Incorporate temporal topology in a deep‐time knowledge base to facilitate data‐driven discovery in geoscience

Data‐driven discovery in geoscience requires an enormous amount of FAIR (findable, accessible, interoperable and reusable) data derived from a multitude of sources. Many geology resources include data based on the geologic time scale, a system of dating that relates layers of rock (strata) to times in Earth history. The terminology of this geologic time scale, including the names of the strata and time intervals, is heterogeneous across data resources, hindering effective and efficient data integration. To address that issue, we created a deep‐time knowledge base that consists of knowledge graphs correlating international and regional geologic time scales, an online service of the knowledge graphs, and an R package to access the service. The knowledge base uses temporal topology to enable comparison and reasoning between various intervals and points in the geologic time scale. This work unifies and allows the querying of age‐related geologic information across the entirety of Earth history, resulting in a platform from which researchers can address complex deep‐time questions spanning numerous types of data and fields of study.


| INTRODUCTION
Studies of rock, mineral, fossil and sediment samples generate a significant amount of data for studying Earth history, particularly the co-evolution of life and environments over billions of years.Geologists sometimes use 'absolute dating' methods (e.g.carbon dating) to determine the numerical ages (reported in years) of samples (Watchman & Twidale, 2002).However, these methods typically require time consuming and costly analyses of radiogenic and stable nuclides (isotopes) with mass spectrometers.In addition, many samples lack sufficient concentrations of these isotopes for absolute dating (i.e.their concentrations fall beneath the detection limits of mass spectrometers), and the numerical ages from absolute dating methods are associated with confidence intervals that may represent millions of years of uncertainty.
For these reasons, geologists have created a system of 'relative dating' based on the layers of rock (strata) that accumulated over time and recorded the sequence of events in Earth history, such as changes in sea level, environments and life over time.This system, known as the geologic time scale (GTS), consists of hierarchically organized time intervals (e.g. the Permian, Triassic and Jurassic Periods), which have been identified with standard geologic techniques and principles.Samples (e.g.fossils and minerals) are assigned to these time intervals based on their relative positions in the global stratigraphic sequence.In many cases, the absolute ages of the time intervals are known.For example, the Permian spanned from 298.9 to 251.902 Mega anna (Ma), wherein 1 Ma equals a time span on 1 million years (Cohen et al., 2013).Accordingly, fossils of animals dated to the Permian Period lived between roughly 299 and 252 million years ago.The GTS, therefore, is a crucial component in geoscience research, providing scientists with a means of constraining the ages of geologic samples and describing the order of events in Earth history without absolute dates.
The GTS is the result of research in the fields of geochronology and chronostratigraphy (Gradstein et al., 2012a(Gradstein et al., , 2012b(Gradstein et al., , 2020;;Ogg et al., 2016;Zalasiewicz et al., 2013).Geochronology is the science of absolute dating (i.e.numerical intervals and points) of geologic samples, and stratigraphy is the discipline dedicated to identifying, describing and naming contiguous packages of strata (stratigraphic units).Chronostratigraphy deals with the relative spatial relationships of stratigraphic units on regional and global scales as well as their relation to geologic time.Stratigraphic units accumulate across laterally extensive areas over time, layer by layer, as vertical successions.Stratum age increases with relative depth beneath the surface, except in places where strata have been overturned by tectonic processes, and stratigraphic units can be traced (or correlated) across regions.Chronostratigraphy focuses on stratigraphic units, which may be located in different regions but represent the same intervals of Earth history.These time-equivalent packages provide the physical basis for the GTS.
