Ecological and behavioral dimensions of coexistence in two understory babblers in Subtropical China

Abstract

Species coexistence is pivotal for ecological balance and is shaped by resource dynamics and evolutionary processes. It is generally believed that a certain degree of niche overlap is beneficial for the integrity and efficiency of mixed-species foraging flocks (MSFs) in birds, whereas the coexistence of species requires niche divergence as a fundamental prerequisite. To navigate the paradox of coexistence against the backdrop of MSFs is crucial for understanding avian biodiversity. Here, we took the angle from home ranges to address this question. VHF telemetry tracking was conducted on two widely coexisting understory birds of the mixed-species flocks, the David’s fulvetta (Alcippe davidi) and the Streak-breasted scimitar babbler (Pomatorhinus ruficollis) in subtropical Central China to investigate their spatial interactions during breeding season. Our research revealed significantly greater intraspecific home range overlap (0.376) in the fulvettas than in the scimitar babblers (0.213), which suggests differentiation in intraspecies gregariousness. Further scrutiny of habitat preferences revealed that scimitar babblers had significantly high tolerances to barren lands and croplands. Both species maintained similar home range sizes and foraging timings and paces. Our research underscores the importance of intrinsic gregariousness and convergent land cover type preferences in MSFs albeit other diverging characteristics in the breeding season.

Graphical Abstract

Graphical Abstract.

The coexistence of species has long been the core question in ecology (Darwin 1859). Various factors influence coexistence patterns, such as resource limitations (Gause 1934), foraging guilds (MacArthur 1958), resource partitioning (Schoener 1974), environmental heterogeneity (Whittaker 1956), and evolutionary history (Darwin 1859). Differences in niches are of fundamental importance to the coexistence of species and to the maintenance of biodiversity (Darwin 1859; Grinnell 1917). The overlap in niches is tolerable only to a certain extent (Macarthur and Levins 1967), and quantifying the extent of this overlap is essential for understanding the mechanism of species coexistence (Chesson 2000). For studies on foraging guilds, disentangling the coexistence of members is challenging due to the multiple factors at play, both morphological and behavioral (Jones 1977), and both the driving force of intrinsic gregariousness of the participants (Moynihan 1962) and the drawback of competition based on diet similarity that brings them together (Morse 1970).

The foraging guild is a ubiquitous phenomenon in birds. Passerine birds have developed robust foraging guilds known as mixed-species flocks (MSFs; Hinde 1952; Goodale et al. 2020). The core species of MSFs are usually highly gregarious, vagile, and vocal (Moynihan 1962). Intrinsic gregariousness could be one of the main drivers of the formation of MSFs (Morse 1970; Goodale and Beauchamp 2010). The extent of niche overlap between members of a mixed-species bird flock can vary, but it is generally believed that a certain degree of niche overlap is beneficial for the integrity and efficiency of flocks (Munn and Terborgh 1979; Schoener 1982). Further development of the niche theory becomes less applicable for the detailed roles of the MSFs by emphasizing impacts and responses targeting one species (MacAuthur 1972; Chase and Leibold 2003); therefore, only observation-based research on a broad scale from the perspective of the associations between abiotic factors and flock assemblies has attempted to address this issue, e.g., the assembly patterns of Andean bird MSFs (Colorado and Rodewald 2015) and the trait‒environment interactions in MSF assembly structures in the Nanling Mountains in southern China (Zhang et al. 2020). The fine-scale overlap of habitat preferences, intra- and interspecific gregariousness, and foraging maneuvers could provide more insightful basis for understanding the niche partitioning of bird MSFs.

Here, we take an alternative perspective on niche overlap during the breeding season by investigating the home ranges of coexisting babblers in subtropical China. The home range of an animal is represented by an area where its daily activities occur, such as foraging, mating, and caring for the young (Burt 1943). It reflects how animals respond to environmental heterogeneity through their movement, resulting in variable geographic and environmental space usage. The connection between home ranges and habitat selection via animal movements was studied in moose using movement patterns (Van Moorter et al. 2016). To study social communities with multiple members requires a more comprehensive approach to monitoring and tracking cooccurring individuals and to consider the overlapping and interacting members of different species to address how they resolve niche conflicts during the resource-demanding breeding season. The terrestrial mixed species flocks in tropical China are diverse and abundant (Zou et al. 2018) but are less commonly reported in subtropical or temperate China. We chose two understory species widespread in tropical and subtropical forests in southern, southwestern, and central China, David’s fulvetta (Alcippe davidi, fulvetta hereafter) and the streak-breasted scimitar babbler (Pomatorhinus ruficollis, scimitar babbler hereafter). The fulvetta is a small babbler of the A. morrisonia complex and one of four acknowledged, closely related, and allopatric species (Zou et al. 2007; Song et al. 2009; Gill et al. 2022). It is approximately 132 mm long and weighs approximately 17 g (Kirwan et al. 2021). The scimitar babbler is larger than the fulvetta being approximately 170 mm and weighs about 26 g (Collar and Robson 2020). Both species are sexually monomorphic, have largely overlapping insectivorous and granivorous diets, and breed from April to June (Yen 1990). In southern China’s tropical region, the fulvetta (A. morrisonia sensu lato) is the core species, and the scimitar babbler an occasional follower in the mixed species flocks (Zou et al. 2011).

The overlap of the home range reflects the extent of the bird’s overlap in habitat preferences and tolerance of cooccurring inter- and intraspecific individuals. During the short breeding season, as the requirements for resources increase, there are greater levels of intra- and interspecific overlap both spatially and temporally, causing other dimensions of niches to diverge more significantly (e.g., Lack 1946; Hutchinson 1957). In nonterrestrial communities, empirical research on this topic suggests greater dietary niche overlap in alcids during the breeding season (Gulka et al. 2019), but the dietary composition could vary among species (Jenkins and Davoren 2021). Niche hypervolumes in pygoscelid penguins are divergent against the backdrop that the species share common diets (Wilson 2010). It is crucial to have a closer understanding of how understory birds coordinate their activities during the breeding season to unveil their coexistence.

