Developing a Historical Phenology Dataset through Community Involvement for Climate Change Research

There has been an increased effort to utilize untapped sources of historical phenological data such as museum specimens, journals written by nature-minded citizens, and dated photographs through local phenology projects for climate change research. Local phenology projects have contributed greatly to our current understanding of phenological changes over time and have an important role in the public’s engagement with natural history, but there are also significant challenges in finding and analyzing historical data. The West Virginia Climate History Project collected historical phenology data (1890-2015) from citizens in West Virginia with scientific and cultural results. We discuss the development of the project, issues we overcame, recommendations for future projects, and the conservation value of local phenology projects. Local phenology projects lend value to climate change research and conservation education. Policy directed towards supporting humanities grants and museum collections, including herbariums, is crucial in the success of these projects.


Introduction
Phenology, the timing of biological events in relation to seasons, has become popular due to its public accessibility [1], potential for scientific education [2], and use in tracking responses to climate change [3]. With the rise of phenology, there is a need to uncover historic phenology data so researchers can track changes How to cite this paper: Petrauski, L., Owen, S., Constantz, G. and Anderson, J.T. (2020) Developing a Historical Phenology Dataset through Community Involvement for Climate Change Research. American Journal of Climate Change, 9, 11-22. over time. There has been a surge in attempts by researchers to utilize untapped sources of historical phenological data such as museum specimens, journals written by nature-minded citizens, and dated site-specific photographs [4] [5] [6].
Historical phenological observations have been used to create baselines for bird migration [7] [8] first flowering dates [9] [10] and leaf-out dates [11] [12]. Researchers have used phenological data found in journals of naturalist-minded authors as a baseline to compare with current dates of phenological events to estimate how climate change is affecting the timing of phenophases [12] [13]. These isolated studies, which involve local communities in research on familiar flora and fauna, are important for understanding individual organisms' responses to changes in climate.
Currently, the USA National Phenology Network (USA-NPN) attempts to synthesize data from individuals and small phenology projects, increase internetwork communication, and promote the practical use of phenological data [14] [15], as well as increase public awareness of phenology, climate change, and the scientific process [16] [17]. However, the implementation of these databases is relatively recent (USA-NPN was created in 2007) and establishing baseline data is still a challenge for phenological research [18] [19] [20]. Researchers have had to use creative methods to gather long-term baseline data, including the use of herbarium specimens, observations from diaries, and photographs [21].
There is a need for long-term phenological datasets in West Virginia to assess the current and future effects of climate change, as well as to set the stage for future citizen science projects. These are the goals of the West Virginia Climate History Project (WVCHP). Previous long-term studies have been done using historical observations from well-known people such as Henry David Thoreau [21] or Aldo Leopold [13], but this project demonstrates that reliable, long-term datasets can be found in any area with observant citizens. The purposes of this paper are to outline the WVCHP, advise other local phenological projects, and discuss the conservation value of local phenology projects.