Overall, the culmination of work on the chronostratigraphy of the world is a formal GTS called the International Chronostratigraphic Chart (ICC; Cohen et al., 2013); it was referred to as the International Stratigraphic Chart (ISC) prior to 2010.The chart divides Earth history into hierarchically organized geologic time units, which are called 'chronostratigraphic units' or 'geochronologic units', depending on context.By convention, a 'chronostratigraphic unit' is a body of rock, and a 'geochronologic unit' is a past interval of time.Nonetheless, chronostratigraphic and geochronologic units are somewhat interchangeable.Whereas the time-rock (chronostratigraphic) units include eonothems, erathems, systems, series, and stages, the time (geochronologic) units include the corresponding eons, eras, periods, epochs, and ages (Cohen et al., 2013).Therefore, in practice, the Permian is both a system and a period.The boundaries between the chronostratigraphic units of the ICC are physically represented by Global Boundary Stratotype Sections and Points (GSSPs), which are real places located around the world and marked by golden spikes placed between strata (Gradstein et al., 2012a(Gradstein et al., , 2012b(Gradstein et al., , 2020;;Zalasiewicz et al., 2013).
The dataset and code presented here are from our recent work on a deep-time knowledge base, which includes knowledge graphs for a list of global and regional geologic time scales, an online service for the knowledge graphs, and an R package of functions for accessing and using the service in data-intensive studies.A highlight of the work is the incorporation of temporal topology in the knowledge graphs and the R functions.More details will be illustrated in the following sections.

KNOWLEDGE BASE AND SERVICE CONSTRUCTION
There are many versions of the GTS in existence due to its long history of the development and the logistical challenges of creating an accurate chronology for Earth history (Gradstein et al., 2012a(Gradstein et al., , 2012b(Gradstein et al., , 2020;;Kulp, 1961).Regional time scales (e.g.Siberia, China, Australia, and North America) were developed for each region, and they can be different from each other in terms of the numbers, names and hierarchical relationships of their chronostratigraphic units and the durations of their corresponding geochronologic units (Menning et al., 2006;Ogg, 2004).Then, ICC was created to represent a synthesis of time scales with global significance, and it also provides the framework for regional divisions.Nonetheless, the units from different regional time scales may be correlated with each other if they represent overlapping time intervals.For instance, the Kungurian Age (global) -a formal subdivision of the Permian System in the ICC -signifies an interval of time (280 to 270.6 ± 0.7 Ma) that overlaps with the following regional units: the Hessian Age (North America), Mangapirian Age (New Zealand), Kungurian Age (Russia), Luodianian Age (South China), Bolorian Age (Tethyan) and Rotliegend Epoch (West Europe).Due to the complexities of dating and correlation timerock units, the GTS continues to undergo revision with advancements in geochronology and stratigraphy.The ICC/ISC has been updated more than 19 times since 2004.For instance, the numerical ages of the lower and upper boundaries of the Permian System were recently revised between 2004 (299.0-251.0Ma; Gradstein et al., 2004) and2020 (298.9-251.902 Ma;Cohen et al., 2013) based on new absolute dating work, and in 2020, the Middle Pleistocene Stage received a formal name for the first time: Chibanian.This sort of semantic heterogeneity hinders progress in data-driven discovery.Data resources use a variety of time scales, including regional and global versions, and legacy data do not always reflect recent changes in the ICC/ISC (Mascarelli, 2009).In this context, the study of Earth history would benefit from a knowledge base and toolkit for reconciling heterogeneous terminology among versions of the GTS.
To tackle the challenges associated with heterogeneous GTS terminology, knowledge graphs (e.g.ontology and vocabulary) have been developed with semantic technologies (Berners-Lee et al., 2001;Fox & Hendler, 2009).Each knowledge graph represents a piece of knowledge in the real world, in which nodes are entities of interest and edges are the relationships between these entities (Hogan et al., 2020).Knowledge graph is trying to define the knowledge in a quantitative and machine-readable way.It normally stores the knowledge in a graph database, which is different from relational database.In this way, it is faster in querying and ready for knowledge reasoning (Ma, 2022).Early works were focussed on the formal machine-readable model of GTS.Cox and Richard (2005) created a model of the GTS using the Unified Modelling Language (UML), which can be converted to eXtensible Markup Language (XML).The NASA Semantic Web for Earth and Environment Technology Ontology (SWEET) contains a GTS vocabulary, with standardization of numerical time points (Raskin & Pan, 2005).Ma et al. (2011) designed an ontology for GTS with multilingual labels.Ma et al. (2012) presented a case study of linking databases and data visualization powered by the GTS ontology.In recent years, Cox and Richard (2015) developed a new ontology based on their previous model (Cox & Richard, 2005), and created different versions of vocabulary for the international GTS.Most recently, the Time Ontology (Cox et al., 2016;Cox & Little, 2017) has been implemented into the GTS ontology to refine time reference systems and topology of geologic time units.
