Science is on an irreversible trajectory towards greater openness. Geochronology is no exception to this trend, as this journal illustrates with its open access and review policies. Funding agencies increasingly demand that research results and data are shared with the public. Currently, geochronological data are generally provided as flat tables of dates or isotopic ratio estimates. However, in other fields of science such as physics, it is common practice to share the raw unprocessed measurements along with processing instructions e.g.,. This paper moves geochronology in the same direction. It presents a mechanism to generate and store fully FAIR Findable, Accessible, Interoperable and Reproducible; data in the context of fission track analysis.

Unlike most other geochronometers, which require mass spectrometers to estimate parent-daughter ratios, fission tracks are observed under an optical microscope and counted by a human observer. In recent years, digital microscopy has moved fission track data acquisition from the ocular lenses of a microscope to the computer screen . Ongoing developments in artificial intelligence generate further opportunities to improve the throughput and accuracy of fission track data . But despite the richness of the digital datasets produced by these novel tools, fission track data are still reported as numerical summary tables. These tables require an unnecessary degree of trust between the “producer” and “consumer” of the data.

geochron@home is a free and open software platform that allows geochronologists to share their raw fission track data over the internet for perusal by peer reviewers, colleagues and the general public. geochron@home is a virtual petrographic microscope connected to a database with digital image stacks of etched fission track samples. The platform can be used to acquire, archive and inspect fission track data in full adherence to the FAIR data principles. In Sect. , we will describe geochron@home's software architecture in five steps. We will show that this architecture accommodates imagery from any type of digital microscope. It enables flexible workflows that can be adapted to four different applications.

Using image stacks of Mount Dromedary apatite, we will show how geochron@home can be used to count fission tracks in “private mode” (Sect. ); to archive published fission track datasets in “public mode” (Sect. ); to build tutorials for training purposes (Sect. ); and to crowd-source fission track data on the internet (Sect. ). Because geochron@home is free and open, it can be extended and improved by any interested party. We make some suggestions for future improvements in Sect. . We hope that the geochron@home's example will be followed in other geochronological disciplines, as this will benefit not only geochronology itself, but all the other disciplines that depend on it (Sect. ).

The geochron@home workflow separates the acquisition of microscope images from their analysis, providing the flexibility to accommodate data from different microscope manufacturers. The workflow can be broken down into five steps.

Acquisition of z-stacks of microscope images in reflected and transmitted light for each of the grains in a sample and, optionally, for the accompanying external detector (Fig. ). At University College London (UCL), this first step is currently accomplished by a Python macro within Zeiss' Zen Blue software. However, geochron@home can also accommodate imagery from other platforms, such as Fission Track Studio Zeiss; and TrackFlow Nikon;.

The five-step workflow can be used for several applications, including (1) conventional fission track analysis; (2) archiving published fission track results; (3) building tutorials; and (4) crowd-sourcing fission track data. The next sections of this paper will illustrate these applications with real world examples.

Administrators can define regions of interest (ROI) and count or edit the fission track coordinates of any grain in a geochron@home database. They can also assign other users to groups, and give these groups access to a subset of the projects in the database. Administrators have fine control over the permissions of the groups. For example, they can allow the members of one group to define their own ROIs, whilst requiring members of another group to count fission tracks in predefined ROIs. Groups provide a mechanism to compare and combine the results of multiple analysts of the same sample. Figure  illustrates this with two sets of fission track density estimates for the same sample of Mount Dromedary apatite , analysed by two users (PV and AC).

Let N1 and N2 be the numbers of spontaneous tracks counted by the two users over areas A1 and A2, respectively. Then their estimated track densities are given by ρ^1=N1/A1 and ρ^2=N2/A2. In this situation when the two areas overlap, N1 and N2 cannot be treated as independent Poisson counts because the two analysts will count some of the same tracks. It can be shown that, under simple (ideal) assumptions, the uncertainty of the ratio of the estimated track densities is given approximately by 1se(ρ^1/ρ^2)ρ^1/ρ^2≈1N1+1N2-2N0N1N2 where N0 is the number of tracks counted by both observers in the area of overlap (A0, say) between their respective ROIs.

The combined data plots of Fig. b and c contain two sets of counts for 25 grains, with ∑N1=686 (PV) and ∑N2=679 (AC), and a weighted mean track density ratio ρ^1/ρ^2=0.94 with relative standard error 0.013. PV counted 600 tracks in those common areas A0, of which 549 were also counted by AC. Conversely, AC counted 622 tracks of which 549 were also counted by PV. The ratio of the two analysts' track density estimates based just on the common area is therefore 600/622 = 0.965 with relative standard error 0.018, calculated from Eq. (), with N1=686, N2=679 and N0=549. This is close to the weighted mean ratio of 0.94 (Fig. c), and is slightly nearer to 1. It indicates that PV under-counts the Mount Dromedary apatite by 3.5 % relative to AC. The existence of “observer bias” is well documented . It is one of the reasons why fission track analysis is often done relative to age standards: observer bias does not have to be a problem provided that it is consistent between samples and standards.