The dense vegetation, combined with the high levels of rainfall and humidity, creates an environment in which it is difficult for researchers to access and monitor birds in tropical and subtropical forests, even more so for the target species in this study, which favors mountainous understories. Despite they are large in numbers, they are also inherently cryptic and elusive, making direct observation of the behavior of single individuals challenging. The observation of breeding birds or nest visits could disturb the birds. To accurately track and study understory birds in these habitats, our research employed VHF (very high frequency) radiotelemetry by manual tracking and triangulation to overcome the challenges posed by the environment.

This study set off to explore the niche overlap of two understory passerine species by testing the home ranges and investigating their activity patterns in response to land cover configuration during the breeding season. Despite their different body sizes, our results show that the average home range sizes of scimitar babblers and fulvettas are relatively comparable, contrary to the expectation of larger animals having larger home ranges (Harestad and Bunnel 1979). Further estimation of the overlap ratios of home ranges revealed significantly greater intraspecific overlap in the fulvettas than in the scimitar babblers. Overall, species sociality or intrinsic gregariousness likely plays a significant role in shaping their daily activities and interactions during the breeding season. While both fulvettas and scimitar babblers show a strong preference for forest habitats over other land cover types, they exhibit significant differences in tolerance to habitat disturbances. Multiple ecological and behavioral characteristics were at play in facilitating their MSF behaviors. By providing preliminary quantitative insights into the movement patterns of these two species during the breeding season, this study suggests a possible basis for the coexistence of understory bird communities in subtropical and tropical forests.

Materials and methods

Study site and timing

Our research was carried out in the Enshi Autonomous Prefecture located in Hubei Province, China. This region is located in Central China, characterized by deciduous subtropical forests, and the semimonsoon climate. The experimental site was situated within a secondary forest interspersed with human settlements at the geographical coordinates of approximately 30°37′10″N, 110°5′40″E (Figure 1). This site was selected due to its substantial populations of fulvettas and scimitar babblers. Historically, before the 1970s, the absence of robust wildlife protection legislation allowed local hunters capturing birds for subsistence. Villagers developed a profound understanding of local birds. Leveraging this local expertise, we design the experiments and constrain the field site’s boundaries with the knowledge that they travel limited distances. The experimental field was selected because it had sufficiently large coexisting populations of the two species and met the requirement of accessibility of the operators. The major activities of the birds were concentrated in a forest segment bisected by a narrow and overarched trail approximately 2 m wide. We acknowledge the possibility that individuals could move beyond the detection range of our radio receivers, potentially leading to an underestimation of their actual home ranges. To mitigate this, we conducted our study when home range sizes were typically the smallest due to nesting and territorial behaviors (Wiktander et al. 2001). We initiated our experiments in May 2023, the expected peak of the breeding season for the species under study. Throughout the month, two operators conducted observations and telemetry triangulation of the species along the forest trail.

Figure 1.

A land cover map with 75% KDE home range centroids for all 28 individuals.

Bird capture and equipment mounting

To determine optimal sites for mist netting, we conducted one preliminary morning observation along the main trail and adjacent side trails, focusing on the foraging grounds of the two target species. The study area encompassed one single continuous population of both species. Two mist nets measuring 3 × 12 m with 36 mm mesh sizes were deployed 200 m apart targeting two foraging grounds where the subject species were most frequently observed. We carried out bird capturing over three consecutive days from 7:00 a.m. to 7:00 p.m. Operators checked and collected the birds every hour with no breaks. At most three individuals of the same species from each net in one collection were collected to carry VHF transmitters. We weighed and assessed the captured individuals to ensure that they were strong and healthy enough to carry the VHF transmitters and band rings (Table 1). Sex and age were not included in our study considering sex and age have limited influence on home range sizes (Rolando 2002; Guppy et al. 2023), and our sample sizes for different sex and age classes were insufficient for reliable statistical analysis. Both species carried backpack PicoPip VHF radio transmitters from Lotek Corporation (USA). The model for the fulvettas weighed 0.46 g, and that for the scimitar babblers weighed 0.70 g; both of these values are less than 5% of the birds’ body weight (Table 1). Captures were then banded and released. Due to the presence of thick understory vines and shrubs at the field location, no exploration for the confirmation of nesting or breeding behaviors was attempted. Previous studies have shown that the fulvetta breeds from April to June, after which they start forming larger family flocks for foraging between June and July (Zhou 1989). The scimitar babbler also breeds during a similar period (Yen 1990). While the two species share overlapping nesting habitat types, there is less overlap in terms of nest height and substrate preferences (Li et al. 2019). The potential competition for resources between the two species likely lies in territorial space rather than in nesting substrates or vertical height niches. We acknowledge that tracking both breeding and non-breeding individuals may introduce certain limitations. First, it increases variation in home range sizes, which could underestimate the divergence of their spatial behaviors; second, it introduces unaddressed variation into the modeling analysis of the subjects’ movements. We performed additional observations the day after the conclusion of the tracking without specific targets and found five banded fulvettas and two banded scimitar babblers in the experimental field with the transmitters still properly attached, confirming the validity of the handling process. Furthermore, these observations served as an indication that these species were prone to staying within a particular region at this time of year. The sampling of individuals in our study resulted in unbalanced numbers between the two species due to differences in population sizes and behavioral properties. According to previous research on tropical populations, the scimitar babbler was observed at approximately 60% the frequency of the fulvetta (Zou et al. 2011) and tended to stay close to the ground (Zou et al. 2005).

Table 1.

Telemetry and home range information of all individuals.

