the sciences. Mutually beneficial research can take the form of working with collaborators in such a way that their needs or interests are met, in addition to an individual or study's original research needs or questions; However, mutualism can and often should go beyond the individual researchers involved so that the wider community, including stakeholders, land stewards, and beyond, are considered. A key point underpinning the "Networked" principle is that designing research to be mutually beneficial for people involved and/or impacted is inherently linked to diversity, equity, inclusion, and, in the geosciences, often to environmental justice. One component of this is considering current and historical disenfranchisement that restricts certain groups from participating in the economic marketplace, scientific forums, governance, and other spaces that ultimately affect decision making. In part, this requests that researchers ask themselves questions before proceeding with a study design. In the Hydrology article (Acharya et al., 2021), provide a specific example binned into four categories: "(1) 'Who is doing the hydrology?' How will marginalized communities be involved? Will they have the same 'power and privileges' as non-marginalized communities? Who will own the scholarly outputs (e.g., data, grant proposals)?; (2) 'Who uses the water?' If marginalized communities are main water users, will they (or their communities) be able to sustain or use the hydrology knowledge research/work effectively (e.g., beyond the end of a project)?; (3) 'Who benefits from this activity?' Will marginalized communities get appropriate and meaningful attribution for their contribution? Will resources and infrastructure be available/sustained to marginalized communities after a project ends?; and (4) 'Why?' What is the purpose of this work and how will marginalized communities benefit and be supported?" The same article provides an example of work being done to strengthen the access and role of indigenous peoples in water research affecting their communities (Acharya et al., 2021). In the GeoHealth article, Barnard et al. (2021) highlight the importance of valuing the expertise of local leadership and communities in an effort to strengthen scientific arguments. In the Biogeosciences article, Dwivedi et al. (2021) suggest that a key challenge to networked efforts are the international cultural differences and resource variances that can cause the contributions of researchers in low-income and under-resourced countries to be undervalued or diminished. Ultimately, this disconnect can lead to a lack of understanding of historical scientific content, and subsequently misinterpretation of results and improper conclusions. This can lead to unintentional hard from research efforts. The "Networked" principle is intended to elevate equity by identifying where sciences can be built on the foundation of mutual benefit through strategic scientific resourcing. An important component of this is considering not just the benefit but also intentional reduction of harm. Many of the articles in the ICON Collection have identified that the "Networked" principle is anticipated to have the greatest benefit to the sustainability of the respective fields.

Integrated, Coordinated, Open, and Networked
As discussed above, ICON science is focused on using all four principles together, and many articles recognized the value of doing so. For example, the Education article discussed how that community has actively expanded ICON capacity through access to and use of shared resources and research findings, enhancing data sharing and publication, and developing leadership. This has led to greater capacity to address environmental and resource issues in just ways, and support equity and inclusion needed for a diverse geoscience workforce (Fortner et al., 2022). Likewise, the Biogeosciences commentary points out efforts like the U.S. National Science Foundation's Long Term Ecological Research program supports integrated, coordinated, and open science to address ecological challenges along with networking opportunities needed to understand needs across collaborators to enhance research development (Dwivedi et al., 2021).

Links to Other Heuristics
ICON is explicit in its definitions that FAIR principles are an integral part of its "Open" principle. Here we very briefly describe the philosophies of three other heuristics and their linkages to ICON.

CARE
The CARE principles (https://www.gida-global.org/care) are specifically founded in indigenous data governance. The letters stand for Collective benefit; Authority to control; Responsibility; and Ethics (Research Data Alliance International Indigenous Data Sovereignty Interest Group, 2019). In addition to the work on CARE individually, there is also work that intentionally links FAIR and CARE principles (Carroll et al., 2021). Much like ICON's emphasis on open throughout the entire research lifecycle, CARE takes a full lifecycle view of data governance that begins in the early phases of study planning and design. There are tremendous opportunities to explore how ICON and CARE can integrate together into studies, particularly for those deeply invested in the "Networked" principle of ICON. The examples described above in Section 1.1.4 from individual articles in the Collection have many points of connection with some of the critical components of CARE, and it is clear there is a path for more extensive application of CARE principles as ICON research grows.

TRUST
The TRUST principles were designed for data repositories with the foundational goal of guiding infrastructure that maintains FAIR data through time (Lin et al., 2020). The letters stand for Transparency; Responsibility; User focus; Sustainability; and Technology. The TRUST principles pertain to the "Open" principle in ICON, with an emphasis on the later phase of the research lifecycle when data are already generated. The TRUST principles have led to the identification of specific data repositories that meet the principles, which are an important consideration as researchers assess how and where they publish their data. We cannot draw strict boundaries to suggest that data must be published in data repositories that comply with TRUST principles in order to follow ICON principles, given the many factors that drive data repository choices, including funding agencies. However, the expansion of TRUST principles to more repositories seems poised to support both FAIR and ICON principles as it continues.

