V3Geo was conceived to simplify access and sharing of virtual 3D models in geology and geoscience. Due to the large 3D datasets, publishing models at their original resolution, sharing results with collaborators or project sponsors or distributing a dataset to support a university class have been major barriers to acceptance of virtual methods. Consequently, the initial specification for V3Geo was devised to meet the following key criteria.

Support large 3D models with multiple model sections. Standard solutions allow a single model upload, which precludes the ability to combine different sections, such as representing multiple faces of an outcrop or a continuous exposure that stretches over a large distance and requires splitting into multiple chunks to facilitate processing.

The V3Geo repository is implemented as a PostgreSQL database hosted on Amazon Web Services (AWS). PostgreSQL (https://www.postgresql.org, last access: 7 March 2022) is a well-established, open-source relational database system and is used to store user profiles and model metadata as tables, which are queried using Structured Query Language (SQL). The virtual 3D models, images and other static files are stored using the AWS Simple Cloud Storage (S3) accessed through the CloudFront content delivery network (CDN). A GraphQL API is used to interact with the underlying database. This was chosen over more traditional representational state transfer (REST) APIs for a number of reasons: to limit the number of API endpoints to implement and maintain, to separate the API from the needs of specific client applications, and to minimise the number of API calls needed to retrieve data from the server. From our experience, use of a graph approach reduces the burden on the server side, allowing client applications to tailor their queries and data usage according to actual needs without having to specifically alter the API endpoints. User authentication and security are performed by the external Auth0 service, which is in turn used to manage the authorisation within the V3Geo database through the GraphQL API. This allows user profiles to be established and single sign-on (SSO) to be used to manage user identities, data protection and roles within V3Geo. Finally, two client web applications provide the V3Geo front end for users. The main website application (https://v3geo.com, last access: 7 March 2022) is the primary front end to browse, view and manage model data. Virtual 3D models are visualised on demand using a client web viewer, which is based on WebGL for 3D graphics.

This section describes the main user-facing characteristics of the V3Geo system at the time of writing.

An important advantage of V3Geo is the ability to store and access large 3D models with multiple sections, allowing wide areas and high-resolution data to be incorporated into the repository. In addition, we believe that a technical quality control ensures a minimum standard to the models in the database so that they can be relied upon in scientific and professional use and thus furthers the community expertise level and adoption of virtual 3D models in geology and geoscience. This reflects the general aim of ensuring rigorousness in 3D modelling approaches in photogrammetry and laser scanning to ensure (1) reliability of results and error reporting in scientific publications based on the 3D models and (2) a general reliability when these models themselves are made available for use by others. It is out of the scope of this article to discuss potential error sources in detail; readers are instead referred to earlier work by e.g. Chandler (1999), Buckley et al. (2008a), Eltner et al. (2016) and James et al. (2019) for reflections on issues related to data quality.

New public submissions to V3Geo are therefore subject to a basic technical “review”, so that they meet fundamental requirements related to suitability of content, georeferencing and scaling, 3D mesh and texture size, and model formatting. This practice has long been built up through workflow developments and internal quality control by the authors, and our experience of interacting with potential contributors, gathered since the initial release of V3Geo in April 2020, shows that there is value for the wider community in putting in place an approach to ensure reliability. It is important to note that the technical check is only to ensure the 3D models themselves meet a minimum standard: it is not a traditional scientific review, such as applied to a journal contribution, and any metadata descriptions that include observations, interpretations, or references to further work are the responsibility of the author to scrutinise prior to submission. It is instead analogous to an initial assessment by a journal's editorial team to ensure a submission meets the published guidelines. The V3Geo technical review is currently performed by members of the author group, though we plan to establish an “editorial board” of technically savvy geoscientists who have experience with 3D modelling methods and applications.

To facilitate high-quality contributions to V3Geo that successfully meet the technical requirements (and therefore a fast turnaround time), we have prepared guidelines for model preparation, published for the benefit of the geoscience community (Appendix A and in Howell et al., 2021). Although prepared with V3Geo in mind, they are equally applicable to more general 3D modelling and viewing software. The provided workflows document an approach to 3D model acquisition and processing that we have found to be successful for many applications, but they are not the only way to reach high-quality models. Rather, they provide one route that we hope is useful to new users of 3D models in geoscience and offers some guidance for more experienced users who wish to contribute models to V3Geo at a level of detail not supported by contemporary alternatives to sharing.