Uncovering and Processing Historical Sources
The WVCHP began with public outreach in January 2015, in the hopes of finding individuals with archival sources of phenology data. Through presentations with local naturalist groups, informational fliers sent to public libraries, a press release through the university, an online project blog, and a West Virginia Public Radio interview, we attempted to reach interested citizens and find leads on potential data. With each outreach effort, we were offered suggestions, email addresses, and names of other people to contact for assistance with the project.
The most significant breakthrough in finding phenological data was the start of our relationship with the Brooks Bird Club, a naturalist group in West Virginia that was founded in 1932. The club's official publication, The Redstart, was the single most data-rich source used during this project. Our communications with their current members led us to other individuals with excellent historical records.
Relationships we forged with local naturalist groups contributed to the success of this project more than any other factor; 78% (n = 5614 observations) of the data gathered through this project were directly related to these relationships.
Once the data sources were located, the task of transcribing the records began.
Transcribing data from handwritten notes, journals, and lists (24% of the data, n = 1350 observations) required patience and investigative skills; whereas capturing data from newsletters, lists, or herbarium-labels was more efficient. The species, location, date, and phenophase were recorded for each phenological observation into an electronic format. One issue we encountered when working with these sources was the question of the difference between a phenological and a biological observation. For example, an entry in a personal journal could mention an American black bear (Ursus americanus) sighting in West Virginia, but the observation is not associated with seasonality. On the other hand, a journal entry that notes a bear with a young cub would have merit as a breeding phenology observation. We recorded phenophases and their definitions based on guidelines set by the USA-NPN [22].
At the start of our project, we had no preferences for specific species, phenophase, or season. After we discovered the first several sources, we saw the most potential in finding data regarding avian spring migration and spring wildflower blooming. Our interest in avian phenology was inspired by the Brooks Bird Club and the data potential of herbarium specimens triggered our interest in wildflower phenology. Advantages of using herbarium specimens for phenological studies include long-term temporal coverage, broad geographic area, and availability of valuable data plus reliable metadata [23]. Most of our search focused on where we could find first arrival dates of migrating birds through West Virginia and flowering dates of spring ephemeral wildflowers, which describes much of the data collected. We collected phenological data from personal journals, dated photographs, personal birding lists, herbarium specimens, Brooks Bird Club records, and center collections (Table 1) as long as they were accompanied by sufficient metadata.
Once we began analyzing specific phenophases and species, we had to overcome the issues associated with combining datasets from multiple observers. Observer skill, time taken each week for observation, and distance covered during observation could differ among observers and bias the data [21]. Phenology data are especially prone to this issue due to the sensitivity to timing of these observations. For both analyses, avian first arrival dates and wildflower blooming dates, the data sources were evaluated for credibility and accuracy before combining the datasets. Specifically for first arrival dates, more observers may result in seemingly earlier arrival dates [24] [25]. We reconciled our avian sources with these issues by arguing that all observers were experienced naturalists who visited a variety of habitats to observe migrating species and that the data from each source were gathered by groups of people more often than by single individuals, which improved consistency among sources. Similar avian datasets from multiple observers and sources have been analyzed with defensible results [5] [8]. Possible issues with the wildflower blooming analyses were reduced by choosing species with a short blooming-window of one to two weeks, which negated problems associated with "first" phenophase analyses [26]. The main concern we faced with analyzing blooming dates was identifying locations of each blooming observation to calculate elevation and local climatic conditions. The location information offered on herbarium labels was used to pinpoint blooming locations with ArcGIS Version 10.3, but the level of detail varied for each location description. This variation in detail increased the likelihood that factors such as hillshade and aspect, which have been shown to explain variation in flowering phenology [27], could be confounding variables. Still, the high number of flowering observations analyzed was sufficient to overcome sampling bias within the data [28]. Overall, issues with combining historical datasets should be evaluated on a case-to-case basis that depends on the phenophase and species being analyzed.

Conservation Value
Local phenology projects have contributed greatly to understanding phenological changes over time [5] [8] [29] and have an important role in the public's engagement with natural history [6]. Local phenology projects are especially relevant in the context of climate change [30], due to their contribution to recovering long-term phenological data [31] and the connection between phenology and climate change [32]. Because recording phenological data is relatively straight-forward, phenology attracts a wide range of people with varying scientific experience, and therefore can impact how people understand climate change [1]. Local phenology projects can take advantage of this accessibility and offer a rare connection between familiar organisms and a global context.
The data collected by local phenology projects can be scientifically and culturally significant. Education and engagement of the public with climate change issues remain key components in the future success of conservation efforts [33].
Additionally, historical insights into species loss, changes in plant communities, the effects of land-use change, and urbanization are unique to each locality and can be used as a public reflection of human impacts on familiar places [34]. Specifically, the WVCHP encompassed a large, topographically variable area (62,259 km 2 , covering four ecoregions, average elevation = 420 m, elevation range = 73 -1482 m) across a great temporal extent (1890-2015). Such scope is a common advantage of citizen science projects and makes this type of research important for landscape-level conservation [35]. This project enabled us to collect data on private lands, which pose challenges for traditional sampling methods [36].
Long-term phenological datasets can be used to explore how relationships between native and invasive species phenologies may influence management practices [37], such as the scheduling of prescribed burning [38] and herbicide application [39]. In these cases, understanding the phenologies of native and invasive plants have facilitated improved management practices to protect native species and prevent the spread of invasives. Local historical phenology datasets can be used to understand the makeup of past plant communities [21], predict future competitive advantages of native and invasive plants [12], and give valuable perspectives on the locale of invasive species [40] [41].
Many studies linking phenology to climate change focus on sweeping conclusions that do not specifically address local management actions [42], but we argue that local phenology projects can be used to impact management, especially if it is prioritized at the beginning of the project. For example, determining historical patterns of phenology to predict future changes in migration or breeding of game birds could help optimize hunting seasons [43]. Similarly, exploring the historical relationships among temperature, aquatic insect emergence, and fish American Journal of Climate Change spawning could contribute to revising fishing regulations and identifying potential threats to aquatic ecosystems [44]. Climate-driven phenological mismatches could affect local populations including those species that are already threatened [45], such as the Cerulean Warbler (Setophaga cerulea) in the Central Appalachians that relies on mature forests for breeding. The ability to predict the vulnerability to climate change of certain species is a crucial first step for adjusting local management actions.
For example, the West Virginia white butterfly (Pieris virginiensis), a rare butterfly found in the Central Appalachians, has been declining in part because of the spread of exotic garlic mustard (Alliaria petiolata) throughout their range. A. petiolata is in the same family (Brassicaceae) as P. virginiensis's host plants, and P. virginiensis lays its eggs on A. petiolata where the larva cannot survive therefore negatively impacting butterfly reproduction. Understanding the phenology patterns of both A. petiolata and P. virginiensis would allow for better local management of both the threatened butterfly and the invasive plant. For example, A. petiolata control should be implemented in early spring before the emergence of P. virginiensis.
Butterflies have shifted their phenologies three times more than spring plants in recent years [46]. Butterfly species shift their emergence phenology relative to temperature [47], but their response to changes in climate is highly dependent on their larval host species [48]. Understanding not only flowering patterns, but also butterfly phenology in relation to climate change is important for pollinator conservation [49]. For example, P. virginiensis depends on early flowering species for food and oviposition. The early-flowering Cardamine diphylla, is a common host plant of P. virginiensis. The early senescence of C. diphylla has resulted in decreased larval survival or reproductive success of P. virginiensis. A local phenology project investigating this host plant and butterfly mismatch could increase the success of managing a threatened species.