JEDI, IDEA, DEI
JEDI, IDEA, and DEI are more diffuse than the heuristics described above, and the words and accompanying acronyms vary. JEDI: Justice; Equity; Diversity; and Inclusion, or IDEA: Inclusion; Diversity; Equity; and Accountability; or DEI: Diversity; Equity; and Inclusion are only a few of the options. Similar to FAIR and the "Open" principle, the concepts in this heuristic space are integral to ICON as they are critical in understanding the mutual benefit that underpins the "Networked" principle. However, this extends beyond "Networked". At its core, ICON science is science that connects people. None of the four principles can be achieved without this, whether by gathering experts in different fields, understanding how others generate or use information, building open outputs that others can use, or operating for mutual benefit. As such, the pursuit of all ICON principles must be done through a lens that considers the people doing the research and affected by the research, and in order to do that successfully, JEDIA principles are foundational to every piece of ICON work.

Goal of the Special Collection
The ICON Collection was created to be a resource for researchers aiming to advance the geosciences through intentionally doing science following the ICON principles. Using ICON principles can be challenging due to the need for more a priori planning, logistical coordination, and stakeholder engagement, relative to many (but not all) traditional ways of doing science. How ICON principles are used also varies across research settings due to variation in numerous practical factors such as discipline-specific technical considerations, available funding and instrumentation, stakeholder needs, and science objectives. An additional challenge is that most scientists are not trained in how to intentionally develop and implement research projects that fully embody ICON principles. These challenges and lack of training are roadblocks to broad use of ICON principles. A primary goal of the collection is to bring together diverse perspectives on challenges, solutions, and opportunities associated with ICON science to reduce roadblocks and enable broader use of ICON principles across the geosciences and beyond.

Overview of Structure
The ICON Collection was meant to span all AGU sections using a crowdsourced collaborative writing approach. Each AGU section was allotted one commentary article comprising contributions from up to three independent writing teams. Most writing teams centered around a theme. The process of creating the ICON Collection is described below, and Tables 1 and 2 provide details about team formation and writing. Through this process we observed the emergence of common themes as well as discipline-specific perspectives across the contributed manuscripts, which are also discussed below.

Conceptualization
The approach used to create the ICON Collection was intentionally designed to follow ICON principles and provided valuable examples of opportunities and challenges that result from implementing ICON. Below we describe the approach used to create the Collection with the intention of helping to facilitate other crowdsourced paper collections in the future. A Town Hall led by members of the ICON Collection leadership team at the AGU 2019 Fall Meeting was a launch point for the Collection. The Town Hall, "Coordinated Open Science by Design to Transform the Geosciences," aimed to catalyze the idea of a special collection by bringing together geoscientists across fields and engaging in active discussions about examples, opportunities, and challenges of ICON science. We invited several panelists that spanned disciplines to provide a base of perspectives and discussions inherently integrated across disciplines. Because only AGU Fall Meeting attendees could participate, using the Fall Meeting also meant that some people were excluded from the opportunity. We accepted the limitations of the Town Hall, because the actual engagement in creating the Collection articles would be open to anyone that wanted to participate. This exemplifies an easy pitfall of trying to pursue open and equitable science throughout the research lifecycle; many scientific opportunities are not fully open, and it is critical to consider who is being excluded and why. As part of small group activities, Town Hall attendees discussed and wrote responses to the same list of questions, including whether they were interested in contributing to a special collection. This coordinated approach allowed us to compile an initial spreadsheet of ICON challenges and opportunities across disciplines that helped guide early development of the Collection structure. Soon after the Town Hall, we worked with AGU journal staff to identify a target journal and develop a special collection proposal.

Action Motivation Trade-offs
During sign up, writers have the option to write in suggestions for ideas/topics of interest to include in the articles Allows all sign ups to express ideas they are interested in focusing on Only the people who submit suggestions have their voices included in the subsequently formed five themes Leadership team reviews all submitted topics and groups them into five overarching themes Brings together people with shared interests Very narrow-focused topics are put into broader categories Each writer submits a ranking of the five themes Allows all sign ups to identify their priority teams and which themes they would not be comfortable or interested in contributing to based on their expertise Requires writers to rank all the themes, even if they only have experience in some of them Leadership team reviews all rankings and assigns writing teams with the aim of 1-3 evenly divided teams per article, depending on the total number of sign-ups. Writers are assigned to their first or second choice team. Articles with only one team are not assigned a theme Solves the logistical challenge of organizing over 180 individuals into writing teams Some writers were not placed in their first choice of teamAll five themes were not represented in each article When team assignments are distributed, teams are told they can modify and alter their themes as needed, and individuals can change teams upon request Provides all writers with flexibility and agency in their teams and themes Some teams change after initial assignment, which needs to be clearly communicated to all team members Writers who join the effort after teams have been assigned are incorporated into the teams following the same process or join teams directly without the leadership team's awareness Creates a mechanism for people to join the effort if they hear about it later than others Requires teams to integrate late joiners