ID	Species	Weight (g)	Starting date	Tracking days	No. of positions	95% MCP area (ha)
F01	A. davidi	16.19	May-10	14	33	2.62
F02	A. davidi	17.23	May-10	12	39	5.11
F03	A. davidi	16.25	May-11	9	20	3.69
F04	A. davidi	17.15	May-10	14	37	3.05
F05	A. davidi	15.71	May-10	12	35	2.26
F06	A. davidi	15.56	May-10	13	34	4.38
F07	A. davidi	15.75	May-10	13	37	1.42
F08	A. davidi	16.50	May-10	11	20	1.98
F09	A. davidi	17.68	May-10	15	35	4.27
F10	A. davidi	21.03	May-10	6	11	2.98
F11	A. davidi	18.75	May-10	12	31	1.70
F12	A. davidi	15.96	May-10	14	36	1.83
F13	A. davidi	16.73	May-10	11	28	1.47
F14	A. davidi	15.97	May-10	14	42	1.52
F15	A. davidi	16.38	May-10	14	47	2.16
F16	A. davidi	17.92	May-10	13	36	3.52
F17	A. davidi	16.65	May-10	12	24	12.9
F18	A. davidi	17.50	May-10	13	38	3.58
F19	A. davidi	15.87	May-10	13	31	0.577
F20	A. davidi	17.00	May-10	12	28	3.20
P01	P. ruficollis	28.03	May-10	21	84	3.44
P02	P. ruficollis	25.71	May-10	11	24	1.13
P03	P. ruficollis	26.78	May-10	11	24	1.98
P04	P. ruficollis	28.00	May-10	20	68	1.51
P05	P. ruficollis	29.02	May-10	10	19	5.12
P06	P. ruficollis	26.71	May-12	20	70	0.846
P07	P. ruficollis	25.33	May-12	18	72	3.03
P08	P. ruficollis	30.21	May-12	18	81	4.76

Telemetry triangulation

Each subject’s telemetry tracking began the day after capture and continued until the transmitter’s battery ran out or until the 20th day since the capture time (16 days for the fulvetta). We performed positioning from 6:30 a.m. to 8:00 p.m., spaced one to three hours apart. Two operators who communicated via two-way radio stood at least 50 meters apart, and directed the target transmitter signal using a radio receiver model SRX1200 (Lotek Corporation, USA). Coordinates and bearings were recorded using the dual-frequency positioning system on an Android smartphone. Each positioning event was performed within 10 min by two operators. At each positioning event, the time, coordinates of the operator and bearing were recorded independently by each operator. Unexpected circumstances such as predation, weather, and transmitter detachment could lead to loss of signal. Due to the extremely lightweight design of the transmitters, their operational lifespan was short, with the signal gradually becoming undetectable after approximately 10 days for the fulvetta model and 14 days for the scimitar babbler model. Our tracking attempts were mostly conducted within 500 m from the mist net sites, which we denoted as a regular range. We considered a signal lost if an individual’s signal was undetectable over three attempts. Upon the first signal loss of a subject within the regular range, we arranged one expedition tracking to search along other trails or roads up to a maximum of 1 km away to reacquire the signal. If the signal was found outside the regular range, the new location was included in the regular range for subsequent tracking. If the first expedition search failed to reacquire the signal, or if the signal loss occurred for the second time, we assumed the battery had depleted. In such cases, we performed once-a-day positioning attempts from the regular range until the 20th day (16 days for fulvetta) since the capture time.

Data analysis

Position calculation and home range estimation

We checked the bearing and coordinate data for positioning failures, signal errors, and overtime positioning events. A signal error was indicated by an operator position that was over 5 m away from the trails and roads. Positioning events with a time gap over 10 min were considered overtime. All the records with the anomalies mentioned above were discarded. Target positions were calculated using the trirmr function in the telemetr package in R (Rowlingson 2012). Positioning events with calculation failure were also discarded. Individuals with fewer than ten calculated positions were omitted from further calculations.

All analyses were performed in R 4.3.2 (R Core Team 2020). We used the Minimum Convex Polygon (MCP) to approximate the home range for each individual (Mohr 1947) and the probabilistic estimator of the Kernel Density Estimator (KDE; Worton 1987) to infer utilization distribution of the home range. KDE generates a continuous probability density surface based on spatial relocations, with the density peaks reflecting areas of concentrated space use. Home ranges for each individual were derived using the hr_mcp functions in the amt package in R (Signer et al. 2019) based on 95% MCP. Since our average sample sizes for each individual are small, we adopted the suggestion of Wauters et al. (2007) to delimit areas enclosing higher than 75% percentile density of the KDE inference as the core range of activities using the hr_kde function in amt (Powell 2000; Signer et al. 2019). The centroids of the core ranges deduced using the gCentroid function in the rgeos package were considered centroid sites (Bivand and Rundel 2021). All distances between the triangulated positions and the centroids of the corresponding individuals were calculated with the pointDistance function in raster. For a brief and straightforward understanding of the activity range, we calculated the median distance from positions to the respective home range centroids for each individual using median function in R base code. Then, we performed Wilcoxon’s test to compare the medians between the two species using wilcox.test in R base code. We further compared the home range sizes of the two species with Wilcoxon’s test also using the wilcox.test function in R.