V3Geo expects two main parts to be prepared for any contribution: (1) 3D model data files; (2) metadata describing the model and characteristics of the field area, hand sample, etc. A prospective author to V3Geo should register with the repository using the web application at https://v3geo.com (last access: 7 March 2022). Here, it is possible to create a new model page and decide whether a model has public or private scope. This document covers public contributions with technical review. The model page is populated with the required fields, CC license, image(s), georeference and tags covered in Sect. 2.2.3. In addition, a file dropzone allows model sections to be uploaded (can be several per published model) as zipped files (according to Appendix A). Once completed, a preview of the model page can be checked before it is submitted for technical review. After submission, a simple review process is started, where the review team is notified and the 3D model is inspected according to the guidelines (Appendix A). The technical check will result in either (1) approval of the model and publication to V3Geo or (2) specific feedback for the author to improve the model and invitation to resubmit. Examples of common feedback are provided in Table 1.

Based on our experience, the first model submission from a particular author may result in issues with the data quality being highlighted and suggestions on how to address these sent to the author for consideration. Later submissions are then prepared according to the guidelines and approved. Hence, by publishing this guideline we hope to aid new users and contribute to an increased standard of 3D model quality in the geoscience community.

The main requirements for preparing 3D model files for submission are related to georeferencing, ensuring a balanced resolution of model mesh and texture per input model section and the final export format. Considerations and guidelines for preparing and formatting a model for V3Geo submission are documented in Appendix A.

The result of 3D model processing and formatting is one standard OBJ file per model section, which is then ready for submission to V3Geo. At the time of writing (and see Sect. 5.1, COVID-19 perspective), model files are uploaded as part of the V3Geo model submission process but must be converted by the project team before publication. Submitted model files are checked and, once they meet the technical guidelines, are converted into optimised tiled models for uploading to V3Geo. In this step, an automatic algorithm (based on the initial steps proposed by Buckley et al., 2008b) converts each OBJ file into a hierarchy of LOD tiles suitable for streaming over the Internet. This tiled model is uploaded to the V3Geo storage before the model itself is approved for publication and becomes visible to general site users.

As there is increasing utilisation of the 3D geoscience approach, there is a growing body of scientific literature being published (see e.g. Bond and Cawood, 2021, for coverage of virtual outcrop data). At present the main method to include the 3D model is through static screen grabs, which are typically unsatisfactory given the fundamental benefit conferred by viewing field localities from different perspectives in a 3D viewer. Our goal is that V3Geo will become a standard repository for 3D geoscience models linked to scientific publication. At present V3Geo includes a number of datasets which have been used in scientific publications, such as virtual outcrops Beckwith Plateau (https://v3geo.com/model/101, last access: 6 February 2022; Eide et al., 2014), Panther Tongue (https://v3geo.com/model/42, last access: 6 February 2022; Enge et al., 2010), Kvalpynten (https://v3geo.com/model/90, last access: 6 February 2022; Anell et al., 2021) and numerous Yorkshire coastal outcrops (Rahman, 2019). This number is expected to grow significantly in the coming years.

One of the most obvious applications for virtual 3D models is as the basis for virtual field trips. There are several ways that V3Geo can be used to build VFTs. The simplest is to use the V3Geo web viewer to direct students to virtual outcrops and encourage them to make observations and measurements in much the same way that they would in the field. A simple example of this was during the COVID-19 pandemic, when field teaching was not possible: University of Aberdeen fourth-year undergraduate students used the St Cyrus model (https://v3geo.com/model/120, last access: 7 March 2022) as a replacement for an existing field trip to the same location. Course exercises were modified to work from the V3Geo model, and it was possible for students to observe and map the sediment–lava interaction in the Lower Devonian sections that are well exposed in the cliffs behind the bay. A similar use was made of the Ainsa Quarry model (https://v3geo.com/model/30, last access: 7 March 2022) by Imperial College London students on a VFT to the Pyrenees. As the V3Geo viewer can be embedded in external websites, it is also possible to incorporate multiple models within online virtual field trips, allowing participants to view and learn from the various localities in the context of a more extended virtual field guide. Models can also be combined with Google Earth to provide regional context, while the V3Geo models give detailed views of an outcrop locality.

At UoA, the most widely used method for delivery of VFTs using data from V3Geo is through the connection to LIME. This allowed us to build more complex and interactive VFTs for the students. Pugsley et al. (2021) provide a detailed overview of the VFT creation methodology and teaching methods as well as a review of the outcomes of running VFTs at the University of Aberdeen. This work is based on two VFTs run for MSc students – an 11 d trip to Utah and a 5 d trip to the Spanish Pyrenees. Both trips were run almost exclusively in LIME and made extensive use of models from V3Geo (Fig. 5). The Utah VFT used a total of 48 virtual outcrop models, 21 from V3Geo, all augmented with other geological data. The students ran LIME locally or via a virtual machine hosted by the university, and they downloaded the project files required for the day each morning. Using V3Geo as the repository for the 3D models kept the project files to a minimum size and made it possible for participants with data download restrictions to participate effectively.