Creation of Infrastructure
Members of the Collection leadership team held a workshop for the people who had attended the Town Hall to gather feedback on the proposed vision and structure of the Collection. We created a series of foundational documents informed by the workshop discussions that defined the ICON Collection approach, author guidelines, team norms, writing contribution guidelines, and roles and responsibilities. The guidance documents are available at https://data.ess-dive.lbl.gov/datasets/doi:10.15485/1840779 (Goldman et al., 2022). We expanded the Collection leadership team to five people to span a greater range of geoscience fields, and the new team iterated on the foundational documents to clarify the vision and approach and integrate ideas from the new leadership team members. The foundational documents played a critical role in creating coordination for the Collection. For the published commentary articles themselves, the foundational documents set instructions that allowed for flexibility while assuring the published content would follow a consistent framework to form a cohesive resource. For interpersonal dynamics of the writing teams, the foundational documents set guidelines and expectations with the intent of minimizing conflict, maximizing open communication, and creating an expectation of mutual respect.

Advertisement and Recruiting
The leadership team made the completed foundational documents public and began a multi-month open advertising campaign for people to sign up to get involved in the Collection. The advertising campaign included an Eos Vox (Goldman et al., 2021), a series of Twitter posts, discipline-specific mailing lists, announcements during meeting presentations, emails to colleagues, emails to previously not contacted organizational leadership ("coldemails"), direct engagement with AGU section leadership, and posting to the AGU Connect message boards and associated email newsletters. We particularly reached out to affinity groups like Geolatinas, 500 Women Scientists, Black in Geoscience, and ADVANCEGeo who helped distribute the information in their social media platforms and with their members. We encouraged people to spread the word to their colleagues, collaborators, followers, and beyond. During the advertising campaign, we worked with AGU to present the Collection at a monthly meeting for AGU Section Presidents to better understand how we could engage members across each of the 25 AGU sections. When signing up to get involved in the Collection, people could select interest in being a writer in the Collection, a "section champion," or both. The section champion was a facilitator role so that each article would have one or two people that communicated directly with the leadership team and understood the Collection structure and expectations. The champions were encouraged to reach out to their networks and colleagues during the advertising period. To equip the champions for their role and gather feedback, we held a workshop with the champions that was also recorded and posted to YouTube (https://tinyurl.com/SCworkshopI-CON). The workshop also provided a valuable opportunity to start building a sense of community among those involved in the Collection.
After implementing the strategies described above to recruit people for the Collection, the leadership team faced the challenge of highly variable numbers of sign-ups across the 25 AGU sections. We reached out to the AGU Section Presidents of the sections that had few or no sign-ups. This approach increased the number of participants in some but not all the sections. We then cold-emailed researchers and professors we found online who specialized in the disciplines with few sign-ups. We also cold-emailed geoscientists across disciplines at minority-serving institutions in the U.S. (i.e., Historically Black Colleges and Universities; Hispanic-Serving Institutions), at research institutions located in countries not well-represented by the sign-ups, and from databases such as "Water Researchers of Color" (Hampton & Byrnes, 2020). We cold-emailed over 140 scientists asking them to join the Collection or distribute the information to their colleagues or networks. After several months of the advertising campaign, we closed the registration form in July 2021 when most writing teams were actively writing or had completed their first drafts. However, we included a contact email for people who were still interested in getting involved, so involvement was never fully closed. Writing teams also brought in additional writers at times, and they were integrated into the Collection. Ultimately, the ICON Collection to date has 19 out of the 25 AGU sections represented. Of the six sections not included, three had at least one writer sign up to contribute but ultimately did not come to fruition after struggling to find co-writers or assessing the bandwidth they had available for investing in the effort. We encourage the inclusion of the six sections not represented, and if there are researchers in these disciplines that want to contribute an article, they can reach out to the Collection leadership team to get started. Although not all sections have their own article, we encourage researchers to read the articles across different sections to see the likely commonalities with their experiences.

