The project

Fossil Atmospheres is a large climate science research project housed in the Department of Paleobiology within the Smithsonian Institution’s National Museum of Natural History (NMNH) in Washington DC, USA. Our research uses fossil and modern ginkgo leaves to develop baselines to understand ancient and future climate change (Figure 1; Barclay and Wing 2016). Specifically, we use the leaf stomatal index, a measure of the percentage of stomatal cells in relation to all cells on a leaf surface, as a climate proxy (Woodward 1987). Ginkgo plants are an excellent study subject as the Ginkgoaceae family has a 200-million-year plus fossil record. Today Ginkgoaceae is represented by a single surviving species, Ginkgo biloba L. (henceforth referred to as ginkgo). Ginkgo is a ubiquitous modern landscape tree that is broadly recognizable by the public due to its unique leaf shape and popularized for its purported medicinal uses.

The Fossil Atmospheres project is made up of three initiatives that each contribute a unique source of leaves from which we calculate stomatal indices. The first source of leaves comes from herbarium specimens and fossils dating back to 56 million years ago, which are used to establish historic baselines for the index. The second source consists of leaves from trees grown under experimental CO₂ conditions. The third source, which is the focus of this paper, consists of modern ginkgo leaves from across the full geographic range of ginkgo in the United States to determine how naturally varying the stomatal index is today.

To achieve this third initiative, our team, composed of paleobiologists, science educators, and museum volunteers, developed a hybrid citizen science protocol that asked people throughout the United States to complete two connected tasks: collect and submit physical ginkgo leaf samples, mailed to us at the NMNH, and collect tree images, and location and other site data reported using the iNaturalist platform. Upon receipt, all leaves were accessioned into the permanent Smithsonian collection. By launching a nationwide call for ginkgo leaf samples, we hoped to capture a snapshot of the species across a wide geographic range. As long-term repositories of physical samples and information, museums are an ideal venue for such projects. We anticipate future researchers will access the collection to answer new research questions beyond our investigations, for example those involving genomic sequencing, particularly as new technologies for analysis emerge.

Technology as a tool for collecting citizen science data

Citizen science projects regularly ask for data to be uploaded to a project-specific website or mobile application. Numerous citizen science projects have also been exploring how increasingly popular online platforms, such as eBird or iNaturalist, might support citizen science. Projects are using online platforms to collect robust scientific data (Sullivan et al. 2014) but also to engage, support, and include a community in scientific or policy decisions (Groom et al. 2019). One powerful affordance of online platforms can be that they provide both machine learning algorithms and a community of knowledgeable users that allow for quick and accurate confirmations of the identity of an organism (Unger et al. 2020). Accurate identification has been shown to be a successful tool for projects that are interested in mapping species habitat (La Sorte and Somveille 2020), and for documenting species richness within an area for either specific research projects (Wittmann et al. 2019) or general BioBlitz efforts (Parker et al. 2018).

A second affordance of online platforms is the standardization of data collection protocols, which mitigates a traditional concern within citizen science regarding varying data quality. Using an online form for data entry is one way to control for incomplete or missing data—recognizing that best practice encourages flexibility with data collection tools so that possible participants or locations are not excluded because of limited technology access (US GSA, n.d.). Established online platforms, such as iNaturalist, may also allow individual projects to incorporate customizable data fields beyond those found in a standard record. This flexibility allows timestamped documentation of ecological and environmental conditions associated with a particular sample, adding to the depth and robustness of data associated with it.

Associating digital data, citizen science, and natural history museums allows for additional benefits. Museum specimens can be used to further validate digital records data and vice versa (Spellman and Mulder 2016), providing a strong case for the use of museum and online records in combination, as we have done in the Fossil Atmospheres project. Combining museum collections with iNaturalist records makes best use of the strengths of each of these repositories to result in a long-term physical asset with rich associated date- and location-specific metadata.

Choose

We began by choosing an online platform that would serve as the best digital tool for both our project and our sampling period.

Design for universal usability

We needed a protocol that facilitated easy and accurate collection of both physical samples and associated site data that could subsequently be paired by the research team. Because we expected citizen participation to be self-guided, or perhaps facilitated by staff at botanical gardens and herbaria but never directed by project researchers, we focused on designing a collection protocol that employed materials and platforms that were widely available and user-friendly. In this way, we hoped to lower the barrier to participation while acquiring quality specimens and associated digital data.

