Using community photography to investigate phenology: A case study of coat molt in the mountain goat (Oreamnos americanus) with missing data

Abstract

Participatory approaches, such as community photography, can engage the public in questions of societal and scientific interest while helping advance understanding of ecological patterns and processes. We combined data extracted from community-sourced, spatially explicit photographs with research findings from 2018 fieldwork in the Yukon, Canada, to evaluate winter coat molt patterns and phenology in mountain goats (Oreamnos americanus), a cold-adapted, alpine mammal. Leveraging the community science portals iNaturalist and CitSci, in less than a year we amassed a database of almost seven hundred unique photographs spanning some 4,500 km between latitudes 37.6°N and 61.1°N from 0 to 4,333 m elevation. Using statistical methods accounting for incomplete data, a common issue in community science datasets, we identified the effects of intrinsic (sex and presence of offspring) and broad environmental (latitude and elevation) factors on molt onset and rate and compared our findings with published data. Shedding occurred over a 3-month period between 29 May and 6 September. Effects of sex and offspring on the timing of molt were consistent between the community-sourced and our Yukon data and with findings on wild mountain goats at a long-term research site in west-central Alberta, Canada. Males molted first, followed by females without offspring (4.4 days later in the coarse-grained, geographically wide community science sample; 29.2 days later in our fine-grained Yukon sample) and lastly females with new kids (6.2; 21.2 days later, respectively). Shedding was later at higher elevations and faster at northern latitudes. Our findings establish a basis for employing community photography to examine broad-scale questions about the timing of ecological events, as well as sex differences in response to possible climate drivers. In addition, community photography can help inspire public participation in environmental and outdoor activities specifically with reference to iconic wildlife.

Keywords: citizen science; climate change; community science; elevation; latitude; molting; sex differences; ungulates.

Figure 1

Map showing mountain goat range (IUCN Red List shapefile 2008) and locations of photographs from community scientists (red), professional photographers (also considered community scientists but shown separately here, in yellow), and researchers (gray). A couple of the points outside the range as shown are from zoos (Woodland Park Zoo and Oregon Zoo); there are also photographs from several areas where goats were introduced (e.g., Mount Peale, Utah) that are not part of the available shapefile. The photograph shows an adult female in the Yukon Wildlife Preserve on 27 June 2018 with 12.17% of her winter coat shed (with molt still at an early stage, demonstrably starting at the head and proceeding down the neck)

Figure 2

(a) Locations of community‐sourced photographs. Cross indicates the location of the captive study population at the Yukon Wildlife Preserve (YWP). (b) Relation between latitude and elevation for all photographs. Colors show what is known about the sex and presence of a kid for each animal photographed. There are six possible animal states described by a pair of letters: [first pair] F = female, M = male, X = sex unknown; [second pair] Y = with kid, N = without kid, X = kid status unknown: FN = female without kid, FY = female with kid, FX = female and presence of kid unknown, MN = male, XN = unknown sex without kid, and XX = unknown sex and presence of kid unknown

Figure 3

Fraction of coat shed estimates from all photographs collected during the community science project. The number of photographs taken each year and the number where animal state is uncertain (brackets) are provided in the panels. See Figure 2 for explanation of animal state. Light gray circles depict all shedding estimates. Note that the unusual late‐year, low‐shed values were removed from the analysis. The model was only fit to years where there was at least one photograph where animal state was not ambiguous (i.e., bracketed value is greater than zero)

Figure 4

Observed and predicted shedding patterns. Panels correspond to the three animal states where sex and kid status are known. Shedding fractions for all photographs where animal state was known are presented (points). The predictions are for 2018 at location defined by the z‐transformed predictors being zero, which corresponds to latitude 49.14 and elevation 2,025 m (c.f. Figure 2b). The predictions also correspond to all random effect terms being set to zero. Solid lines depict the median shedding fraction, and shaded regions are the associated 89% credible intervals

Figure 5

Predicted dates for females without kid (state FN) having shed 50% of their coat when at a site defined by z‐transformed predictors being zero (i.e., latitude 49.14 and elevation 2,025 m). Dashed line is the long‐term trend, and shaded region is the 89% CI. Estimated yearly fluctuations about the trend are also presented along with their 89% CI

Figure 6

Observed shedding patterns for 14 captive animals repeatedly observed at the Yukon Wildlife Preserve in 2018. See Figure 2 for explanation of animal state. Shaded region depicts 89% CI for the mean fraction shed