In our work, Ma, Kale, et al. (2020a); Ma, Ma, and Wang (2020b) created a new version control structure based on the ontology and vocabularies of Cox and Richard (2015).Ma et al. (n.d.) created vocabularies for 17 regional GTS across the world.The resulting knowledge graph has 19 international geologic time standards (from year 2004 to 2020) and 17 regions ones (Table 1).For the regional data, there is no historical data considered this time.We stored these knowledge graphs in a graph database instead of relational database for several reasons: (a) Data querying is faster in a graph database than that in a relational database.Although not much difference in our data considering the size, it is prepared for the future big data management.(b) The properties of the edges (relations) and nodes (entities) in the knowledge graph are structured and well defined based on existing communitylevel ontologies, schemas and vocabularies (Table 1 in Ma et al., n.d.).These make the knowledge graph interoperable to meet the FAIR principles.(c) Following community standards makes the edges (relations) and nodes (entities) in our knowledge graph readily be used by other knowledge graphs or applications.The outputs of Ma, Kale, et al. (2020a); Ma, Ma, and Wang (2020b) and Ma et al. (n.d.) are bundled into a comprehensive deep-time knowledge graph, stored in a graph database (Virtuoso) and made open as a service on the World Wide Web.Knowledge about the GTS can be queried through
The DeepTimeKB package was developed using the R language in RStudio, an integrated development environment for R. The deep-time knowledge graphs use semantic Web standards including RDF (Resource Description Framework), RDFS (RDF Schema), SKOS (Simple Knowledge Organization System), and the Time Ontology (Cox & Little, 2020).In particular, the temporal topology specified in the Time Ontology was implemented in all the knowledge graphs for iGTS and rGTS as well as the DeepTimeKB package in R. The 'topology' in our study only limits to topological temporal relations defined in time ontology (Cox & Little, 2020), meaning binary relations on intervals and instants (e.g.meet, overlap, and during; Allen, 1984;Cox & Little, 2020).The service of the knowledge graphs was built by using Virtuoso Open-Source Edition from the OpenLink Software, Inc. GitHub was used for version control of all software code and collaboration between team members.

FUNCTIONS OF THE KNOWLEDGE BASE
The DeepTimeKB package is currently shared on GitHub (Ma, Kale, et al., 2020a;Ma, Ma, & Wang, 2020b) and is in the process of being formally released on the Comprehensive R Archive Network (CRAN).The package is built upon the knowledge graphs (i.e.vocabularies) for international geologic time scale (iGTS) with version control and regional geologic time scale (rGTS; Ma, Kale, et al., 2020a;Ma, Ma, & Wang, 2020b;Ma et al., n.d.).The iGTS covers all geologic time units in the ICC/ISC from 2004 to 2020 and the rGTS contains 17 regions (Table 1).Both iGTS and rGTS record information on (a) the boundary ages of chronostratigraphic units (i.e.start and end times of geochronologic units), (b) their hierarchical structure, such as the assignment of the Jurassic System to the Mesozoic Erathem, (c) the topology of the geochronologic units (Cox & Little, 2020), such as the other time intervals before, after, and contained within a given time interval and (d) identifiers to uniquely mark each GTS.The encoding of the GTS identifiers was based on semantic technologies (Ma, Kale, et al., 2020a;Ma, Ma, & Wang, 2020b;Ma et al., n.d.), such that each version of the ICC/ISC in the iGTS and each region in the rGTS has a unique identifier, which applies to all of the knowledge graphs and establishes a well-organized structure for chronostratigraphic/geochronologic unit management and query.Moreover, the iGTS has specific information for locations and ages of the GSSPs and structures to track different versions of the ICC/ISC.