Once a set of fission track images has been analysed and the analyst is confident that the results are accurate, the status of the results can be changed from private to public. This makes the results visible over the internet as a list of URLs, where each grain number and user ID corresponds to a unique address. The geochron@home archive (GaHa) brings fission track geochronology into the era of FAIR science. It allows peer reviewers to inspect the raw data from which thermochronological inferences are made. The archive is open to submissions from any fission track laboratory, free of charge. However, as mentioned in Sect. , it is also possible to establish a new archive elsewhere. At the time of writing, GaHa contains data for three studies:

: this is an LA-ICP-MS based fission track study of detrital apatite from the northeastern Tibetan Plateau. It contains image stacks of semitracks in 1146 apatite grains from 16 different samples.

Given the right permissions (assigned by an administrator), users can build tutorial pages by annotating features in fission track images. These features can be tracks or other objects such as scratches, inclusions, dislocations or holes. A selection of tutorial pages is presented to new users when they first log into geochron@home. They must complete the tutorial before being allowed to count fission tracks. The tutorial pages can be revisited at any time by visiting the corresponding link on the geochron@home landing page.

Some exemplar annotations in the current tutorial pages were made by an experienced fission track analyst (Andrew Carter). This basic tutorial provides a quick and easy mechanism to train novice users in the art of fission track analysis. The tutorial pages are in their infancy and offer only a limited degree of interactivity. Users can click on features to read the annotations. A more detailed, comprehensive tutorial will grow with the wider input from experienced analysts and future plans include the addition of fully interactive “quizzes” (Sect. ). The limitations of the current tutorials are apparent in the results of the crowd-sourcing experiment described in the next section of this paper.

In 1906, Sir Francis Galton visited a county fair in which a contest was held to guess the weight of an ox. 787 people participated in the event. Galton discovered that the median of all their estimates was within 0.8 % of the true weight of the ox and more accurate than 90 % of the individual estimates. Such is the “wisdom of crowds” . Similar effects are seen in fission track geochronology. An interlaboratory comparison study by showed that the average of several fission track age estimates is closer to the known age of mineral standards than the age obtained by individual observers. Inspired by these previous experiments, we used geochron@home to bring fission track analysis to a proverbial “county fair” of citizen scientists.

We here present the results of a type of “crowd-sourcing” experiment that was carried out at UCL as part of an undergraduate course in isotope geology. 68 students were asked to create an account on the geochron@home server by selecting a unique username and password. After spending a few minutes to complete the compulsory tutorial (Sect. ), they were asked to count fission tracks in randomly assigned image stacks of Mount Dromedary apatite. Each student was required to analyse at least 15 grains, using pre-defined ROIs. In a matter of hours, the students amassed a dataset of 37 331 fission track counts in 34 grains. This large dataset was parsed into separate data files – one for each student – which they processed during an assessed programming exercise.

The raw counts for the 68 students are given in Table  and Fig. , for the same selection of 25 grains that were analysed by AC and PV in Sect. , along with corresponding counts made by PV. Everyone analysed the same ROIs. Variation between the mean values is to be expected because of the differing numbers of tracks due to varying areas and U contents of the grains. But the variation between counts within grains (shown by the standard deviations in Fig. ) is due entirely to differences between the students' recognising and counting exactly the same tracks. These standard deviations increase with the mean number of tracks and are considerable in size, being on average about 40 % of the mean. With expert trained counters one would expect much smaller differences between counts. Furthermore, the vast majority of students counted fewer tracks than PV did, often many fewer, and on two occasions someone counted no tracks at all. PV's count is always above the students' median and nearly always above the upper quartile (Fig. ).

The rightmost three columns of Table  contain summary statistics (ng, rp and sp) giving, for each student, the number of grains counted and the weighted mean and standard deviation of the ratios of the student's count to PV's count for each grain (definitions provided in the caption). The true number of tracks in each ROI is unknown, so we are comparing the students with PV (remembering that PV's count is also subject to error, which is the same for each student). rp measures agreement with PV on average and sp measures consistency over grains. For good agreement rp should be close to 1 and sp should be small. Perfect agreement is rp=1 and sp=0. For comparison, an rp=1.04 and sp=0.69 is obtained for PV and AC using just the counts in the common areas A0 from Sect. . With the caveat that the ROIs are different and PV's counts are different, this calculation shows that there is a group of 6 (possibly 9) students whose agreement with PV is comparable with AC's. This group does not include student 229 whose rp=0.95 but with a large sp=1.62. Students near the bottom of Table  have undercounted the samples by such a large degree that their work can be qualified as vandalism. Sect.  will suggest strategies to detect and counteract this type of behaviour.