Developing a clear, concise sampling protocol

Our sampling protocol involved two tasks: 1) collecting observational data at the tree, recorded digitally and 2) collecting a sample of leaves at the tree to mail to the NMNH. The complete hybrid protocol is included in the supplemental materials (Appendix B). To develop the collection protocol, we focused on identifying single steps and data fields that would allow us to extract multiple pieces of information. For example, team scientists were interested in aspects of the tree’s habitat and specific sampling details. Rather than asking participants to record that information, we instead asked for photos of the base of the tree. The research team could later use these photos to determine that information. In this way, our protocol shifted the data processing from citizen scientists to our team as much as possible.

When a participant submitted their observation to the Fossil Atmospheres iNaturalist project, they included two photos, the geo-tagged location and date of the observation, a positive identification of the pictured tree as ginkgo, and three custom data fields specific to our project. All custom fields provided options from a drop-down menu, ensuring data standardization (Appendix C). We first asked participants to estimate the height of the tree from two defined ranges (10–30 feet or 30+ feet). We then asked that participants determine the sex of the tree (male or female). Because iNaturalist requires only one photograph for an observation to be submitted to their platform, our third field prompted participants to confirm that they had uploaded two photos to meet protocol requirements: of the tree and of the base. Finally, we asked participants to determine, if possible, which side of the tree the leaves were collected from (north, south, east, west, or unknown). This was not required, as we did not want to preclude or discourage any participants who may not have had the technology or the skills to easily collect this information.

The protocol for submitting the leaf samples was designed to minimize the risk of physical damage during shipping or biological damage from factors such as mold if mailed packages were not received in a timely manner. We also wanted a package that could be mailed cheaply given the United States Postal Service’s parcel size thresholds. We focused on using packaging materials that were easily accessible around a home or office, eventually deciding on cardboard, tape, and newspaper. The shipment protocol was iteratively tested and refined by mailing leaves to ourselves at the intended destination address at the NMNH. We found that making a “cardboard sandwich” (Figure 3), with the sample leaves surrounded first by newspaper and then by cardboard, all of which was secured by tape and then placed in an envelope for mailing, was the minimal amount of effort required to adequately protect the leaves. International citizen scientists were asked to dry specimens thoroughly before shipment and provide extra documentation to comply with government requirements. All samples, including those sent internationally, were received in good condition.

Figure 3 

An excerpt of the protocol instructions detailing how to create the cardboard sandwich used to ensure intact delivery of leaf samples to the Smithsonian. Full protocol instructions are available in the supplemental materials (Appendix B).

To ensure an adequate amount of sampling, participants were asked to collect at least six leaves from a single shoot on a tree. Critically, each sample of six leaves was to be contained within its own cardboard sandwich with the exterior cardboard layer clearly marked with the participant’s iNaturalist username, date, and time of observation. This information allowed the Fossil Atmospheres team to identify which unique iNaturalist observation the leaves were associated with and precluded the need to collect any additional personal information not already publicly available on iNaturalist’s website.

Engage

Our promotion efforts began at the start of our August collection period and continued for approximately three weeks. We worked to identify potential partners that could help us communicate the project to broad audiences and engaged consistently across social media. We found particular success through Facebook, posting our call-to-action video on the platform on August 2nd and then re-engaging through that video and other content every 5–7 days throughout the month. Fossil Atmospheres also had active social media accounts on Twitter and Instagram. One member of the research team was assigned to each social media account and worked to be highly engaged and responsive on their assigned platform throughout the month. We also had a designated email account for the project through which we responded to requests for clarification or other help from more than 90 people.

Pairing physical samples with their online record

We received 562 physical samples and 608 digital observations. Multiple observations were made in the iNaturalist Fossil Atmospheres project for which no physical paired sample was received. We also received physical samples that did not have an associated digital record. Of the 562 physical samples received, 367 (60%) were easily paired with project-linked observations on iNaturalist. This left 195 physical samples and 241 digital records that were ambiguous. After additional investigation by the research team 65 samples (12%) were found to be unpairable or paired but missing key scientific data. We concluded with 496 samples (88% of all physical samples received) that met our scientific requirements and were able to be included in the Fossil Atmospheres sample set (Figure 5).

Data integrity

The leaves of ginkgo have a unique and highly recognizable morphology, making it easy for both citizen scientists and iNaturalist’s computer vision algorithms to correctly identify them. Our project benefited from this; 100% of the digital and physical samples we received were correctly identified as ginkgo. Only one sample, composed of leaves from the wrong part of the branch (see Appendix B), did not meet project requirements. Participants were very successful at following the protocol requirements for packaging and sending the physical sample. Of the physical samples received, 558 (99%) had complete or traceable information included on the cardboard sandwich (Table 1). In the instances where the information was incomplete or different from the iNaturalist observation, the name, return address, or postmark from the outer packaging could often be used to facilitate pairing the physical sample to an online observation. If more clarification was needed, we messaged users directly through the iNaturalist platform.