Based on the home range inferred by 95% MCP, the overlap of the home range between all pairs of individuals was calculated with amt::hr_overlap. This process created an asymmetric matrix of overlap for further comparison. Due to the unbalanced sample sizes between the two species—having fewer individuals of the scimitar babbler than the fulvettas—we incorporated permutation methods to discern the variance in home range overlap both within and between species. We conducted two sets of permutation tests targeting intraspecies and the complete overlap data respectively (Table 2). First, we focused on intraspecific overlap based on pairs of fulvettas and scimitar babblers. We permuted (n = 999) subsamples from the overlap matrix of the fulvettas to be equal to the sample size of the scimitar babblers and tested for differences in median values with Wilcoxon’s test. Second, we also tested whether each species exhibited affinity or avoidance to conspecific or interspecific individuals. For intraspecies interactions, there could be breeding competition for identical resources, and it is expected that increased territoriality during the breeding season would lead to lowered overlap level. We shuffled the species identifiers across the dataset for all individuals and extracted the resulting intraspecific and interspecific subsamples. On the one hand, should a species exhibit pronounced conspecific gregariousness, we would anticipate a lower conspecific overlap in the shuffled dataset compared to the observed dataset, and vice versa; on the other hand, a marked interspecific avoidance would manifest as a greater interspecific overlap in the shuffled dataset compared to the observed data, and vice versa. The niche partitioning theory (Schoener 1974), which contradicts the foraging similarity in MSFs empirically (e.g., Sridhar et al. 2009), leaves the outcomes open-ended. Species may exhibit greater intraspecific gregariousness due to an intrinsic impulse driving their flocking behavior, or they may show less intraspecific contact compared to interspecific contact due to more intense intraspecific competition (Lotka 1925; Volterra 1926) during the breeding season when the flocking behavior is less prevalent. All analyses were performed using customized R code.

Table 2.

Permutation results of median overlap level inter- and intra-specifically.

		Median of the first sample	Median of the second sample
Intraspecific overlap	Fulvetta vs scimitar babbler	0.368(nperm = 999)	0.213
	Fulvetta vs shuffled dataset	0.376	0.341(nperm = 999)
	Scimitar babbler vs shuffled dataset	0.213	0.338(nperm = 999)
Interspecific overlap	Observed vs shuffled dataset	0.305	0.346(nperm = 999)

Land cover utilization

To investigate the habitat preferences of the two species, we manually categorized land cover types for the field located at approximately 30°37′10″N, 110°5′40″E using satellite imagery from Mapworld (www.tianditu.gov.cn) from 2021 in ArcGIS 10.2. One day prior to our fieldwork, we conducted a brief drone-based survey to validate the consistency of the satellite-based map from 2021. We used a DJI Mini 3 Pro (48MP CMOS, 1/1.3-inch sensor) flown at 120 m altitude to acquire images and videos over the study area, and manually compared them to the 2021 Mapworld data. Our observations confirmed that the land cover types identified from the 2021 imagery accurately reflected the current conditions, with no significant changes in any habitat borders. The land cover types were simplified into five categories: cropland, forest, shrubland, barren land, and water. Our data found no birds were positioned in the water category; therefore all further analyses omitted this land cover type. Given that our target species are predominantly understory dwellers, all non-vegetated land, human buildings, carways, and pavements were collectively classified as barren lands. Local people intensively cultivate potatoes, beans, and corn, which we designated as cropland, while all other exposed lands with sparse vegetation were classified as shrublands. There was no grassland in the experimental field. We extracted the land cover types of all positions with extract from raster package (Hijmans 2022). To restrict the area included in the land cover type calculation, we cropped the experimental field map. We excluded the three most distant samples in each direction to mitigate potential edge effects. We then defined the map’s extents by the maximum and minimum positions of the remaining subjects. The resulting boundaries are between 30°36′43″N and 30°37′39″N latitude, and between 110°5′22″E and 110°6′0″E longitude (Figure 1). The overall area sizes of each land cover type in the experimental field were calculated using the area function in the raster package on the cropped map. The compositions of four land cover types were calculated. We performed Chi-squared tests on the position counts to assess the differences in land cover usage between the two species, and two Chi-squared goodness-of-fit tests for each species against the land cover composition of the map. Both using the chisq.test in R base code. After finding that each species’ land cover usage differed significantly from the expected distribution based on the map, we calculated the selection ratios for the four land cover types for both species, following Manly et al. (2007), to determine their preferences or avoidance. We assessed significance by examining whether the 95% confidence intervals (CIs) of the selection ratios included 1.0. If the 95% CIs included 1.0, the land cover type was used in proportion to its availability, suggesting no significant preference or avoidance. Conversely, if the 95% CIs did not include 1.0, it signified a preference or avoidance for that land cover type. Specifically, a selection ratio significantly greater than 1.0 indicated preference, while a ratio significantly less than 1.0 indicated avoidance. Additionally, to examine species-specific sensitivity to disturbed land covers, we compared the distances from the centroids of the 75% KDE home ranges to the nearest border of barren lands and croplands between the two species using the Wilcoxon’s test. The centroids represented the most visited area of the birds, and testing the distances revealed the difference in the tolerance to disturbed habitats between the two species.

Movement steps

We further examined the movement patterns of the two species, which are influenced by a complex interplay of environmental and biological factors. We transformed the dataset of positions into steps using a customized procedure in R. Due to the manual nature of our data collection, the triangulated positions were discrete and unsuitable for stepwise modeling, such as Hidden Markov processes. As an alternative, we used Generalized Linear Mixed Models (GLMMs) to analyze the average translocation of steps, with the individual ID included as a random effect. We calculated the step length using the pointDistance function from raster package (Hijmans 2022) and time intervals using difftime from R base code. Then, the average translocation was calculated for each step by dividing the step length by the time interval. While the average translocation does not directly measure velocity, it provides an indirect reflection of the species’ vagility, foraging intensity, and possibly the difference in foraging strategy over time. This interpretation aligns with correlated random-walk models, which indicate that higher directional persistence yields straighter, faster trajectories toward a target (e.g., Codling et al. 2008). For the convenience of reading, we would henceforth refer to the average translocation as “speed.”