Writing
The writing process operated within a framework set forth by the leadership team and supported by section champions, but the writing teams intentionally operated independently. The guidance documents provided to the writing teams are available at https://data.ess-dive.lbl.gov/datasets/doi:10.15485/1840779 (Goldman et al., 2022). The leadership team formed writing teams within articles based on themes submitted, collated, and then ranked by the writers (Table 1). Up to three writing teams each wrote an independent theme-based section, and these sections were collated into a single commentary article. Most writers did not know the other people in their assigned team. This approach allowed the writers to guide specific directions of the manuscripts, while still creating a sense of connection and consistency across the entire collection. This approach also intentionally created teams in which many people did not know each other or had not previously collaborated before but had shared interests, with the goal of sharing new perspectives, creating new connections, and maximizing innovation. Each writer came to the project with a firm understanding of their field of work and an interest in ICON principles. Whenever possible they brought in additional expertise to discuss the challenges, tools, and opportunities to advance their field. What was new and sometimes more difficult to connect were the ICON principles to these challenges and opportunities. The leadership team met upon request with section champions and writing teams and provided clarifications and links to guidance materials frequently. Most communication with the leadership team was done over Slack and email, including bi-weekly check-ins, and many writing teams held frequent virtual meetings for collaboration without leadership team members. The emphasis on communicating within writing teams rather than with the leadership team was intentional. We wanted the articles to reflect the perspectives and opinions of the writers and their experiences. Allowing for flexibility in interpretation of the article goals and themes allowed for the writers to more clearly emphasize what stood out specifically to them. In some cases this led to repetition by multiple writing teams within the single article, which was a valuable indicator of the importance of a topic to the discipline.
The maximum level of interaction between the leadership team and the writers came during two rounds of revisions to each draft ( Table 2). The feedback provided by the leadership team on the drafts was focused on the following: 1. General light editing (i.e., clarity, coherence, critical grammatical errors); 2. Verifying there were examples for points made (i.e., describing "how" not just "what"); 3. Clarifying ICON definitions and descriptions as needed (e.g., "networked" is more than conference interactions); 4. Verifying the overall article framing was around ICON (i.e., specific principles are called out and applied); and 5. Suggesting specific text/topics, improvements, ideas, and ways to think about components differently.
The leadership team also provided front-end language for the titles, abstracts, and introductions of the articles to help with cohesion and to provide the reader with context and connection to the rest of the ICON Collection. The leadership team provided the AGU journal requirements and left the submission duties to the writing team. The final submission was determined by the writing teams. Since the articles for most sections were made up of individual pieces written by independent teams, author order is often alphabetical and readers should not necessarily interpret author order as indicative of contribution.

Composition of the Writing Teams
An important component of transparency of the Collection is communicating the composition of the writing teams with the awareness of the biases that come from backgrounds, experiences, and perspectives that are absent or less represented. When recruiting the participants for the Collection, we asked them to fill out their demographics to be aware of the scientists' background behind the commentaries. Out of 201 participants who expressed interest to be part of the collection (sign ups), 135 end up being part of the group of final authors who wrote articles. From the final list of authors who participated in the ICON Collection, 25% did not register through the form that we used during the recruitment process (Section 2.4). Figure 1 displays six categories of demographics. For authors who selected more than one race/ethnicity, each race/ethnicity was counted separately. The most common gender identity and race/ethnicity across both sign ups and writers was male and "White or Caucasian." "South or Southeast Asian" was the second most common race/ethnicity. The two most common races/ethnicities that were selected at the same time were "White or Caucasian" and "Hispanic and/or Latinx". Of the 6% of "Hispanic and/or Latinx'' authors in Figure 1f, half also checked the box for "White or Caucasian". The most common age range of sign ups who expressed interest in the Collection and who participated in the process was 30-39 years. This correlates well with almost half of the authors identifying as early career scientists. To assess how the demographics of the ICON Collection participants compare to AGU members, we compared the final authors' demographics with the 2020 AGU's Diversity, Equity and Inclusion dashboard data collection (AGU, 2021; Figure 2). We compare demographics from the ICON Collection to AGU demographics as a point-of-reference. Authors without demographics data were categorized as "unknown." To have comparable categories in the race/ethnicity data to AGU, we re-grouped the ICON data from East Asian, Middle Eastern, and South or Southeast Asian into "Asian or Asian American". An important difference between the ICON Collection and AGU race/ethnicity is the AGU race/ethnicity is U.S. only, whereas the Collection data is from all the ICON participants. From the total authors who submitted commentaries to the collection and submitted demographics information, 55% are based outside the U.S. In the context of the total 181 authors in the collection, this translates to at least 20% of authors are based outside the U.S.