Using an established online platform as a data collection tool gave a high level of confidence for digital data. Drop-down data fields requesting the height and sex of the tree were required before the iNaturalist observation could be submitted, and so were uniformly complete. Location and time information were automatically generated by the platform and so these fields were also uniformly complete.

Participants were more challenged by digital data entry than by the leaf shipping procedure. Many participants created multiple digital observations for the same physical sample: 46 records of the 241 ambiguous iNaturalist records were in effect duplications that could be collapsed into one observation and then included in the project. A more serious, but possibly correctable, challenge with the digital data involved citizen scientists entering their tree data on iNaturalist without linking that observation to the Fossil Atmospheres project. Because a standard iNaturalist observation requires only species identification, photo, location, and time of observation, entries that were not linked to our project did not contain responses to our required project fields. It was often possible to find these unlinked observations through a direct search of iNaturalist using either the project-specific data they included on their cardboard sandwich packaging or by using information from their outer mail packaging. As administrators of the iNaturalist project, the research team was then able to link these observations to the project digitally. However, that did not resolve the missing data. We could subsequently reach out to specific participants using iNaturalist’s direct messaging feature to prompt them to provide the missing tree data. These additional efforts allowed us to successfully pair another 130 physical samples to their associated digital data, increasing the usable sample set from 367, which were perfect upon arrival, to 497 that were finally accepted by the project (Table 1).

On the rare occasions when participants mailed us a sample but did not generate any iNaturalist observation, there was no option for correction. In these cases, we knew the participant had read and followed the protocol, as evidenced by a correctly collected and packaged sample. Occasionally some or all of the required digital data were written on the cardboard sandwich, but more often, these samples presented without any paired tree data and we were therefore not able to include them in the data set.

Designing a hybrid protocol

We realized success at all three phases of project development and implementation (Figure 2). We present these actions as a template that other projects may follow.

Limitations of online platforms

Online platforms are extremely useful for organizing the collection of data but have profound limitations for long-term data storage. Although it would be tempting to consider a professionally maintained network like iNaturalist as a permanent repository for data, such a practice would not be considered good data management. Internet-based databases may have a lifespan and are accessible to outside users for only as long as the sites are functional online. Databases cannot be guaranteed to be backed up, engulfed, or accessible in perpetuity. Online platforms for long-term data storage also frequently lack database versioning that allows previous database iterations to be referenced later. Though download dates and version numbers are standard to report, there is no guarantee a reported dataset would be reproducible, as iNaturalist and other online databases are constantly under development and growing.

Lastly, how and what data is accessible is dictated by the online platform, which may lead to unanticipated restrictions. For example, iNaturalist limits file download size and suggests that users utilize the Global Biodiversity Information Facility (GBIF 2021) database for larger downloads. This is reasonable, but there are caveats. iNaturalist uploads only “research-grade” observations to the GBIF website. Our observations of Ginkgo biloba are considered by iNaturalist to be a “casual” observation, because effectively all specimens of current ginkgo were planted. Despite the fact that Fossil Atmospheres is collecting these casual observations for research purposes and that we have acquisitioned the physical specimens into our collections at the NMNH, Ginkgo biloba records are not included in databases like GBIF, which exists to document natural diversity patterns. For these reasons, we highly recommend that researchers create their own permanent and versioned copy of data when utilizing online platforms for citizen science, as we have for these collections, which are permanently housed at the NMNH.

Limitations of the protocol

A primary limitation of our hybrid protocol involved universal usability. We took care to design alternative paths for participation that did not require a smartphone, but at minimum, participation did require a camera, a computer, and internet access. We had 24 participants send us physical samples that were not associated with iNaturalist in any way. Some participants wrote some, or even all, of the required digital data on the cardboard sandwich, but without photos and a confirmed location, these samples could not be included in the Fossil Atmospheres data set. We can’t know why these participants chose to eschew the online platform tool. At this time, it is also not possible to know how many people were discouraged from participating by the technological requirements.

Our hybrid protocol also did not allow us to collect information on who the citizen scientists were or exactly how they interacted with the protocol. For example, we were unable to determine which form of the protocol guide participants found most useful or if they were using the app or the website to enter their digital data. This information could be collected by future projects by including unique data fields that asked demographic questions or queried the user experience. Further analysis, which is outside the scope of this paper, could also give some indication of whether participation in the project impacted people’s use of the online platform—for example, how many participants were regular or first-time users of iNaturalist and whether, when a ginkgo observation was their first action on the platform, they stayed and became regular users.