Perch selection of three species of kingfishers at the Pantanal of Brazil

.


Introduction
A habitat provides the conditions and resources necessary to support the population of a species (Whittaker et al., 1973).The amount and quality of resources vary in space and time, conditioning behavioral decisions and, consequently, their spatial distribution (Cody, 1985;Litvaitis and Villafuerte, 1996;Manly et al., 2007).Animals tend to select microhabitats that maximize their fitness (Chalfoun and Schmidt, 2012), depending on their specific needs, balancing the associated (energetic and non-energetic) costs and benefits (Hutto, 1985;Block and Brennan, 1993;Houston and Lang, 1998;Johnstone and Earn, 1999).It is expected that animals select -that is, use a resource in greater proportion than its availability-the microhabitats where they achieve higher reproductive success and survival (Levins, 1968;Orians, 1980).At the same time, animals should avoid (that is, use a resource in a lower proportion than its availability) those places where chances of reproduction and survival are low (Orians and Wittenberger, 1991).Here lies the importance of studying how and why animals select the different resources available (Cody, 1985).Moreover, to answer the question 'why do animals select a determined habitat?' it is important to learn about behavioral and evolutionary ecology so as to manage and preserve relevant habitats (Chalfoun and Schmidt, 2012).
For many species of birds, the structure and composition of vegetation and water bodies are key factors influencing the diversity and abundance of trophic resources, perches, and nesting sites (Cueto, 2006).Perches are important habitat elements for many bird species since they relate to multiple ecological functions (Cody, 1985).For example, birds use perches to defend their territory (Wiens 1969), and for reproduction (Krams, 2001), foraging (Wolff et al., 1999), feeding (Glinski and Ohmart, 1983) and resting (Thompson et al., 2022).Due to their relevance for many vital processes, selecting suitable perches may have important ecological consequences.For fisheating birds, the microhabitat use is closely related to the temporal and spatial distribution of water bodies (Ferreira, 2013).Kingfishers are an important group of continental fish-eating birds.In the Pantanal wetland in Brazil, three species of kingfisher are particularly common: the ringed kingfisher Megaceryle torquata (Linnaeus 1766), the Amazon kingfisher Chloroceryle amazona (Latham 1790), and the green kingfisher Chloroceryle americana (Gmelin 1788).They are sympatric and differ in body size, M. torquata (~39.6 cm of mean body length) > C. amazona (~29.5 cm) > C. americana (~21.8 cm) (Gwynne et al., 2010;Rodrigues et al., 2019).All three three species inhabit temporary ponds of the Pantanal, sharing habitat and diet (Donatelli et al., 2014).They mainly feed on fish in shallow water, although they may occasionally eat terrestrial and aquatic arthropods, and even lizards (Rosa, 2009).At these temporary ponds, perches serve to visualize prey and prepare the attack from a certain height (Gwynne et al., 2010).Previous research pointed to perch height as an important trait partitioning the habitat of kingfishers (Remsen, 1991;Bitterman, 2012;Chodacki and Skipper, 2019).These studies compared various traits of perches used between species but did not test habitat selection by comparing the traits of used perches with those available in the habitat.
The Pantanal is a natural reserve, holding one of the highest bird abundance worldwide (Tubelis and Tomas, 2003).In the Anthropocene, however, artificial elements, such as electricity lines, wooden bridges, and traffic signs are present even at the most isolated natural reserves.These artificial structures are thus an option for fish-eating birds to use as perches, and they are commonly observed as such (Asokan and Ali, 2010).Power lines and other artificial elements may offer good visualization of prey, free of the visual obstacles of branches, for example, normally present in tropical areas (Lammers and Collopy, 2007).However, a bird perching on a power line would also be more exposed to potential predators, and may even risk electrocution (D'Amico et al., 2018).This cost-benefit relation, however, has not been tested in analyses of perch selection by kingfishers.Our study aimed to understand perch selection of three kingfisher species and elucidate the role of artificial elements in this choice.We hypothesized that artificial elements could be selected rather than natural perches because they could enhance kingfishers' fishing success, that kingfishers would select perches with greater vegetation cover to avoid predation, and that selection of perch heights and distance from the water could contribute to reducing interspecific competition among the three species (fig.1).We did not have a priori expectations about the differences among the three species regarding perch selection.