Group Dynamics
The ICON Collection leadership team requested feedback from participants to understand more about their experiences of writing in this crowdsourced approach. We heard from 76 of the 181 authors. It is important to recognize that this is a small portion of total authors, nonetheless their insights can still be very useful. Of those 76, most were interested in getting involved in another crowdsourced open science collaborative writing opportunity. Although they began this process without knowing the people in their writing teams, most felt that in their writing teams their ideas were heard and included and they were respected. One goal of this effort was creating a foundation for future collaborations, and most of the 76 thought there could be future collaborations created from this effort. One of the writing teams has already begun working on a new project.
The same 76 participants also provided input on what the writing teams and the leadership team could do to create a more inclusive culture and a more equitable culture. Several recurring themes emerged from the feedback: (a) Create opportunities for social engagement and communication early in the process to build trust and better understand people's working styles and needs; (b) Increase diversity, including international representation, and relatedly, improve scheduling for different time zones and create space for different languages; (c) Facilitate more direct communication between the leadership team and the authors; (d) Provide more clarity on authorship guidelines and verify agreement of all participants at the start of the process; (e) Increase advertisement of opportunities to get involved; (f) Provide examples of expected outcomes; (g) Make sure collaboration tools are accessible by all participants; (h) Increase use of virtual meetings rather than relying on written tools; and (i) Provide more time for participants to accomplish tasks. These themes specifically tie into "Coordinated", "Open", and "Networked" and illustrate not only important areas to improve upon in the future but also the value in critically assessing our approaches and tools through the ICON lens-not just at the beginning of the process, but repeatedly throughout the process.
It is important to recognize that even with intentionally designing the process of writing the Collection to align with ICON, we saw that at times people felt like they were not being fully heard depending on the dynamics of their team, or that differences in time zones were prohibitive for coordinating meetings with writing teams. As described above, we placed individual contributors in writing teams within their discipline based on a ranking system of possible themes of interest, and although the responsibility to make sure teams were coordinating well was given to each section champion for the section, retrospectively it may have been useful to establish teams in a way that was structured by time zones or more involved based on communication styles. For some articles, no writer volunteered to be section champion, so a leadership team member stepped into that role. This approach did not hold the same weight as having a champion from the discipline who could understand more nuances of the discipline-specific dynamics and was available to be more hands-on. For a collection of this size, it is not feasible for five leadership team members to structure the full list of authors into individual personalized groups, but it would have been helpful to have more section champions and have each of those champions be more involved in establishing the teams based on the dynamics they saw. This likely would have addressed some of the comments that mentioned individuals who were more outspoken or more senior within their career stages had a disproportionate voice within their groups. Groups that were, by chance, structured by earlier career stage individuals seemed to have had pleasant experiences with their opinions being heard and valued, and thus providing support with a more involved grouping dynamic may have helped mitigate some of these issues. It also may have been helpful to hold a virtual meeting space where the leadership team could oversee the introduction and dynamic of the different writing teams, as some people noted that they would have liked a more involved role from the leadership team to establish the teams.
Interestingly, even within a group of writers focused on ICON and using an ICON approach to the Collection, we had some difficulties regarding authorship order and authorship contributions. This suggests that even people who recognize the importance of what the ICON framework represents struggle with implementing it when the benefit structures in science have not yet adopted similar mindsets for collaborative work. This experience demonstrates that fervent effort is needed to shift the scientific culture toward a more open, equitable, and collaborative perspective of authorship while also changing common metrics of success. The success of such a cultural shift relies in part on institutions and funding agencies recognizing and emphasizing different metrics of success beyond first-author publications (Davies et al., 2021;Moher et al., 2018). A few such metrics can include (a) type of role in a publication and frequency of that role; (b) FAIRness of dataset publication; (c) preprint publication; (d) preregistration of studies; (e) publication of protocols; (f) number of or types of collaborations beyond a home institution; (g) stakeholder outreach; and more. The expanding use of the CRediT (https://casrai.org/credit/) system for describing authorship contributions could eventually allow for an automated system to pull out what roles an author filled in their publications, which would allow for less emphasis to be placed on author order and more on specific author contributions.
Finally, the bias toward a lack of underrepresented groups and marginalized communities within STEM fields is prevalent within the ICON Collection even after the leadership team's attempts to reach out to specific groups and organizations in an effort to increase the overall representation. We recognize that not all voices in the geosciences are represented in the Collection, and that greater efforts must be taken to capture these voices. It is possible that some scientists we reached out to from marginalized groups could not afford to take time to write in the Collection, and that further placing the onus on these communities to navigate a way to become involved seems like an inappropriate way of making their voices heard. In an effort to provide greater inclusivity within future collections, financial support or other tangible resources may help mitigate the disparity in the demographics. As it was put by one of the writers who provided feedback: "we still have a ways to go." It is our hope that the ICON Collection serves as a primer to help people understand what we need to move toward, and how it can be done to enable scientific pursuits to be more aligned with the foundational goals of ICON.

Defining ICON
Throughout the writing process and most clearly during the leadership review of the first drafts of the articles, it was clear that there was variation in how people understood some of the ICON principle definitions. Teams were provided with written definitions at the beginning of the process in the article advertising involvement in the ICON Collection (Goldman et al., 2021). They were also provided the link to an example of ICON in practice on the website for the Worldwide Hydrobiogeochemistry Observation Network for Dynamic River Systems (WHONDRS; https://www.pnnl.gov/projects/WHONDRS/icon-fair-framework). There were three recurring experiences across the writing teams: (a) Teams expanded definitions to better fit their experiences; (b) Teams wrote extensive content related to a specific ICON principle but did not realize that the content was related to the principle; and (c) Teams misunderstood or partially understood the definition of one or more ICON principle. Having teams expand definitions to better fit their experiences was an outcome we hoped would occur during the writing process, and the content and nuances in the articles is valuable in understanding how different disciplines engage with ICON. Teams writing content without realizing it applied to a principle or misunderstanding a principle occurred most frequently with the "Networked" principle. Many first drafts identified engaging with colleagues at conferences and workshops as the source of "Networked" in their discipline and separately wrote about the importance of mutual benefit and stakeholder engagement without linking it to an ICON principle. This highlights that an important component of expansion of the ICON framework is clear communication about the meaning and foundation behind each principle. When a concept is already embedded in someone's mind, it can be challenging to incorporate a broader or different definition. This was also a challenge with the "Open" principle, which required people shifting from the concepts of open data or open publishing to open and FAIR science throughout the research lifecycle. Iterating with the writing teams during the two rounds of leadership team-provided feedback was a valuable way for the leadership team to reflect and learn from how writers were interpreting the ICON principles and to provide guidance when appropriate.

