Introduction

Many butterfly species are in decline or at-risk of extinction around the world. At the same time, the climate is changing. One of the most documented “fingerprints” of climate change is advance in phenology, that is – changes in the seasonal timing of biological events. A concern for many in the conservation community is that changes in phenology may be causing further stress for already at-risk species. Our project was focused on asking if trends in phenology (e.g., changes in butterfly flight seasons) are associated with trends in population status (i.e., how “at-risk” a butterfly species is). By focusing on at-risk butterfly species, we can make inference about the degree to which climate-change driven changes in phenology are a primary factor of concern for these species, or if other factors (habitat loss, habitat degradation, invasive species, etc.) may be a higher priority for managers and wildlife biologists to address when planning conservation efforts for these species.

Methods and Definitions

To assess the relationship between phenology and abundance for at-risk butterflies, we first needed to identify which butterfly species to include. With help from Candace Fallon at the Xerces Society, we designated a butterfly species as “at-risk” if it is listed as threatened or endangered at the federal or state levels (see individual state wildlife agencies), if it is considered a Species of Greatest Conservation Need (SGCN) within a state or the Nature Serve Rank is S3 or higher at-risk. The at-risk species were those listed in the described databases in 2018.

Next, we compiled and curated long-term data from as many of these species as we could find sufficient datasets. We defined “sufficient” as datasets with at least ten years of surveys in which surveys included each at-risk butterfly species for multiple surveys each season. This broadly includes two types of datasets – 1) butterfly community datasets which include habitat for at-risk butterflies, and 2) single species datasets for which the monitoring programs were established specifically to monitor focal butterfly species. Because butterflies are easy to see and count, there are multiple programs in which community scientists participate in systematic programs to count and identify all butterflies in a community. Many of these use a “Pollard Walk” technique, in which observers walk a known route each week and identify and count all butterflies within a visual range (often defined as 5 m on either side of the observer and 5 m in front). This technique has been used for decades in the UK as part of the United Kingdom Butterfly Monitoring Scheme.

Three large US programs use Pollard Walk methodology at multiple sites within a state. These include Art Shapiro’s butterfly monitoring at ten sites in a transect across California, the Ohio Lepidopterists’ Butterfly Monitoring and the Illinois Butterfly Monitoring Network. The latter two are networks of sites monitored by community scientists with volunteers curating at the state-level. For these three datasets, we identified all species listed as “at -risk” in these states for which there were sufficient data in the datasets for use in our analysis. Second, we contacted agencies and non-profit organizations (e.g. The Nature Conservancy, Xerces Society, and others) which survey federally listed at-risk species. When there were documented at-risk species on Department of Defense land, we made additional efforts to access butterfly survey data if surveys had been systematically collected. In addition, in a few cases, we learned of community-type surveys that were started at the same time as starting surveys for focal at-risk species. We included these community-surveys in our datasets if they included additional at-risk species.

We excluded two types of datasets from this analysis. First, we excluded sub-tropical regions (e.g., southern Florida) because the butterfly population dynamics are known to be aseasonal and are not expected to track changes in climate as well as excluding species that are outside the continental US. Second, we excluded monarch butterflies because this species is treated in-depth in other parts of this project and because its migratory biology is distinct from the others in the dataset; all other at-risk species in the dataset are resident species within each focal state. In addition, years in which captive-reared individuals were added to a site, or sites were created by de novo restoration, were excluded from the analyses because the releases and colonization of habitat may be at different times from natural eclosion in the wild or have trends driven by the restoration efforts.

After compiling datasets for at-risk butterfly species, our next step was to collect management information for each site at which we had data for at-risk butterflies. For many species, these data are not systematically gathered. We used a combination of published documents, reports, and expert interviews to compile the management data. For each species, we reached out to staff with federal and state agencies, land managers, and volunteers who have a history of collecting butterfly data at the sites in the dataset. If there were no written reports, staff and volunteers were contacted to conduct phone interviews. For each species-site-year combination, first we asked if the site is actively managed for butterfly habitat or not. If the site is actively managed, then we asked for details on the management actions (mow, fire, graze, herbicide-based management, planting of host or nectar plants, or general manual methods to reduce weeds). Collection of these data was at a coarse scale. For example, we asked whether fire was used each year – not about the area or intensity of burns. Thus, some sites were actively burned yearly but only a portion of the site is burned in any prescribed fire event– and the finer details that include which parts of a site were burned in a year are not included. This allowed us to collect data on the frequency with which a management technique was being used but not the frequency with which a given patch of land was burned.

To estimate trends in phenology abundance (i.e. butterfly activity index), we fit separate smoothing splines for each site for each species through time. Butterfly activity index is a relative measure of the number of observations for a butterfly at a given site but is not necessarily indicative of the absolute population sizes. Phenological shifts were measured as change in median activity date. We used these analyses to look at associations between trends in phenology and trends in abundance. Finally, we assessed if trends in abundance were associated with site-level management. More information on technical details of the analysis can be found in the full report (link to report will be available Summer 2022).

Results

Overall, these populations were declining, but there was no systematic trend in phenology. Moreover, for the at-risk butterfly species that we assessed, we found no correlations between phenological shifts and changes in abundance over time including five butterfly families in 10 states.

However, we found strong associations between trends in abundance and habitat management. In general, increases in butterfly activity were associated with active management of the sites. The most frequently selected type of management was mowing management. Over 60% of sites had mowing at least once during the period for which we have butterfly monitoring data, 45% used fire, and 40% applied herbicides to control invasive species. At 40% of the sites, host or nectar plants were planted/seeded during the survey period. The least selected management option of the ones we asked about was conservation grazing. In this case, fewer than 10% of the sites used conservation grazing at any time during the period during which we had survey data.

We do not have analyses by spatial location to date, general locations for sites are in our map below.

Discussion

Several studies across all butterfly species in the US have documented advances in phenology, often accompanied by increases in population size. In contrast, we see no similar pattern in our dataset for at-risk butterfly species. Here, our strong associations between abundance and presence of habitat management indicates that conservation efforts to improve conservation outcomes are working. Managers should be encouraged to continue developing site and species-specific management actions which promote butterfly populations. Our results suggest that active habitat management promotes climate resilience in an era of global climate change and shifting phenology.