We defined a step as two sequential positions recorded within a 130-min interval based on the availability of data and to maximize the dataset for modeling. This threshold was selected after testing multiple intervals to maximize the number of movement steps available for modeling while maintaining the most temporal proximity (Supplementary Figure S1). For an initial understanding of the speed of the two species, we calculated the median speed for each individual and compared the medians between the two species using the Wilcoxon’s test to determine whether the two species differ. Then, we computed eleven variables that could be predictive of the step, including time of day, distance to trail, distance to barren land, distance from centroid, acuteness from centroid direction, whether or not crossing a trail, density of four land cover types (that the birds actually used), and the possible confounder of the distance of the further operator for each positioning event (Table 3; Supplementary Figure S2). While these values are not direct measures of the birds’ movement speed, they serve as indicators of their mobility or foraging behavior during the observation period. For example, birds that are brooding or roosting, and spend most of their time still sitting could have lower values; for birds on a strongly biased random walk away from or toward the centroid site, we may detect significant association between speed and acuteness from centroid direction. Flocking birds likely follow persistent foraging courses (Mcclure 1967), a pattern related to the area-restricted foraging tactic theorized in forest birds (Robinson and Holmes 1982). Under such tactics, movement speed and direction could present certain biases (Kareiva and Odell 1987; Grünbaum 1998). The coarse temporal resolution of our data prevents us from addressing stepwise directional changes, we incorporated the acuteness from centroid direction as an explanatory variable, substituting for turning angle to gain insights into the birds’ biased movement patterns and habitat utilization.

Table 3.

Explanatory variables used in GLM modeling.

No.	Variable name	Content
1	Time of day	Numerical, the time of the start point in decimal numbers;
2	Distance to trail	Numerical, closest measurement of start point to the trail;
3	Distance to barren land	Numerical, closest measurement of start point to any patch of barren land, 0 if inside barren lands;
4	Distance from centroid	Numerical, distance between start point and centroid, calculated with pointDistance function in raster package (Hijmans 2022);
5	Acuteness from centroid direction	Numerical, the angle formed by the step and the vector from start point to centroid;
6	Crossing trail	Binominal, whether or not the step crossed the trial;
7-10	Land cover density (four variables)	Numerical, the proportion of each land cover type within the buffer range of the step, including forest, shrubland, cropland and barren land sequentially;
11	Distance to operator	Numerical, the distance between the start point and the more distant operator of the two.

Before fitting the GLMMs, we calculated the Variance Inflation Factors (VIFs) for the predictor variables to assess and reduce multicollinearity among the predictors. This was done using the vif function from the car package (Fox and Weisberg 2019) on an equivalent Generalized Linear Model (GLM) excluding the random effect. The GLMs were fitted using the glm function in R base code. Variables were selected to ensure that the VIFs of the variables included in the model were lower than five (Supplementary Table S1). This process removed two variables: distance to trail and density of cropland. To account for the different land cover compositions resulting from different search patterns, we calculated land cover percentages based on two hypotheses representing distinct foraging behavior: the fixed-course and the exhaustive search. The land cover types could be predictive of the movement patterns, and the buffer distance reflects the subject’s sensitivity to land cover types in a step. For the fixed-course movement hypothesis, we calculated the land cover along the direct passage between the two endpoints of each step. This was done by creating buffer zones around the straight-line path using the st_buffer function from the sf package (Pebesma 2018). We generated 50 different sets of land cover percentages by varying the buffer distances from 10% to 500% of the step length, in increments of 10%. For the exhaustive search hypothesis, we simulated thorough and extensive search behavior by calculating the land cover within the smallest circular area encompassing both endpoints of each step. This approach also used the st_buffer function from the sf package. Combining these approaches, we obtained a total of 51 sets of land cover percentages.

We then iteratively ran 51 GLMMs with the glmer function in lme4 package (Bates et al. 2015) for each species, modeling movement pace as the response under a Gamma distribution with a log link. This iterative modeling allowed us to assess how these different representations of land cover, reflecting distinct search patterns and sensitivities to land covers, influenced movement speed. We compared models using Akaike’s Information Criterion (AIC) to select the best-fitting model; the model with the lowest AIC was considered the best fit. To evaluate model adequacy, we conducted simulation-based residual diagnostics with the DHARMa package, employing 1000 simulations using simulateResiduals (Hartig 2024). We assessed residual uniformity with testUniformity, which performs a Kolmogorov-Smirnov test (KS test), and checked for over- or underdispersion using nonparametric dispersion test with testDispersion function. Finally, we inspected Q-Q plots and rank-transformed model prediction values of the optimal models to confirm there were no systematic departures from model assumptions.

Upon finding that crossing trails was significantly predictive of step speed via the GLMMs, we performed a series of Wilcoxon’s tests to verify the variation in step speed and identify the range of effect of the possible breeding and roosting behaviors. Some animals are known to vary in speed in open environment, possibly as a mechanism to increase foraging efficiency or avoid predation risks (Turchin 1998; Da Silveira et al. 2016). Other factors such as distance from core areas and breeding ecologies, have rarely been addressed. If the time birds spent stationary around the centroid area significantly influenced their speed, we would anticipate an absence of association between speed and crossing a trail when a step started far away from the centroid, where the birds would unlikely return to the centroid in one step. The marginal distance inferred from the window scan scheme, where the birds were more prone to leave the centroids (indicated by the lack of association of speed and crossing the trail), would serve as referable information of movement choices for modeling and simulation of forest birds. The test was based on a window scan scheme on the distance from the starting point of the steps to the centroids. The scanning scheme had a one-meter step length and a window size of 10 m. The speeds of steps crossing and not-crossing trail were compared by Wilcoxon’s tests using wilcox.test in R base code.

Results

Overall patterns of bird positions and centroids

In total, our triangulation methods yielded 1084 positions across 20 fulvettas and 8 scimitar babblers (Table 1). Positioned points were taken from ca. 6:15 to ca. 19:45 with an intermission at 13:00 (Figure 2). The samples were consistent throughout the day and comparable between two species, resulting in 642 effective coordinates for fulvettas and 442 coordinates for the scimitar babblers in total. On average, each fulvetta was positioned 32.1 times, averaging 2.5 positions per individual per day, while each scimitar babbler was positioned 55.2 times, averaging 2.6 positions per individual per day.