Common Themes
We found common themes across people's experiences creating the articles and across the key points defined in the articles. Although all articles aimed for the same goal of exploring ICON science within their field, in practice, each discipline is at different stages of enacting science following ICON principles. For example, some sections focused on the difficulties of collecting and sharing data and how the cultural and historical hierarchies within the field make this difficult. Other sections highlighted struggling with an excess of publicly available data that was not coordinated and as such, unavailable for meta-analyses or cross-study interpretations. However, across all of the articles, even for fields actively implementing ICON principles, there was a recognition that there are opportunities for growth and improvement that will ultimately help the discipline as a whole.
Perhaps the most common theme across manuscripts was the two-fold perspective that the geosciences would benefit from more use of ICON principles, but that using these principles also presents risks. For example, several articles mentioned the risk of "parachute science" and "helicopter science" in which samples and/or data are extracted for the benefit of researchers without providing commensurate beneficial outcomes to those providing resources and/or impacted by research outcomes (Minasny et al., 2020;Stefanoudis et al., 2021). This occurs most often in the context of researchers from wealthier countries traveling to developing or lower income countries and collecting data and resources for the purpose of taking it back to their original institutions. This is also common in work with indigenous communities, and the CARE principles for indigenous data governance were designed for improved research approaches (Section 1.2; Research Data Alliance International Indigenous Data Sovereignty Interest Group, 2019). Collecting data and resources from lands and retreating to home institutions can result in detrimental effects to the community that helped provide the samples/data/resources and divorces the scientific products from the locations, cultures, and communities from which they are sourced, often resulting in a lack of critical insights into the systems and environments and subsequently incomplete and improperly analyzed data.
In a related theme, many manuscripts highlighted the need for greater equity in science and discussed ways in which this could be achieved. Across manuscripts, it is clear that the geosciences community feels strongly that the risks of ICON must be considered and minimized through careful planning and community engagement. The issues can be context dependent and there is a need to work with stakeholders to understand risks and generate/ use mechanisms that minimize these risks. This risk evaluation is part of the "Networked" component of ICON, which is focused on pursuing research in a way that is mutually beneficial for the primary research team and multiple stakeholders involved in and/or impacted by the work. The repeated focus across manuscripts on the value of mutually beneficial research indicates a need to more fully develop and formalize strategies to achieve the ICON vision for "Networked" science. This goes hand-in-hand with increasing equity in science by using ICON principles to increase opportunities for researchers across diverse settings in a way that is mutually beneficial for those engaged and impacted. Ultimately, although each of the sections identified challenges and risks within their fields, there was a general consensus that implementing ICON principles will lead to successful scientific advances.

