In developed cities, bird communities are typically comprised of a few dominant invasive species that can cause considerable social and economic costs. While various studies advocate restricting anthropogenic food as a suitable management approach, a significant knowledge gap persists regarding how these species interact and respond to such an intervention. Here, we evaluate if limiting a shared food resource may affect their abundances similarly and assess if such limitations influence their niche dynamics. In Singapore, open food centres for people, colloquially known as hawker centres, serve as key food sources for three highly adapted urban birds: feral pigeons, Javan mynas, and house crows. We counted these three species across 63 hawker centres and analysed their niche dynamics across different phases – before, during COVID-19 social restrictions when dining-in was prohibited, and during an enforcement phase mandating the return of crockery. We modelled their counts, diet niche widths, and niche overlaps, considering predictors which include the sampling phases, food availability, structural characteristics of hawker centres, and spatial attributes such as distance to public housing. During social restrictions, feral pigeon and Javan myna counts showed a significant decline, while the count proportions of the three species compared to each other remained relatively stable. Hawker centres closer to bridges and public housing, and those that structurally more open, attracted more birds. The niche widths of feral pigeons and Javan mynas significantly narrowed during social restrictions due to reduced food availability. However, their niche overlaps remained consistent across sampling phases, indicating resource partitioning strategies to cope with extreme food shortages – feral pigeons adapted by foraging more on grass verges outside, while Javan mynas frequented tray return stations. This resilience in maintaining species proportions and the absence of significant niche overlap suggested the existence of an ecological balance despite substantial reductions in available food. Synthesis and applications. Our study underscores the importance of controlling human-provided food to collectively manage dominant urban bird commensals. Beyond the two social restriction phases, curbing the availability of anthropogenic food through enforcement also kept nuisance birds away, validating a cost-effective approach in reducing their counts.

A total of 63 hawker centres in Singapore were surveyed from 2019 to 2021. We conducted area searches and recorded the counts of feral pigeons, Javan mynas and house crows within the hawker centre premises and 10m of its perimeter (i.e., outside). To minimise double-counting, we only counted individuals that were detected in frontal view and omitted individuals that flew in from other sections of the hawker centre. The same surveyor conducted the surveys between different sampling occasions. We also observed their activity (i.e., feeding, perching and others which constitute moving either by flying, hopping or walking), perch location (i.e., ground, table, chair, elevated structures, tray return station and outside) and foraging substrate (i.e., ground, table, chair, crockery, tray return station and others). A tray return station is an open shelving unit for returning crockery, utensils and trays.

Predictors affecting bird commensal counts included food availability, structural attributes, spatial factors and weather. Food availability metrics included the number of food stalls operating, the frequency of feeding incidences, food waste detected (within a 1m² grid), uncleared crockery (i.e., plates and bowls that were left unattended), and the proportion of uncleared to total tray station levels. Additionally, we quantified the number of active cleaners and the extent of deterrent measures used (i.e., bird spikes, screens at tray station, nets, dummy predators and wires) by rating them from a scale of 0 to 3 (not used, low, medium to high coverage respectively). The scales for each measure were summed to obtain an overall deterrence metric. We also measured two structural characteristics of each hawker centre – the total length of elevated structures (e.g., beams, pipes, stall sign-boards and vents) representing available perches for birds to rest, and the openness as a proxy of accessibility, measured as a cross-sectional area in m² (including access points for patrons and ventilated areas) with the aid of a laser rangefinder. Spatial predictors included the proportion of public residential housing within a 200m radius of a hawker centre, and the distances to nearest Mass Rapid Transit viaduct and bridge; structures commonly used by urban adapted birds for nesting. Last, we noted if the weather was clear (i.e., no rain or otherwise) during our surveys.

To test the effect of food reduction on our bird counts, we surveyed the same 63 hawker centres during two social restriction phases and a tray return enforcement phase when mandatory return of crockery was enforced in hawker centres (Table 1). These three surveys were conducted at 1330 and 1500 as our previous surveys showed the highest bird count at these times.

Table 1: Effective dates and survey dates during pre-COVID-19, social restriction and tray return enforcement phases.

We used aggregated counts of feral pigeons, Javan mynas and house crows separately at different perches for each sampling phase and hawker centre to quantify their niche widths. The niche widths were calculated using the Shannon-Weiner index since our data were discrete. For niche overlaps between the feral pigeon and Javan myna, we used Morisita’s index of similarity since it showed the least bias with changing sample size, number of resources, and resource evenness.

We used general linear mixed models (GLMM) with negative binomial error distribution and log link function to determine the factors affecting their counts. To determine the effect of food availability on feral pigeon and Javan myna niche widths at our hawker centres, zero inflated gamma GLMM with log link function was used since our response included zero and positive values. Zero-one inflated beta regression (ZOIB) with random effects was used to determine the factors that influence the proportion of niche overlap between the feral pigeon and Javan myna.