Figure 2.

Number of effective positions throughout the day for David’s fulvetta (blue) and streak-breasted scimitar babbler (red).

We also calculated the distances of the birds’ positions from their home range centroids. The median distance of the fulvettas from their centroids was 51 m (range 20–87 m), while for the scimitar babblers, it was 43 m (range: 10–96 m). Wilcoxon’s test revealed no significant difference between the two species (P = 0.601). The distance from centroids to barren lands is indicative of the tolerance to human disturbance. Wilcoxon’s test revealed significant differences between the species in terms of the proximity of their centroids to barren lands (P < 0.001), with scimitar babblers having a median distance of 68 m (range 7–119 m) and fulvettas having a median distance of 119 m (range 44–170 m). The distance to croplands showed less variation and was not statistically significant (P = 0.182), with medians of 44 m for fulvettas and 32 m for scimitar babblers.

None of the positions fell in waters. Of the 642 triangulated positions of fulvettas, 35 (5.4%) occurred in cropland, 582 (90.7%) in forest, 16 (2.5%) in shrubland, and 9 (1.4%) in barren land; of the 442 triangulated positions of scimitar babbler, 38 (8.6%) were in cropland, 381 (86.2%) in forest, 6 (1.4%) in shrubland, and 17 (3.8%) in barren land (Figure 3). The land cover composition of the map area included 26.8% cropland, 59.3% forest, 6.3% shrubland, and 7.6% barren land. Waters were omitted. Both distributions significantly deviated from the land cover composition of the map (Chi-squared test, P < 0.001 for both tests). Chi-squared test on the counts revealed that the two species also differed in land cover preferences (Chi-squared test, P < 0.05). Although the two species diverged in overall counts, the 95% CIs of selection ratio suggested that both species significantly prefer forests and avoided all other land cover types significantly (Table 4).

Figure 3.

Percentages of four types of land cover for David’s fulvetta, streak-breasted scimitar babbler, and the experimental field.

Table 4.

Selection ratios of the two species. The counts of each land cover types were extracted from all positions of the two species.

Land cover types	Utilized frequency	Counts of land cover usage	Selection ratio
			Lower CI (5%)	Upper CI (95%)
(A) David’s fulvetta
Cropland	0.055	35	0.147	0.280
Forest	0.907	582	1.492	1.568
Shrubland	0.025	16	0.243	0.640
Barren land	0.014	9	0.097	0.353
(B) Streak-breasted scimitar babbler
Cropland	0.086	38	0.237	0.434
Forest	0.862	381	1.401	1.509
Shrubland	0.014	6	0.097	0.476
Barren land	0.038	17	0.318	0.808

Home range, sizes, and overlaps

Home-range sizes calculated with 95% MCP areas varied among individuals but not between species. For fulvetta individuals, the average home range size was 2.80 ha (n = 20, range 0.85–5.12 ha), and for scimitar babblers, it was 2.50 ha (n = 8, range 0.58–12.88 ha). The home range sizes of the two species were comparable (Wilcoxon’s test, P = 0.710).

We also calculated the percentages of neighboring home range overlaps using asymmetric overlap matrices with zero values removed (Supplementary Table S2). The average overlap rate for adjacent home ranges for all individual pairs, regardless of species, was 0.342 (n = 540 of 756 pairs of individuals; range 0.000–0.995). For the fulvetta population, the home range overlap was high, averaging 0.376 (n = 306 of 380 pairs; range 0.008–0.958) intraspecifically. The scimitar babbler was less enthusiastic about conspecific daily activity overlap, with a lower overlap rate of 0.213 (n = 28 of 56 pairs; range 0.039–0.968). The mean overlap rate between the interspecies individual pairs was 0.305 (n = 206 of 320 pairs; range 0.000–0.995), which was greater than that of the scimitar babblers but lower than that of the fulvettas. We also tested the significance of the difference between the conspecific overlap of two species by permuted sampling from the fulvettas. The overlap rates were significantly different between the permuted subsamples of intraspecific fulvettas and scimitar babblers (t test, P < 0.001). The results showed that the fulvettas were significantly more tolerant of conspecific encounters in their daily activities than the scimitar babblers, probably even more than the interspecific level.

The two sets of permutation tests with shuffled species identifiers further demonstrated the elevated intraspecific overlap in the home range of the fulvettas (Table 2). For within-species overlap, the shuffled dataset with a sample size identical to that of the fulvettas dataset had a median of 268 pairs of adjacent individuals, which was slightly lower than the observed value (n = 306). In contrast, the shuffled scimitar babblers had a higher number of adjacent pairs, at 42, compared to the observed value (n = 28). However, both differences were only highly probable but not statistically significant (Wilcoxon’s test, P = 0.099 for fulvettas and P = 0.098 for scimitar babblers, respectively). The extent of overlap was also different from the observed value. For fulvettas, the median intraspecific overlap was 0.341 based on a shuffled sample size of 20 individuals, which was lower than observation; for scimitar babblers, it was 0.338 based on a shuffled sample size of 8 individuals, which was higher than observation. The differences were both significant (t-test, P < 0.001 for both tests). These results support that fulvettas were more likely to gather, while scimitar babblers were more likely to disperse, regardless of sample sizes.

For between-species overlap, the permuted dataset showed a significantly greater number of bird pairs with overlapping home ranges based on shuffled species identifiers, at 232 pairs compared to 206 pairs in the observed dataset (Wilcoxon’s test, P < 0.05); the extent of the overlap was only likely greater than the observed rate, at 0.348 and 0.305 (t-test, P = 0.070). The two species significantly avoided interspecies overlap in their daily activity ranges and probably overlapped less with interspecies neighbors.