Evaluation and Recommendations for Future Projects
Overall, the WVCHP collected 7181 historical phenology observations covering 404 different species, including 258 bird species, 37 species of trees and shrubs, 100 species of herbs and forbs, 2 insect species, 1 mammal, 3 amphibian species, and 3 grasses. Most of the observations (80.2%) were used to analyze avian spring migration and wildflower blooming. The remaining 19.8% covers a range of species and phenophases that we were unable to analyze due to too few consistent observations.
The success of this project in finding useful avian and wildflower phenology data was due to our concentrated focus on these two categories. This highlights the potential for more phenological data to be uncovered in this region, if a focus is set on particular species and phenophases. We expect that there are more useable data in this region that we have not had the opportunity to compile.
Phenological information about spring insect emergence would contribute to our L. Petrauski et al. American Journal of Climate Change understanding of ecosystem phenology and the potential risks of mismatch, because insects are a link between avian and plant phenologies [6], both of which have been studied more thoroughly than insect phenology. One of the greatest potential sources of phenological data is herbarium specimens, which have been highlighted in recent literature [11] [50] [51] and confirmed with this study. Herbariums are rich sources of phenological information that can be used to evaluate species presence and geographic distributions [23]. There is a disconnected relationship between the usefulness of herbarium specimens for research and the amount of funding given to enhancing these collections. There has been a decrease in funding for herbariums over the last few decades [52] despite an increase in using herbarium specimens as data sources [53]. Digitizing herbarium specimens has increased the usefulness and availability of these data and has contributed to the increased use of herbarium specimens in research. Unfortunately, small collections are seldom cataloged due to tight budgets, which makes them difficult to access [54]. Another issue facing herbariums is a decrease in addition of new plant specimens [55] [56], which could be contributing to the antiquated reputation of herbariums and their subsequent decrease in funding. The herbariums used in this study were strewn throughout West Virginia in various universities with no central herbarium database or consistent organization. Using herbarium specimens from these under-utilized and hard-to-access collections increased the distinctiveness and temporal range of our dataset. This demonstrates the value of small herbariums and the importance of continued funding for maintaining and growing herbariums and museum collections.
Humanities grants have also experienced decreases in funding, but we argue that these grants play an important role in local phenology projects and ensuing climate change research. Historical personal journals were a key source of phenology data in this study, and humanities grants allow for libraries to preserve these irreplaceable data points. The handwritten journals we uncovered carried both cultural and scientific significance that enriched the breadth of our project. This highlights the importance of funding for humanities grants.
We did not encounter a shortage of historical data, but finding current data in handwritten journals and herbariums was more difficult. To create a complete temporal dataset, we used alternative data sources to augment our current phenological observations. For our avian analysis, we supplemented current data with observations from eBird, an online community of birders [57]. For the wildflower analysis, we used dated photographs from a West Virginian citizen to increase the number of current blooming observations. The increased use of electronic phenological monitoring programs, such as Project BudBurst [58] and Nature's Notebook, are instrumental sources of current data that should be integrated with historical phenological data.
Local phenology projects are categorized as citizen science [59], but unless there is substantial effort to include educational outcomes and create educational materials, there is little evidence that the project will have educational impacts American Journal of Climate Change