Data Collection
Core unit associations were recorded during behavioural follows conducted over 21 months between August 28th, 2017 and May 13th, 2019 (243 days) by two trained field assistants (E. Mujjuzi and H. Kakeeto). From August 28th, 2017 to Aug. 22nd, 2018, 12 core units were sampled and from Aug. 29th, 2018 to May 13th, 2019, 13 units were sampled because it became obvious that an all-male unit (AMU) had formed by the splitting off of seven adult males from the largest core unit (Lovoa), which subsequently became an OMU that was markedly less cohesive than the other units. We deemed 21 continuous months of data as sufficient to answer our questions because four rainy seasons were covered and clans had sufficient time to change in core unit membership. One focal unit was followed each day between 8:00 h -16:00 h and all individuals were identified based on physical characteristics (e.g., broken fingers, tail shape, nipple colouration). Scan samples on core unit association were taken every two-hours, where the number and identity of core units within a 50 m radius of the focal core unit was recorded along with the time and date (overall N = 907 scans). We chose a two-hour interval between scans to help ensure their independence. We reasoned that two hours was enough time for core units to shift their position relative to one another (Stead & Teichroeb, 2019). Our data collection regime led to a relatively even distribution of focal days among core units during the study (mean N days/unit = 20.17, range: 14-25; mean N scans/unit =75.25, range: 56-90). Dispersals of individuals within the study band were recorded on notice of occurrence and a date range during which the dispersal occurred was generated based on the last time an individual was noted in their original core unit. The month of dispersal was determined to be the month with the most potential dates within that range.
To examine the seasonality of association patterns, we considered three ecological variables: rainfall, the availability of young leaves, and the availability of fruits. Rainfall data (mm per month) was obtained from https://www.worldweatheronline.com/masaka-weather-history/masaka/ug.aspx for the nearby town of Masaka (12.5 km away). We considered the availability of young leaves and fruits as these food items comprise the majority of the C. a. ruwenzorii diet at this field site (96%, JAT, unpubl. data). Food availability indices were calculated for each of these plant parts, for each month of the study period. We used a line-transect survey to estimate tree species abundance (i.e., number of trees and their basal area) within the home range of the C. a. ruwenzorii band. Thirty-two parallel transects set 100 m apart were cut throughout a 140 ha section of the forest and all trees >10 cm DBH within 5 m of either side of the transect were identified and measured (covering 9.702 ha) (Teichroeb et al., 2019). The seasonal availability of these plant parts was estimated using monthly phenology surveys of 126 trees of 44 species that were known to be consumed byC. a. ruwenzorii. During phenology surveys, observers indexed the percent canopy cover of mature vs. young leaves, ripe and unripe fruit, ripe and unripe seed pods, and buds vs. flowers in at least three sample trees of each species. We calculated the food availability index for both young leaves and fruits separately by multiplying the mean monthly phenology score for each plant part in each of the 44 species by the total basal area of that species, and summing these values for all the tree species consumed (Dasilva, 1994; Fashing, 2001; Saj & Sicotte, 2007).