Translocations and movement patterns

We converted the point dataset into 138 steps for the fulvettas and 174 steps for the scimitar babblers. The average translocation is 26.4 m (6.6–82.8 m) per hour for the fulvettas’ medians and 23.5 m (6.6–103.8 m) per hour for the scimitar babblers. The Wilcoxon’s test showed no significant difference between the two species (P = 0.897), suggesting similar movement paces between the two bird species.

In the iterative process for the optimal GLMM, all candidate GLMMs exhibited no evidence of residual misfit by KS test and dispersion test, validating that our GLMMs satisfy key assumptions of residual uniformity and dispersion (Supplementary Table S3; Supplementary Figure S3). The optimal model for the fulvetta and the scimitar babbler were different. For the fulvettas, the best-fitting model was based on the fixed-course hypothesis with land cover sensitivity within a buffer distance equals to 350% of the step length along the straight line (AIC = 143.3, Table 5). In contrast, for the scimitar babblers, the best-fitting scenario was the exhaustive search hypothesis (AIC = 133.1, Table 5). Different models of fit between the two species possibly indicate divergence in sensitivity to foraging environments or difference in foraging strategies, but further data was required to validate this result considering the oversized suggested buffer distance.

Table 5.

GLMM results for the two species. For David’s fulvetta the land cover variables were generated from 350% step length buffer area of the fixed route; and exhaustive search for the streak-breasted scimitar babbler.

(a) David’s fulvetta
Variable	Estimate	Standard error	t value	P-value#
Time of day	−0.025	0.070	−0.358	0.720
Distance to barren land	0.105	0.080	1.321	0.187
Distance from centroid	0.534	0.157	3.398	0.001***
Acuteness from centroid direction	0.171	0.072	2.368	0.018*
Crossing trail	0.422	0.160	2.633	0.008**
Forest density	0.091	0.080	1.135	0.256
Shrubland density	0.046	0.079	0.584	0.559
Barren land density	−0.142	0.079	−1.802	0.072
Distance to operator	0.101	0.161	0.625	0.532

(b) Streak-breasted scimitar babbler
Variable	Estimate	Standard error	t value	P-value#
Time of day	0.037	0.062	0.587	0.557
Distance to barren land	0.206	0.083	2.485	0.013*
Distance from centroid	0.233	0.084	2.767	0.006**
Acuteness from centroid direction	−0.024	0.063	−0.376	0.707
Crossing trail	0.405	0.162	2.497	0.013*
Forest density	−0.155	0.066	−2.373	0.018*
Shrubland density	−0.022	0.063	−0.353	0.724
Barren land density	0.034	0.066	0.519	0.604
Distance from operator	0.341	0.085	4.002	0.000***

^#Significance codes: 0 “***” 0.001 “**” 0.01 “*” 0.05.

GLMMs of the two species revealed multiple variables associated with the speed. Both crossing a trail and increased distance from the centroid were significant predictors of an increase in speed for the two species (Table 5, Supplementary Figure S4). Distance from the home-range centroid served as a proxy for spatial familiarity, with locations nearer the centroid denoting core, frequently visited areas. This result indicated the birds accelerated when their route involved gap crossing or in unfamiliar environments, likely as a strategy to optimize foraging efficiency while minimizing predation risk (Brown and Kotler 2004). In fulvettas, greater angular deviation from centroid direction was positively associated with step speed, an effect absent in scimitar babblers and one that may inform future research. This association probably reflects heightened vigilance or a change in pattern from more persistent outbound activities to tortuous inbound returns. For the scimitar babblers, distance to barren land and forest proportion were associated with the speed. Notably, the scimitar babblers moved slower in forests, which may indicate intensive foraging behaviors. However, we also found that the significance of the association between crossing a trail and speed could vary depending on the starting distance from the centroid when partitioning the steps. According to the window scan results (Supplementary Figure S5), within ca. 13 m from the centroid, the birds traveled slower in this distance range than in other ranges, and were significantly faster when crossing the trail; beyond 23 m, the birds traveled faster regardless of whether they crossed the trail or not. In the intermediate range of 12–23 m, birds were more likely to keep moving if they crossed the road compared to not crossing, represented by a significantly greater speed when crossing roads, as supported by Wilcoxon’s test (P < 0.001). The scale of these distances provided information for understanding the condition of the transition in movement states, e.g., from roosting to active foraging.

Discussion

The scimitar babbler and the fulvetta are widespread birds in tropical and subtropical understories and common members of mixed species flocks. The coexistence of these species during the breeding season presents an opportunity to investigate the intricate mechanisms of niche differentiation and overlap. In this study, we focused on their home range sizes and overlaps and land cover preferences and explored their movement patterns using manual telemetry positioning.

Commons and variances between home ranges

The two species exhibited similarities in several aspects, including their home range sizes, preference for forests, and paces of movements. The convergence of home range sizes of the two species against the backdrop that they are different in body sizes is probably a result of the decrease in home range size during the breeding season (Sorato et al. 2016). It has been widely recognized that larger-bodied animals have lower population densities (Juanes 1986), and larger body masses are associated with larger home ranges (Harestad and Bunnel 1979). This association was confirmed in the study by Ottaviani et al. (2006), which showed that larger body mass correlates with increased home-range size in Italian birds. Notably, the definition of home range in the present study differs from that in our analysis, as it was “considered as the whole area used by a species in a year” (Burt 1943). A study on group home range size in the chestnut-crowned babbler suggested a decrease in home range size during nesting (Sorato et al. 2016). Thus, the convergence in home range sizes observed here may reflect species-specific reductions during the breeding period. Forest preference was indicated by selection ratios based on land cover types, yet a Chi-squared test revealed a significant association between species and land cover use, highlighting a divergence in detailed patterns. Additionally, the centroids of scimitar babblers were significantly closer to barren lands and probably cropland than those of fulvettas, indicating a greater tolerance for disturbed habitats. This difference in habitat usage could contribute greatly to the partitioning of niche hypervolumns.