Study area
Our study took place at the Pantanal of Miranda (Corumbá, Mato Grosso do Sul, Brazil) (19º 34' S, 57º 01' W).The Pantanal is one of the largest wetlands in the world.It covers an area of approximately 150,000 km 2 , and is characterized by flooded grasslands and savannas.The climate is tropical with marked dry and wet seasons and a mean annual temperature of 25 ºC (Bergier and Assine, 2016).We conducted the fieldwork during the dry season in September 2018.We sampled 13 temporary ponds located along the path named 'MS-184' (an unpaved road that crosses the Pantanal Matogrossense National Park at the municipality of Corumbá).The ponds have circular to ellipsoidal shapes of ca.50-100 m at their widest part at this time of the year.These temporary ponds are similar: they are well preserved, maintaining the native vegetation, and abundant in fish (Alho, 2008).Although the area is well preserved, because the path crosses the ponds, each pond holds artificial elements: a wooden bridge, power lines, and traffic signs.
As we did not capture or mark the individuals, we sampled the ponds sequentially, sampling each pond just once, to avoid pseudoreplication (Millar and Anderson, 2004).All observations were conducted in the period of maximum activity of kingfishers, between 07.00 hrs and 11.00 hrs (pers.obs.).We visited the ponds and recorded four variables at the site where an individual kingfisher was perched.We systematically recorded available perches, noting for each used perch, and the closest four available perches (unused perches).We defined the available perches as the closest potential perch in a straight line to the perch used by the bird (i.e., > 5 cm thick, enough to support individuals of all three species), either artificial or natural.We used a case/control design, pairing the data of each perch used to data of four unused systematically sampled perches, assuming they represent the perches that are available for a given individual at a given moment (Manly et al., 2007;Duchesne et al., 2010).The studied variables were: (1) perch type (artificial vs natural), (2) perch height, (3) perch distance from the water, and (4) perch vegetation cover.Due to the difficulty in approaching the perches, the same previously trained observer (Laura C. Peinado) estimated their height and distance to the water by eye.The distance to the water was measured as the horizontal distance to the pond shore, considered zero if the perch was exactly at the shore, negative if it was within the pond border, and positive if it was outside the pond border.We quantified vegetation cover as the percentage of vegetation at the angle of view of the observed kingfisher, dividing the plane into four quadrants to facilitate the estimation.We considered this view plane as the sphere above the perch viewed through a densimeter.By simultaneously recording our four variables of interest (perch type, height, water distance and vegetation cover) for each used/unused perch for each bird and using a case/ control design with the individual as a random factor, we avoided temporal and environmental variation that would confound results (Duchesne et al., 2010).We observed 34 birds using perches at these ponds: 15 M. torquata, 14 C. amazona, and 5 C. americana.Because we identified four unused (available) perches for each used perch, we sampled 134 unused (available) perches for the 34 used perches: 60 for M. torquata, 56 for C. amazona and 20 for C. americana.

