Discussion
Our results demonstrate that community photography, including old and recent images, has potential for investigating phenological differences across wide spatial areas. In particular, we have shown that statistical analyses can be informative even when a significant fraction of the photographs are missing information (e.g., animal states), which is often inevitable with community science data. We were able to incorporate nearly all of the community science photographs that were associated with uncertainty into the statistical analysis by building missingness explicitly into the statistical model expectations. Conclusions drawn from the community science analysis are supported by similar conclusions drawn from the analysis of our Yukon Wildlife Preserve data that included complete data and repeated observations of individuals.
Our findings (using both crowd-sourced and field-sourced data) on sex differences and effects of new (current year’s) kids are also consistent with earlier published results from Caw Ridge, west-central Alberta (Déry et al., 2019) as well as with the recorded knowledge of a Tlingit weaver (Rofkar, 2014) who observed that nannies finish moulting only after weaning their kids. The stage of hair moult has also been used to make assumptions about sex and reproductive status in the field (Chadwick, 2002).
Community-sourcing of photos was generally successful; every naturalist with existing mountain goat photos already uploaded on iNaturalist granted us permission to have their photos included in our analysis and some people uploaded additional photographs after learning about our study. More than 100 people submitted photos via our designated CitSci platform. That we detected sex and kid effects using photos crowd-sourced over approximately one year demonstrates that community science can be leveraged to boost research scope, scale and speed especially if it can be validated against long-term and/or carefully field-sourced data.
While our sample had broad spatial coverage (Figure 1) directly inferring the effects of latitude and elevation on moult was difficult due to their negative correlation. Where mountain goats have access to a range of elevations (e.g., in the southern Rocky Mountains) there are presumably advantages as the climate warms as average temperature drops by approximately six degrees Celsius for every 1000-meter increase in elevation. However, as observed in cashmere goats (Capra hircus ) wool is also thicker at higher elevation, so presumably there is also more to shed if animals shift higher up, a pattern noted by British military officer Cecil Rawlins when above 4000m in Tibet in 1905 (Berger, 2018). We found later and slower rates of moult in mountain goats where they were photographed in high mountains / higher elevations, although we acknowledge this does not account for their vertical movements as photographs are mere snapshots of locations and not indicative of long-term responses or behaviour (Beever et al., 2017), for example shade-seeking or timing of activity.
We found later moult at higher latitudes in contrast to findings in muskoxen where earlier shedding occurred in animals at 69 degrees N than at 57 degrees N (Wilkinson, 1974). Whereas photoperiod is considered to be a primary cue for moult (Lincoln and Ebling, 1985; Zimova et al., 2018), temperature also plays a role; for example, cattle populations at the same latitude but different temperatures show earlier moult where lower and warmer (Murray, 1965). In northern mountain goats, delayed moult may be warranted if a warm coat is still needed for the late spring cold. Rate of moult was faster at the lower elevation, higher latitude sites, which might also be explained by northern mountain goats’ need to put winter coats on more quickly in time for winter.
We found no clear long-term climate signal, however, consistent with our hypothesis that moult may occur earlier in the year in recent years because of climate warming, we detected some evidence of earlier moult in 2015-2018 (Figure 5A). Our inability to tease apart long-term change with random but recent shifts to earlier shedding may be partly due to a limited sample of historical photographs. Specifically, more than 90% of our 650 photos were from the year 2000 and later. Further outreach and time would be required to meet the objectives of boosting sample sizes and reducing bias, especially for the acquisition of historical images. However, such data are necessary to enable a more rigorous assessment of the extent to which climate is driving shedding schedules. Analyzing shed extent from photographs by hand is slow, thus as data accumulate an important research focus will be how best to employ automated processing of photographs (e.g., Artificial Intelligence, machine learning).
One further caveat, and an indirect effect of climate change, was the prevalence of ticks (Dermacentor andersoni ) and their effect on moult patterns. We noted visible hair loss in mountain goats’ shoulder areas associated with active rubbing, especially at Glacier National Park (the locale of most crowd-sourced photographs) and where others, notably Douglas Chadwick and Sumio Harada observed these tick-affected bare patches (personal communication, 2018). As moult starts at the face, loss of hair and irritated skin at the shoulder area when the face and neck are not yet shed can most likely be attributed to ticks and not to moult onset. However, we did not attempt to distinguish tick-related loss of hair from regular moult. Our Yukon data offer a tick-free baseline for the far north; while ticks were first detected in the Yukon in the early 1990s and have also been found on ungulates in the NWT, these ticks are cervid-specific and it is highly unlikely that they have reached mountain goat range in the Yukon. Future moult research would still more carefully distinguish between regular moult and tick-affected hair loss resulting from rubbing.
Our future work will explore the potential to reduce analysis effort and time by linking the CitSci platform with Zooniverse, where community scientists engage with image classification tasks. From a community science theory perspective, such a linkage offers opportunities to study cross-over motivations between volunteers submitting photos and those classifying images. Given opportunity for additional field research, it would be valuable to compare the captive herds of the YWP with adjacent wild mountain goats (at the same latitude) to explore the roles played by nutrition and relative safety from predators in moulting.