To leverage the usage of the built knowledge base and help researchers explore multiple deep-time data resources in workflow platforms, we created 12 functions for the R package DeepTimeKB (Table 2), which can be divided into 3 categories: (a) knowledge query for extracting information on geologic time units, hierarchical structure, and GSSPs; (b) knowledge visualization for plotting GSSPs on a global map; and (c) knowledge reasoning for deducing topological relationships between chronostratigraphic/geochronologic units.
The gts.listRegion function of the DeepTimeKB package provides an overview of the vocabularies of iGTS and rGTS in the knowledge base.It returns the identifiers and names of all vocabulary schemes in the knowledge base.The function gts.list takes two input parameters: 'region' (i.e. a vocabulary scheme) and 'level' (i.e.eonothem, erathem, system, series, and stage, or eon, era, period, epoch and age), and returns all chronostratigraphic or geochronologic units that meet the parameters.The properties associated with a certain geologic time unit can be extracted by running the function gts.In some cases, the level of a specific geologic time unit is needed for further analysis.The function gts.level takes the literal name of a geologic time unit and returns its vocabulary scheme and level.The function gts.point takes a specific numerical time point and returns all the geochronologic units that contain the time point.The function gts.range can extract the start and end times, vocabulary scheme and duration of a specific geochronologic unit.In this function, a parameter 'iscVersion' can be set for querying the unit from a specific version of ICC/ISC in the iGTS.The parameter 'region' can be set for querying the unit from a specific vocabulary scheme in the rGTS.The function gts.within enables the querying of geologic time units inside a specific geochronological range defined by two time parameters: 'geotime1' (start) and 'geotime2' (end).The function gts.hierarchy can return the subordinate units contained within an input unit, and thus show the hierarchical structure in the GTS.
The function gts.gssp queries the GSSP information from the knowledge base.All GSSP information can be obtained by running gts.gssp(), and the parameter 'is-cVersion' can be set to query the information only from a certain version of ICC/ISC in the iGTS.The returned information includes vocabulary scheme, GSSP name, longitude and latitude.The function gssp.map can visualize the GSSP information by plotting all GSSPs on a map (Figure 1), in which the parameter 'iscVersion' can be used to set the version of ICC/ISC in the iGTS.Examples of the functions mentioned in this paragraph are listed in the Appendix S1.
The temporal topology (Allen, 1984) summarizes the relationships between intervals and points of time.The Time Ontology (Cox & Little, 2020) implements the temporal topology in a machine-readable structure and paves the way for other knowledge graphs to adopt, including our work on iGTS, rGTS and the DeepTimeKB package.There are 13 elementary relationships between time intervals (Figure 2), which can be inferred by comparing the start and end time points of any two intervals.Geochronologic units (e.g.epoch) and boundaries (e.g.GSSP) are also time intervals and points.In the work of Cox (2020), the Time Ontology was used to specify the topology between geochronologic units.We reused that in the knowledge graph for iGTS, in which our structure for ICC/ISC version tracking is also able to show the changes in topology at a few places on the ICC/ ISC, such as the relationship between the Quaternary Period and other intervals.In the knowledge graph for rGTS, the topology between geochronologic units was also included.More specifically, the function gts.topo in the DeeptimeKB package also implements the temporal topology.The input of this function is two geochronologic units with the following parameters: (a) unit name and (b) region in the rGTS or version of ICC/ISC in the iGTS.The output will be one of the relationships shown in Figure 2. Comparing with the specified topology in vocabularies of iGTS and rGTS, the gts.topo function gives flexibility to infer the topological relationship between any two given internals, whether they are from the same GTS or not.