Perceived Benefits Outweigh Costs of ICON Science
As with every approach to doing science, the use of ICON principles comes with both costs and benefits. The benefits should outweigh the costs for any approach that is used. Otherwise, there is no motivation to use a given approach. It is thus important to assess the costs and benefits of all four ICON principles. A formal accounting of all costs and benefits is, however, far beyond the scope of our current efforts. Instead of a formal analysis, each writing team was asked to place each ICON letter within a cost-benefit space. This space was defined by a cost axis and a benefit axis, both ranging from 0 to 10 ( Figure 3). The placement of the letters was inherently subjective and meant to represent each team's perception of ICON costs and benefits. Upon completion, we visually estimated the location of each letter along each axis to the nearest quarter point. This visual approach was deemed suitable, instead of a more precise method, given that the teams placed the letters by simply dragging and dropping them on the computer screen.
Our analyses of the perceived costs and benefits clearly show that writing teams felt the benefits of all four ICON principles outweigh the associated costs ( Figure 4) and that variation in perceived costs was higher than variation in perceived benefits (Figures 4a, 4b and 5). The cost distributions were all centered near ∼5-6, while the benefit distributions were centered ∼8-9. The median benefit was significantly higher than the median cost when pooling data across all four letters and across all teams (Two-tailed Wilcox test: W = 2273.5, p < 0.0001). Not surprisingly, the costs and benefits varied across teams in the same section/article, and the analyses summarized in Figures 4a and 4b do not directly account for this among-team variation.
To directly link perceived costs and benefits, we calculated the cost-benefit ratio for each ICON principle within each team. For all four ICON principles the cost-benefit ratio was significantly less than 1 (Figure 4c), again showing that perceived costs are lower than perceived benefits. This was evaluated with a one-sided Wilcox test for each ICON principle: for "Integrated", V = 21, p < 0.0001; for "Coordinated", V = 14, p < 0.0001; for "Open", V = 6, p < 0.0001; for "Networked", V = 55, p < 0.001. Collapsing all team scores across all eight variables (one cost and one benefit for all four ICON principles) via a principal component analysis (PCA) showed that teams varied primarily in terms of the perceived costs of ICON ( Figure 5). This is consistent with the cost distributions being broader than the benefit distributions (Figures 4a and 4b).
It is encouraging that across diverse geoscience disciplines there is a consistent perspective among the participants that the intentional use of ICON principles outweighs the associated costs. In addition, participants indicated that their perspective on the importance of ICON principles changed through the writing process for this special collection. Specifically, many participants indicated an increase in their perceived importance of intentionally using ICON principles. It is important to recognize, however, that perceived benefits may not all be currently available. That is, some perceived benefits may be thought of as potential benefits presumably via careful implementation that minimizes negative outcomes. We cannot quantify this at present, however, because the cost-benefit analysis did not attempt to parse current versus potential benefits. Future assessments may consider doing so. In addition, the higher level of variation in perceived costs (relative to the variation in perceived benefits) indicates a need for deeper understanding of the costs of ICON. We emphasize that in the analysis, the interpretation of costs was not constrained. Each team interpreted the meaning and scope of "costs" as they felt was appropriate. This could have led to variation among teams, though teams were also free to interpret "benefits" as they felt appropriate. In turn, we hypothesize that higher variation in perceived costs was due to "costs" spanning a more complex suite of considerations than "benefits." For example, participants noted potential risks of using ICON principles that go beyond direct financial and labor costs (Section 4.2). To help evaluate the landscapes of perceived costs and benefits, it would be useful to gather information on the identities and relative importance of specific costs and benefits. More generally, our observations collectively highlight the need to better understand and minimize the inclusive costs and risks of using an ICON approach. As discussed below, the ICON Science Cooperative has been launched as one tool to help address these needs.

Next Steps Identified Within and Across Disciplines
Each of the ICON Collection's individual articles provide next steps and actions that can move each discipline forward. In summation these recommendations and suggestions offer a pathway to continue learning about ICON principles to support advancing science across domains. The steps described could be divided into three themes: funding, infrastructure, and focused community engagement efforts.
Many sections' articles pointed out the need for not only government research funding, but also funding from private and non-governmental organizations (NGOs) that enforces and emphasize policies that support the ICON principles. Almost all the Collection's articles included a suggestion to engage citizen science and to equip it with funding. Other funding related needs were mentioned in the Cryosphere Science article, including support for new types of undergraduate research experiences that can accommodate those unable to travel but who can conduct remote data analysis (Brügger et al., 2021).
Under the infrastructure theme, suggestions included the need for better coordination among scientists to establish data standards, centralized and shareable data and equipment, and better understanding of leadership, opportunities, and frameworks within initiatives. The Collection's Space Physics and Aeronomy article described a unique aspect of infrastructure in which memorandums of understanding (MOU) and agreements to host exchange programs can provide benefits that align with ICON (Sur, 2021). These agreements could increase "Coordinated" and "Networked" efforts, instead of encouraging competition that can be detrimental to the advancement of the field and to the students and early career scientists. Along similar conceptual lines in which formal agreements can help advance the use of ICON, the Collection's Near-Surface Geophysics (NSG) article highlighted a recommendation from the National Academies of Science, Engineering, and Medicine to provide access to NSG instrumentation from a central NSG Facility (Salman et al., 2022). Such agreements align with the "Networked" aspect of ICON in which efforts are made to develop resources that enhance the equity of access to scientific resources. The Collection's Education article also discussed how that community approaches infrastructure. For example, they use web infrastructure to share teaching resources and literacy principles. They further align their "Networked" principles by pairing community visioning and co-creation (e.g., geoscience research frameworks) with network building activities that engage a range of communities (Fortner et al., 2022).
There was agreement across articles that engaging with local communities was an important mechanism aligned with ICON principles, particularly "Networked," that is needed to uphold the societal value for science. The ICON Collection's Hydrology and GeoHealth articles both note the importance of engaging the public interest in critical issues of local interest for each ICON principle are summarized as kernel density functions. On each panel the median value for each distribution is given in the legend. Benefits are significantly higher than costs, and the cost-benefit ratios are significantly lower than 1 (see text for statistics). like water quality (Acharya et al., 2021;Barnard et al., 2021). The Collection's Biogeosciences article encourages the adoption of "people-centric" approaches to build research capacity, understand cultural nuances, and promote research community engagement with open fair research practices (Dwivedi et al., 2021). Several articles point out parachute science, discussed in Section 4.2, and instead encourage developing a relationship with local stakeholders, land stewards, and others, valuing their expertise, embracing the opportunity to learn from local or indigenous knowledge, and providing value back to them. These ideas tie in again to the CARE principles described in Section 1.2. The Paleoclimatology and Paleoceanography article describes "true collaboration," as "co-develop[ing] mutually beneficial projects with the local community, aligning outcomes with both of their goals" (Belem et al., 2022).

