Shorebird populations have declined due to several threats throughout their annual cycle. Anthropogenic disturbance is one of the most ubiquitous threats to shorebird conservation in North America. Here, we studied the influence of human disturbance on shorebird community dynamics during migration and winter in Ensenada de La Paz, a subtropical coastal wetland in Mexico. We used negative binomial generalized linear mixed models to investigate the associations between spatial, biological, and anthropogenic variation and local shorebird abundance that accounted for shorebird body size (small, medium, and large) and foraging strategy (visual and tactile) of 21 shorebird species. After controlling for these different correlates of abundance, human disturbance (people, vehicles, and dogs) was negatively associated with shorebird abundance. During winter, all shorebird species were negatively related to human disturbance but positively associated with presence of raptors. However, small, tactile foraging birds exhibited a proportionally larger negative response to human disturbance than other shorebird types, indicative of guild-level sensitivities to human disturbance regimes. The positive association between shorebird abundance and disturbance from predators was unexpected. Shorebirds likely concentrate in large groups to reduce predation risk, resulting in higher densities of shorebirds occurring in areas with high predation risk. Understanding factors influencing the abundance and habitat use of shorebirds on their non-breeding grounds is paramount to support management and conservation policies for shorebirds and their habitats.

We conducted an observational cross-sectional study with repeated surveys to determine effects of potential raptor and human disturbance on abundance of shorebirds. Area search surveys at 12 sampling units followed a monitoring protocol described in detail by Reiter et al., (2020), including documentation of environmental and habitat conditions in each sampling unit during each survey. The area flooded, vegetated, and bare ground was visually estimated at each sampling unit during field surveys. In addition, number of potential human disturbances (e.g., number of humans and number of dogs) and avian predators were recorded as they occurred during the survey period at each sampling unit (see details below). Twelve sampling units were surveyed twice a month, during neap rising tides, from September 2014 to April 2015; each sampling unit was visited 15 times. Sampling units were fixed areas of homogeneous habitat delimited by a polygon and included a mosaic of natural and human-made habitats within Ensenada de La Paz. Intertidal mudflat was the dominant habitat type among sampling units and are mud-sandy areas exposed during low tides, with less than 10% vegetation cover, and was found in Marina Sur (6.1 ha), Zacatal (30.6 ha), Chametla 1 and 2 (7.8 and 23.2 ha, respectively), Centenario 1 and 2 (22.7 and 9.2 ha, respectively), and Granjas Centenario (27.8 ha) sampling units. In addition, sandy beach habitat occurred at units Comitán 1, 2, and 3 (8.7, 9.3, and 12.6 ha, respectively). Human-made habitats included an abandoned shrimp farm (Granja Abandonada, 10.2 ha) and sewage ponds in the wastewater treatment plant of La Paz (Lagunas de Oxidación, 9.9 ha).

During each survey, a team of one observer and one annotator counted all shorebirds using the sampling unit, including shorebirds arriving or leaving the sampling unit, but not those flying over the sampling unit. Small flocks (˂300 birds) were counted, and we used the “blocks” method (Howes & Bakewell, 1989) to estimate numbers in flocks with ˃300 birds. Distance from the observation point to the birds in each count was generally less than 200 m. We sought a quick survey to assess bird estimates and behavior, so we tried to limit the time at each sampling unit while counting all birds accurately. Across all surveys, time spent at each sampling unit was between 8 and 44 min (mean = 26 min, SD = 20); duration depended on species richness, number of individuals, and size of the sampling unit. Long-billed (Limnodromus scolopaceus) and Short-billed (L. griseus) dowitchers were grouped as dowitchers (Limnodromus spp.) because their identification in the field was difficult. All surveys were conducted during daylight hours, in the same tidal condition (neap rising tides), when wind was < 25 km/hr, and when there was no rain to control for any effects on bird abundance behavior.

We counted avian predators during surveys and considered them as the baseline of potential natural disturbance as they can disrupt normal activities of shorebirds. Avian predators included Peregrine Falcons (Falco peregrinus), Cooper’s Hawks (Accipiter cooperi), Crested Caracaras (Caracara cheriway), Zone-tailed Hawks (Buteo albonotatus), Merlins (Falco columbarius), and American Kestrels (Falco sparverius). In addition, avian predators in, perched adjacent to, or flying over the sampling unit were counted during each survey.

We considered a potential cause of human disturbance to be “any human-induced activity that likely constitutes a sufficient stimulus to disrupt normal activities or the distribution of shorebirds compared to a situation without such activity” (page 2, Fox & Madsen, 1997). We counted number of potential causes of disturbance, as they occurred, during the survey period in the sampling unit (Sutherland & Green, 2004). We categorized potential human disturbances as dogs (without a leash, either alone or accompanied by people), people (running or walking, in a group or alone, or dogs with a leash), vehicles (cars, motorcycles, and ATVs), and other (horses, drones, and fishing boats).

We used generalized linear mixed models (GLMM; Zuur et al., 2009) to quantify associations of the count of shorebirds in different groups in each sampling unit with raptor and potential anthropogenic disturbance (events/minute), type of habitat, and percentage of the sampling unit that was flooded. Area (ha) of the sampling unit was also used to control for different sizes of sampling units. All models included a random effect of the sampling unit and a random intercept for species to account for different abundances among species and random slopes of the effect of human disturbance by species to control for variation in the effect of human disturbance across species. In addition, we consider random slopes for other covariates (e.g., raptor disturbance); however, the models would not converge, and since we were most interested in the effect of potential human disturbance, we considered random slopes for that covariate only. Because the distribution of shorebird counts had excess zeros and was overdispersed, we assumed a zero-inflated negative binomial distribution. Shorebird abundance was compared between migration (fall/spring) and winter, between three size classes (small, medium, and large shorebirds, sensu Page et al., 1997), and between two feeding strategies (tactile and visual, sensu Barbosa & Moreno, 1999). Species with <0.02 probability of occurrence were excluded from the analysis. Western Sandpipers (Calidris mauri) represented 67% of all counted birds. Therefore, similar analyses were performed for this single species during winter and migration. In total, we evaluated the model for 14 response variables, each representing a different combination of counts by shorebird species groups and/or seasons, but all with the same covariates considered. We considered migration (September-October and March-April) and wintering (November-February) seasons separately because we hypothesized that the responses of shorebirds to disturbance might differ during these different periods of the annual cycle.