AND DISCUSSION
Different components and spheres of the Earth interact with each other, forming a dynamic system.This planetary system encompasses geochemical, biological, geological and geophysical processes operating across a wide range of spatiotemporal scales (NASEM, 2020).Cyberinfrastructure, analytical tools and computational approaches can support far-reaching investigations that help improve our understanding of this system (Fan et al., 2020;Hazen et al., 2019;Morrison et al., 2020).Geologic time units/concepts/intervals are key components of many open-access geoscience databases, such as EarthChem (Walker et al., 2005), the Paleobiology Database (PBDB), Macrostrat (Peters et al., 2018), the Geobiodiversity Database (Fan et al., 2014), the OneStratigraphy Database (Fan et al., 2020).Examination of these resources suggests that, while geologic time represents a common axis for collecting, presenting, and synthesizing multi-source data, the heterogeneity and ambiguity of GTS information hinders the effectiveness and efficiency of data integration and analysis.The geologic time units/concepts/intervals data be divided into two categories.One is the data collected from standards and community-level references.The other is time scale or interval table to inform their numeric ages, such as those used in different databases (e.g.Macrostrat 1,2 PBDB 3 ).The first one is partly professional readable (although no single professional knows all of them) but not machine readable.The second one is 'local' machine readable, because these dictionaries are based on their own rules and showed in a way that professionals can understand, but not a universal way that can be interpreted by all machines.For example, 'min_age' 1 and 't_age' 2 used in Macrostrat and 'lag' 4 in PBDB represent the top age of an interval.The data with 'local' dictionaries are machine readable by 'local' machine, while it is less readable for external machines (e.g.software programs or applications).Furthermore, most of them do not track the version history.For example, the 'Jurassic' concept collected from publications in 2000 might have a different indication of time span comparing with that collected from publications in 2020.Thus, if a system or a machine is going to utilize the data from these databases, a problem is the automatic understanding of these geologic time concepts, which is also called heterogeneity or lower-interoperability.Here, in this work, we try to build a 'global dictionary' through knowledge graphs to overcome the issue of GTS heterogeneity.In one of our pilot studies (Wang et al., 2018), we showed that the machine-readable vocabularies of rGTS are a complementary way to explore the fossil collection data in PBDB.This pilot study is available on the website of the deep-time knowledge base (DeepTimeKB, 2021; Figure 3).
The deep-time knowledge base (Figure 4) will assist researchers in overcoming challenges related to GTS heterogeneity.Our knowledge graphs consist of different versions of the international GTS and regional GTSs (Ma, Kale, et al., 2020a;Ma, Ma, & Wang, 2020b;Ma et al., n.d.).This work translates a comprehensive set of geologic time standards into a structured, machinereadable format.A free online service is available for knowledge query.The R package DeepTimeKB allows end users, particularly those lacking advanced coding and programming skills, to follow straightforward steps and expedite the query of GTS information.When using the package in a workflow platform, a user does not need to write detailed code in the raw query language.Instead, the user can simply specify the parameters for a function related to the GTS, and then quickly receive results (see examples in the Appendix S1).Although this comprehensive knowledge base includes a wide array of vocabulary schemes and geologic time units, the DeepTimeKB provides several categories of functions for users to familiarize themselves with the knowledge graphs and then implement sophisticated tasks.