Expanding the Use of ICON
Pursuing research that fully embodies and uses all ICON principles is challenging, and there is a need for structural/cultural change and additional resources that collectively help reduce these challenges. There is a need to support and reward the time/energy individuals spend building collaborative efforts that make use of ICON principles. For example, it takes time to engage with diverse stakeholders to genuinely understand their needs so that research efforts can be designed for mutual benefit. Similarly, it takes time to ensure methods and (meta)data structures are consistent enough with other efforts to enable (meta)data interoperability. It also takes time to think through how to tangibly integrate one's science with other disciplines. Furthermore, it requires taking on some risk-some perceived and some real-to be truly open throughout the research lifecycle. Research institutions and funding agencies could foster the use of ICON by recognizing the value of that kind of time/energy investment and the risks that researchers take on when they aim to facilitate those beyond themselves. The associated recognition would need to have tangibly positive effects on career advancement.  Figure 4b) and showed little variation among teams. These inferences are based on the cost-associated arrows being much longer than the benefit-associated arrows; arrow length is proportional to the loadings of those variables on each of the first two principal component (PC) axes. Each filled circle represents one writing team, with colors indicating the associated AGU section. Larger distances between any points indicates larger differences in their perceived costs and benefits of using ICON principles; teams within some sections cluster closely while others are divergent.
As a complement to structural and cultural change, there is a need to develop and share resources to maximize the value and minimize the effort of doing ICON science. There are numerous resources and efforts to draw upon and continue to develop. For example, AGU's Thriving Earth Exchange (https://thrivingearthexchange.org/) helps scientists work with local communities to address environmental challenges. This is an example of being intentionally "Networked" to design and implement efforts that achieve mutual benefit. The wisdom of those engaged in the Thriving Earth Exchange could be brought together with related efforts to further advance our collective understanding of how to best achieve mutual benefit. For example, the Education commentary discussed how the Science Education Resource Center (https://serc.carleton.edu/index.html) supports an open community of practice and resource sharing. Similarly, the ICON Science Cooperative (https://ICON-science.pnnl.gov) was recently launched to help bring resources together to facilitate robust use of ICON principles. While the Cooperative will leverage other efforts that touch components of ICON (e.g., The Center for Open Science), the Cooperative addresses the unique challenge of simultaneously using all ICON principles. The Cooperative and related efforts could be brought together to more formally share knowledge and potentially co-develop resources to solve pressing challenges.
As discussed above, one of the pressing challenges identified in manuscripts contributed to the ICON Collection is the need to understand how to implement the "Networked" component of ICON. This is potentially the most challenging component of ICON because it requires understanding and meeting the needs of multiple stakeholders. Associated needs and benefits are often subjective and may be in conflict across stakeholders. This has the potential to lead to difficult situations for researchers, who are often not trained in how to find common ground among or even assess multiple stakeholder needs. As such, there is particular value in developing guidance and other resources around the vision for and implementation of "Networked" science. There is, however, also a need to develop strategies for simultaneously using all four components of ICON in a way that maximizes benefits and minimizes risks. ICON science is ultimately about being more intentional in how we design and implement research efforts to enhance the transferability of our understanding and the mutual benefit of research outcomes. We can all find deeper connections to and value from science if there is more forethought about how to integrate disciplines to draw in multiple perspectives, to be consistent in our methods so others can reuse and connect with our work, to find value in openly sharing and receiving knowledge and data from those beyond our immediate collaborators, and to make genuine efforts to understand how even small changes in what we do can have large positive (and negative) effects on others. ICON science can enhance the value of scientific efforts by directly and indirectly connecting people, ideas, data, models, and knowledge across diverse settings. The ICON Collection is an example of this in action. Each person that contributed to this collection has their own perspective on ICON. Those individual perspectives are highly valuable, yet may go unheard without a critical mass of other voices. ICON principles themselves helped enable the collection to be a platform for those voices. By spanning AGU sections the collection itself strove to integrate perspectives across disciplines. Using a coordinated approach to crowdsource the manuscripts allowed for consistency in the focus and structure of the manuscripts. Being open throughout the process allowed all those with interest to join and share their perspective on ICON. Listening and responding to the needs of contributors throughout the process helped generate outcomes that are-we hope-beneficial to both the writers and the readers. ICON science pulls together existing ideas and ideals into a cohesive heuristic that can be applied to all science domains to broadly enhance outcomes. This will only happen if scientists and stakeholders sincerely and intentionally apply the full suite of ICON principles, while simultaneously looking for ways to improve this heuristic tool.

Data Availability Statement
The data and code (R scripts; version 3.6.1) used for plotting and statistics are available at https://data.ess-dive. lbl.gov/datasets/doi:10.15485/1840779 (Goldman et al., 2022). The foundational documents associated with the ICON Collection can be found at the same link. The data associated with demographics are not published to protect the anonymity of participants.