In spite of the under-counting by the students, the ratio of their pooled track counts (rp) for two different grains is often close to PV's ratio, especially when comparing grains with similar track densities. This is illustrated in Fig. a for grains 23 and 25. For any point, the ratio of the two counts is given by the slope of a line from the origin to that point. Although the individual slopes vary considerably, the line with slope equal to the ratio of the total counts for all students (which is 0.97) passes close to PV's pair of counts (52, 50) shown by the blue dot (with ratio 1.04). It is interesting that the students with lower counts are mostly above this line and those with higher counts are all below it. There is a positive correlation between the pairs of counts, consistent with systematic observer effects (i.e., people who count a low value in one grain tend to count similarly low in the other, and vice versa). Comparing other pairs of grains shows similar results, with correlations varying between 0.4 and 0.9.

As mentioned in Sect. , geochron@home stores the actual track positions marked by the users. This raw data can be downloaded as a .json file and inspected in detail. Figure b shows two-dimensional histograms for the x–y positions of all the track positions generated by the students in grains 23 and 25. Visual comparison of the histograms with the optical images confirms that the students unanimously identified the most obvious semitracks, which contain a clearly visible etch pit and tail. Shorter and fainter tracks received fewer clicks. Datasets like this can be used to replace integer counts of fission tracks with probabilities, reflecting the ambiguity of some fission track datasets. Figure  also shows that some students counted the tails of the fission tracks rather than their etch pits, despite being told the opposite in the tutorial. Fixing this issue will require some improvements to the tutorial pages (Sect. ).

geochron@home has been in development for a decade and remains a work in progress. Planned improvements include:

Length measurements. The geochron@home archive already contains horizontally confined fission track measurements . However, these results must be generated externally (e.g., using Fission Track Studio) and uploaded via a .json file. The virtual microscope currently lacks the functionality to generate length data within geochron@home. This functionality will be added in a future update.

These improvements will be made by ourselves pending additional funding. However, because geochron@home is free and open software, we invite any interested parties to join the effort and extend or improve our code.

This paper introduced geochron@home, a software platform for FAIR fission track analysis. We demonstrated four different applications of this platform using real data. Putting the FAIR data paradigm into practice, all the imagery, counts and source code for this paper are publicly available on the geochron@home archive. Using these resources, the reader can reproduce all the results that were presented in this publication. We encourage other geochronologists to follow this example. FAIR workflows promise to address the reproducibility crisis in science .

Increased transparency will also introduce new challenges. As FAIR data practices become more widespread, the fission track community will need to define how extensively reviewers are expected to re-examine primary imagery and what level of discrepancy constitutes sufficient grounds for rejecting or disputing raw counts. Although these issues are beyond the scope of the present paper, they will require community-wide discussion as FAIR data reporting and review become more widely adopted.

geochron@home's rich archive of raw data can be reanalysed in the future. We anticipate that the adoption of FAIR data processing workflows will open up new research opportunities. For example, archived pairs of peer-reviewed fission track images and counts could be used to train the next generation of automated machine learning algorithms. Conversely, it is also possible that future improvements in fission track images analysis will be used to update the count data for published datasets, improving their accuracy.

Another advantage of the geochron@home workflow is the clear separation between image acquisition and image analysis. This separation lowers the hardware requirements for fission track geochronology and enables resource sharing. Because state-of-the-art digital microscopes are expensive, geochron@home makes it possible to envision a model in which sample preparation and scanning are outsourced to a centralized facility, while individual scientists analyze their own data online. In this way, a single microscope could serve multiple laboratories, making fission track analysis more accessible and affordable.

The fission track method has always been a test bed for new geochronological developments. Because fission track data are imprecise, the fission track community has solicited the help of statisticians and mathematicians to develop its analytical protocols. Other geochronological communities are still catching up with concepts and tools such as overdispersion, mixture modelling and radial plots, which have been commonplace in fission track analysis for decades . In a similar vein, the subjective nature of fission track identification has prompted the fission track community to organise inter-laboratory comparisons and round-robin studies long before other geochronological communities .

With the development of geochron@home, fission track thermochronology is once again ahead of the pack in terms of FAIR data analysis. geochron@home currently only stores images and counts. This is enough to reproduce the results of fission track studies using the external detector method, but not for LA-ICP-MS based data. FAIR data processing of LA-ICP-MS data requires a new generation of mass spectrometer data reduction software. The development of such FAIR ICP-MS data pipelines will not only benefit fission track analysis but other chronometers as well, such as in-situ U–Pb, Rb–Sr and Lu–Hf.

With the establishment of FAIR data, geochronology will be well placed to avoid the reproducibility problems that have plagued other fields of science.