F I G U R E 4 Components of the deeptime knowledge base and its potential applications in data-driven geoscience discovery
The combination of the deep-time knowledge graphs, the online service, and the R package into the so-called deep-time knowledge base will enable exploration of various geoscience databases (Figure 4).Specifically, the DeepTimeKB package is intended for use in workflow platforms, making it convenient to access and explore multiple online geoscience databases in real time.One example is data exploration with the PBDB, as illustrated in Ma et al. (n.d.).The total number of regional geologic intervals in DeepTimeKB is 496, while that in PBDB is 926.The number of overlapped intervals is 257.For example, 'Saxonian', 'Kathwai', 'Khorokytian', 'Suiningian' and 'Kulyumbean' are a few unique time concepts in DeepTimeKB.In comparison, 'Rancholabrean', 'Badenian', 'Colhuehuapian', 'Monroecreekian' and 'Dienerian' are a few unique concepts in PBDB., and 'Haweran', 'Wanganui', 'Castlecliffian', 'Nukumaruan', 'Mangapanian' are in both DeepTimeKB and PBDB.Our deep-time knowledge base offers a complementary resource that helps retrieve the boundary ages of rare units and then use them to query fossil collections from the PBDB.The large number of regional intervals used in PBDB may lead to future extension of our deep-time knowledge base, which will grow through detailed analysis of the sources and organizational schemes of additional intervals (i.e.there are potentially many other versions of the regional GTSs).
The deep-time knowledge base improves the machine readability of geologic time units, solves the issue of heterogeneous GTSs and enhances semantic interoperability between databases.It can be used to automate the process of tagging geologic time units in massive geoscience data.As such, the knowledge base can save a significant amount of time that is usually spent on manual conversion among GTSs and on the study of the complex topological relationships among chronostratigraphic and geochronologic units.Data-driven discovery requires data from multiple databases, affirming the need for database linkage and integration (Peckham et al., 2014).The deeptime knowledge base could serve as a common reference for connecting deep-time data from various data resources into a semantically cohesive framework, and then conducting data integration and analysis (Hazen et al., 2011;McGuinness et al., 2009).Given the various nominal and numerical representations of geologic time in geoscience data, the temporal topology embedded in the DeepTimeKB functions will offer powerful support for data cleansing and help set up a clear sequence of deep time across data resources.
Through workflow platforms, the data analysing processes are portable, maintainable, reproducible and shareable (Perkel, 2019), which will benefit big geoscience data analysis through open science and community of practice.For example, the developed deep-time knowledge base could be used in automatic chronostratigraphy, that is, a potential method for automatically developing stratigraphic time frameworks based on all available chronologic data.The international and regional GTSs extracted by text mining from multiple sources could constrain age models in such frameworks.For example, if a 'Jurassic' label is extracted from a literature or a database, the region and version this concept refers to can be determined by the workflow along with the metadata of the literature or database, then the age of this 'Jurassic' label is constrained.Furthermore, the deep-time knowledge base can be implemented with geoscience databases that require formal, comprehensive geologic time units.For example, if the PBDB integrates the deep-time knowledge base developed in this study, then the database will grow its options for end users that attempt to access data from rare geologic time units in regional GTSs.Such implementations may ultimately enhance the usability and value of cryptic data based on outdated GTSs.
The resulting knowledge graphs and knowledge base will facilitate FAIR principles from different aspects.In a broad sense, the knowledge graphs we created are data too.They were serialized in standardized formats and stored in a graph database.We also shared the source code of the knowledge graph on Zenendo and GitHub (see links in the section Code availability).As leading open source and open data repositories, both Zenendo and GitHub have their FAIR guiding principles.Moreover, we plan to make the knowledge graphs part of big organizations or programs to make them more FAIR.As for facilitating FAIR principles in other people's work, there are two ways.One way is to provide precise geologic time information for the downstream work, such as multi-source data retrieving and analysis.Our established deep-time knowledge base can quickly provide information for a large amount of heterogeneous geologic time concepts.The other way is for the upstream work.Existing or future databases can use our knowledge base as their local controlled vocabulary to make their databases more interoperable.