Comparison of the speed between two species found no significant divergence in the pace of their movement, which agreed with the similar home range sizes. However, the results from GLMMs suggest that the foraging strategies of the fulvettas and the scimitar babblers are probably diverged, as each species’ movements were best explained by different predefined models. However, validating these results requires more sophisticated data.

The fulvetta and scimitar babbler coexisted through these commons and variances. For their daily activities, shared factors, including the overlapping diets (Yen 1990), common preference for forest habitats and comparable foraging speeds facilitated their bond in the MSF. The sociality of the fulvettas also played a significant role in this process (Morse 1970). Divergent factors, including foraging height (Zou et al. 2005), tolerance of disturbed habitats, and possibly different foraging strategy were undermining this alliance, with the lower level of sociality in scimitar babblers further segregating the two species. The divergences seemingly outweighed the convergences in aspects of their daily activity. However, our study did not address the divergence of nesting ecologies (Li et al. 2019).

Fulvettas are more tolerable of conspecific contact

The intraspecies home range overlap of the fulvettas was evidently greater than that of the scimitar babblers. Compared to the scimitar babblers, the fulvattas are excessively tolerant of conspecific interactions, although the intraspecific overlap rate is far from complete overlap. As Hsieh & Chen (Hsieh and Chen 2011) pointed out in their research targeting the MSF group led by the fulvettas (A. morrisonia sensu lato), it appears that niche overlap could enhance the cohesion of MSF to some degree. The stronger sociality in fulvettas aligns with their nucleus role in the MSF. Our results during the breeding season showed that their intraspecific home range overlap could reach 0.376, denoting the lower limit of the intraspecies overlap because the fulvettas would form large flocks in non-breeding season and likely increase in home range sizes (Sorato et al. 2016). The breeding male Swainson’s Warbler had an intraspecies overlap rate of approximately 0.28 (Anich et al. 2009). We performed data reinterpretation of the study on Blue-crowned Laughingthrush by Liu et al. (2020) and found the average intraspecies overlap rate was 0.33. Observation based research suggests 0.15-0.34 intraspecific overlaps in American redstarts (Cooper et al. 2014), which is also lower than the intraspecific overlap of the fulvetta. The intraspecific overlap in fulvetta exceeds these examples, highlighting its strong sociality and tolerance to conspecific contacts.

The intraspecies home range overlap of the fulvetta also exceeded the interspecies overlap between itself and the scimitar babbler (mean = 0.305). Although comparable to the distributions of some of the samples afore mentioned, the observed interspecies interaction was significantly low both in the probability of overlap and the extent of overlaps than the expected level inferred from the shuffled dataset.

In contrast to the fulvettas, the scimitar babblers show a notable and meaningful tolerance for barren lands and croplands. Considering this, it seems that scimitar babblers may possess a greater degree of adaptability to landscapes that have been modified by human activities. The fulvettas, however, have exceptionally greater overlap of conspecific home ranges, not only exceeding the conspecific overlap level in scimitar babblers but also exceeding the interspecific overlap level. Internal gregariousness is hypothesized to be one of the main driving forces and characteristics of the core species of mixed species flocks (Moynihan 1962). In contrast, the scimitar babbler attends the MSF only 3.7% of the time (Zou et al. 2011) and moves in smaller groups. These results indicated the scimitar babblers probably turned less sensitive to barren land as an adaptive strategy compared to the fulvettas.

Synthesis of behavioral and ecological drivers of coexistence

Our study provides a coarse understanding of the behavioral ecology of these two species in the middle of mixed habitats, highlighting how intrinsic gregariousness facilitates coexistence in the overlapping species pair. The interspecific home range overlap rate of the two species was moderate and not distinctive from other species, while the intraspecific overlap differs significantly from each other. Three factors were underscored in our study to be aliased to MSFs: the sociality of the fulvetta, common preference for forest habitats, and synchronized movement speed (Powell 1985). On the other side of the three factors are low sociality and slightly higher tolerance for disturbance in the scimitar babblers, and diverged movement strategy, which segregate niches.

The two species followed the competitive exclusion principle (Macarthur and Levins 1967) circumventing the limitation that species with identical resource requirements cannot coexist. In this case, through multidimensional differentiation: divergences in sociality and tolerance to disturbance alleviate direct competitive pressures. In the context of Chesson (2000), the potential different foraging strategies could expand niche differences, while adaptive balance could be achieved by high tolerance to the disturbance in scimitar babblers versus high resilience of conspecific society in the fulvettas, thus satisfying the coexistence conditions in the three ecological and behavioral dimensions.

Our observations further mirror the encroaching human influence on the avian community at our study site, which may have already begun to alter the interaction and coexistence mechanisms of these species. Further investigation is encouraged into the temporal and regional variances across various landscapes of more regular members of the MSF following the fulvetta and its closest relatives. Advancing our knowledge in these topics is crucial for obtaining a more comprehensive understanding of the interactions in coexisting bird communities and for understanding the genesis of diversity and adaptability in the interactive tropical and subtropical bird assemblies.

Authors’ contributions

FZ conceived the ideas and designed methodology; ZW and WY collected the data; ZW and FL analyzed the data; ZW and FZ led the writing of the manuscript. All authors contributed critically to the drafts and all author approved the manuscript.

Funding

This study was supported by a GDAS Special Project of Science and Technology Development (2022GDASZH-2022010106), a Funding by Science and Technology Projects in Guangzhou (2023A04J0843), the National Natural Science Foundation of China (31961123003), and DFGP Project of Fauna of Guangdong-202115.

Ethics statement

Fieldwork and procedures were conducted in non-protected area under the supervision of local forestry and law enforcement departments. The experiment protocol was approved by the Animal Care & Welfare Committee of Institute of Zoology, Guangdong Academy of Sciences (Certificate No. GIZ20221101-01). No bird was harmed during the process.

Supplementary material

Supplementary material can be found at https://academic.oup.com/cz.