The first release of V3Geo was made in April 2020, with development fast-tracked in response to the first wave of COVID-19 lockdowns affecting Europe and much of the world from March 2020. Although parts of the system were complete at this point, it was not planned to make the public release of V3Geo until late 2020, when more functionality for registered users had been completed. However, restrictions on travel and in-person teaching in educational organisations resulted in cancellation of conventional field-based teaching and excursions, requiring alternatives to be quickly established. The launch of the browse and viewing functionality of V3Geo was therefore made to help educators and the geoscience community to weather the pandemic, even it was not complete according to the original project vision. V3Geo continues to evolve, with a roadmap for future direction in place to ensure the repository develops as a resource for publishing 3D data for the scientific and professional community. The following section outlines some areas of development prioritised in the shorter term.

V3Geo currently allows storage and viewing of high-resolution 3D models representing the surface state at the time of capture. However, the repository has been developed to incorporate additional interpretations created using the 3D models, such as marking key stratigraphy, structure, or geomorphological or geological features. Authors will be able to submit their interpretations together with a model or potentially create new interpretations based on another published model within the database. Interpretations are planned as overlay layers on top of the 3D model data (e.g. Buckley et al., 2019) and will have access individually governed by similar system roles to those for the model data. The aim is to facilitate publication of interpreted models made within the scope of a student project or research article, allowing open access to datasets that have traditionally been difficult to share in 3D without resorting to 2D screenshots or interpretation panels that are non-interactive and therefore difficult to appreciate and reproduce by publication readers.

Several measures will assist authors in publishing 3D models on V3Geo. It is planned to introduce an automatic converter to facilitate the process of integrating the 3D model files into the V3Geo database. Currently this is performed by the project team following submission by authors, using the tiled model converter (Sect. 3.3) and upload API. In the future this step will be deployable on the V3Geo cloud infrastructure to reduce time from submission to publication. Some further optimisation is being carried out to ensure robustness to the many different 3D model cases and data qualities that are generated – aided by adopting model preparation approaches such as those advocated in Appendix A – before deployment on the server is completed.

File formats for 3D graphics on the web are not yet fully standardised between processing applications and in-browser rendering, particularly for tiled models. As one of the most widely used 3D exchange formats in graphics software, the OBJ format is currently favoured for submission to V3Geo, even if it is not efficient in terms of rendering and streaming. Internally, the V3Geo tiled models are currently stored in a JSON format with supplementary binary geometry arrays and JPG image textures. However, a number of tiled model formats are now gaining traction for representing large, real-world 3D environments required for applications such as V3Geo. Examples are the Open Geospatial Consortium (OGC) Community standards for 3DTiles (https://www.ogc.org/standards/3DTiles, last access: 7 March 2022) and i3S/Scene Layer Package (SLPK; http://docs.opengeospatial.org/cs/17-014r5/17-014r5.html#89, last access: 7 March 2022) formats that have the potential to address issues around interoperability and ensure standard coordinate reference systems are enforced for field-based datasets.

Publication of data is becoming increasingly recognised in science, with open-data initiatives being established to govern the release of publicly funded datasets. Funding agencies require data management plans in grant applications, and permissive licenses for data are becoming standard, even though making large 3D datasets has conventionally been problematic. Publishing 3D datasets to V3Geo will conform to the drive for open-data access, allowing authors to use V3Geo as a repository for public 3D models as well as a landing page for supplementary 3D datasets supporting a journal article. Although acceptance and author credit for data publication is normalising, we plan to introduce the option for supporting digital object identifiers (DOIs) in V3Geo, so that the repository conforms more closely to FAIR principles. This will also consider versioning and a transparent history of changes, allowing older or obsolete 3D models to be superseded by newer versions generated using enhanced processing workflows.

An important consideration for V3Geo, as for other community-driven initiatives realising infrastructure, databases or software, is the future development and maintenance needed to create a sustainable and long-term resource. This is especially pertinent for systems based on rapidly changing web and 3D technology, where innovations and standards are in continuous evolution. For V3Geo, large dataset storage and traffic resulting from high engagement from users as well as maintenance and development by skilled engineers needed to work on a complex interaction of components require an associated model for long-term operation. While this is not yet determined, we are engaged with public agencies, learned societies and even private actors to support and evolve the repository. In the shorter term, a subscription approach for private and organisation storage is one potential means to ensure the system's durability and public accessibility for the scientific community.