Building this knowledge base and the toolkit is a first step towards a big picture of automatic knowledge and data mining in geoscience.A few future works are also on our list.The current knowledge base does not fully track the change of the meaning of a concept and the underlying reasons.For example, the Pleistocene Epoch only included three Ages (Upper Pleistocene, Ionian, and Calabrian) before 2009.In 2009, the International Commission on Stratigraphy (ICS) decided to move the base of the Pleistocene to the base of Gelasian Age because the initial base of Pleistocene (i.e.Calabrian Age) did not reflect a global change (Mascarelli, 2009).There are other such changes that can be checked on the ICS Website (https://strat igrap hy.org/chart) as well as GTS-related books (Gradstein et al., et al., 2012a, 2012b, Gradstein et al., 2012c, Gradstein et al., 2020;Ogg et al., 2016).Such detailed information, if included in the knowledge base, will be a good reference for scientists.The current knowledge base only includes 17 regional geologic time scales, and it does not consider the historical and multi-lingual version of them, which may limit the ability of processing version conflicts and multi-lingual labels from the literatures Nevertheless, our knowledge graphs have a good extendibility as the fundamental conceptual structures are sound and stable.More regional geologic time scales, their geologic knowledge background, and the multi-lingual labels can be gradually added into the knowledge base in the future work.The functions in our R package are the first release and they many not perfectly meet all the needs from geoscience community.We will collect and review feedback from users and update those functions accordingly.
Another challenge in the long term is the sustainability and maintenance of the deep-time knowledge base.It is easy to do maintenance at present as the size of data is small and the system is not too complicated.However, as the ICC chars are continuously updating through ICS, there should also be a mechanism to update the knowledge base.One of the options is to form a working group (such as under the International Commission on Stratigraphy) to continue and expand the scope of the work on deeptime knowledge graphs.Recently, there have been some informal contacts between researchers (including us) across the world on that topic, and we hope to make it a reality.

| CONCLUSIONS
Geologic time units and GSSPs are key components in geoscientific research.As Earth history data continue to accumulate, data-driven discovery will require seamless access to multiple data resources, which in turn, will demand precise machine understanding of various geoscience concepts.The deep-time knowledge base presented in this paper provides an exemplar solution to the issue by focussing on the heterogeneous GTSs.Nevertheless, the full utilization of this knowledge base in data-driven discovery calls for applications of it in more deep-time studies, such as the co-evolution of geosphere and biosphere (NASEM, 2020).Future work should develop use cases in different application scenarios, translate the DeepTimeKB package into other programming languages (e.g.Python), and collaborate with geoscience databases, which need formal vocabularies of geologic time to enrich their discoverability and accessibility.By adopting open source and open science approaches, we hope the developed deep-time knowledge base will be a long-term, reusable resource for data-driven discovery in geoscience.

F
Thirteen topological relationships between time intervals (fromCox & Little, 2020).The grey double-arrowed bar represents time interval i, while the black double-arrowed bars are time interval j. in seven possible situations, there are 13 possible relations between i and j, considering both directions of i to j and j to i.only in the last situation when i and j are equal (last row in the figure), where the temporal relationships are the same for both directions.The boxes on the left show one example for each relationship F I G U R E 3 Using geologic time units in the geologic time scale of North America to explore fossil collection data within the United States.Here, the map shows part of the results within the time interval of the Lenoxian age (under Wolfcampian epoch of the Permian period).The demo Website is accessible at DeepTimeKB (2021).Fossil collection data are from the paleobiology database.Background geologic map is from USGS.All data in this figure are reproduced with permission.The geologic time bar is modified from UW-Macrostrat (https://github.com/UW-Macrostrat/ geo-times cale) All functions in the DeepTimeKB R package T A B L E 2 gts.listRegionGet the identifiers and names of all vocabulary schemes (international or regional) in the deep-time knowledge base gts.pointtime Get all the geologic time units that contain a specific time point gts.rangegeoConcept, region, iscVersion Get start time, end time, and duration of a specific geologic time unit gts.topo geoConcept1, region1, iscVersion1, geoConcept2, region2, iscVersion2 Get topological relationship of two given geologic time units gts.within geotime1, geotime2 Get all geologic time units within a time interval F I G U R E 1 Global Boundary Stratotype Sections and Point (GSSP) map of the 2018-07 international chronostratigraphic chart resulted by running the function gssp.Map (iscVersion = "2018-07")