Statistical analyses
First, we determined perch use of the three kingfisher species regarding the four variables of interest (perch type, height, water distance, and vegetation cover).We explored the data using univariate plots (bar plots or violin plots, depending on the variable;  see supplementary material).Data were not normally distributed, and showed no homogeneous variances between species.We tested for differences in the relative frequency of use of artificial perches (observations in artificial perches/total observations of the species) by species using Fisher's exact probability test.We tested for inter-specific differences in perch height, perch distance to water, and perch vegetation cover using Kruskal-Wallis tests.If differences were found between the three species, we used pairwise Wilcoxon rank sum tests using the Benjamini and Hochberg (1995) p-value correction to identify where the differences were.We then fitted a factor analysis of mixed data (FAMD) to reduce the dimensionality of the values and visualize perch use by the three species (Pagès, 2014) using the R package 'Facto-MineR' (Lê et al., 2008).
To evaluate how kingfishers selected their perches, we used a resource selection function (RSF) approach.An RSF is a function of the probability that an individual uses a specific resource based on its availability in the habitat (Manly et al., 2007).In our case, the resource was the perch, and we were interested in the effect of the four studied variables (perch type, height, water distance and vegetation cover) in their probability of use (fig.1).Here, we used conditional logistic regression (CLR) models to solve the RSF, pairing the data of used and available perches for each individual.In this way we could determine whether available perches were true absences that the bird could have selected at the moment of observation but did not (Jones, 2001;Duchesne et al., 2010;Benício et al., 2019).The response variable was the presence of the bird on the perch (1, used perch; 0, unused perch).In this way, using CLR we fitted a model on the odds ratio of the probability of a bird selecting a certain perch in relation to other available perches depending on the main effects of the explicative variables (artificial/natural perch, height, distance to water and vegetation cover) in the same probabilistic process (Liedke et al., 2018;Ortega et al., 2019).We used the function clogit of the package R 'survival' (Therneau, 2015;R Core Team, 2018).To test our ecological hypotheses (fig.1), we fitted six competing CLR models and selected the best model(s) based on AIC using the 'AICcmodavg' R package (Mazerolle, 2017) 5) vegetation cover model, to test whether the kingfishers selected perches to avoid predation: presence ~ veg_cover + strata (ID); Table 1.Descriptive statistics of used and unused perches of the three kingfisher species studied at the Pantanal of Miranda (Mato Grosso do Sul, Brazil).Mean and standard deviation (SD) values are provided for perch height, distance to water and vegetation cover.The variable 'type' represents the percentage of artificial perches (vs natural ones) of the total for each group.We used a paired design, with four unused perches assessed to each used perch.Sample size is 34 for used perches (15 M. torquata, 14 C. amazona, 5 C. americana) and 136 for unused (available) perches (60 M. torquata, 56 C. amazona, 20 C. americana).
Tabla 1. Estadística descriptiva de los posaderos utilizados y no utilizados para las tres especies de martín pescador estudiadas en el Pantanal de Miranda (Mato Grosso del Sur, en Brasil).En relación con la altura del posadero, la distancia al agua y la cobertura vegetal, aportamos los valores de la media y la desviación estándar.La variable 'type' representa el porcentaje de  We identified the best models as those with delta AIC c < 2 and accumulated Akaike weights up to 0.95 (Johnson and Omland, 2004).If a clear best model was observed (AIC c < 2 and Akaike weight > 0.95), we kept this model as the best one.If this was not the case, we averaged the coefficients of the best models.We assessed the goodness-of-fit of selected models by visually inspecting residuals (Johnson and Omland, 2004).

Results
The three species used perches with similar traits, that is, at a height of approximately 1-4 m, 2-3 m within the pond shore (in horizontal, see Methods section), and with 5-20 % of vegetation cover within the bird's viewpoint (table 1).Most of the used perches (78.58 %) were natural elements (tree branches), and only 21.42 % were artificial.The relative frequency of use of artificial perches was similar for the three species   Selection on CLR models indicated one clear best model, the 'height by species' model, with AIC weights 0.92 (table 2).The other five models did not fit the inclusion criteria to be selected (ΔAIC c < 2 and Akaike weight > 0.95; see table 2).Thus, the kingfishers studied did not select or avoid perches depending on their type (natural or artificial; fig.4s in supplementary material), distance to water or vegetation cover.Results from the selected CLR (fig.6s in supplementary material) revealed that C. amazona selected taller perches than those available in the habitat (table 3), while M. torquata and C. amazona did not select or avoid perches depending on their height (tables 2, 3).The probability of using a perch by C. amazona increased considerably when the perch was above 3.2 m in height (fig.5s in supplementary material).

Discussion
To our knowledge, this is the first study on perch selection of kingfishers.By using a resource selection function approach, assessing the traits of used perches in comparison with the available perches for each bird within a close temporal and spatial scale, we were able to determine the perch selection process beyond describing species' perch use.From our four hypotheses (fig.1), we found support only for hypothesis 2, about different species selecting different perch heights (probably to reduce intraspecific competition).None of the three kingfisher species showed preference for artificial perches, distance to the water, or vegetation cover.
Contrary to our expectations, kingfishers from the Pantanal showed no selection preference for artificial or natural perches.The three studied species, M. torquata, C. amazona, and C. americana, used both artificial perches (bridges and power lines) and natural perches (tree branches) in the same relative proportion as that available in their habitat.This result suggests that artificial structures (at least at this isolated natural reserve) are not altering the behavior

americana en verde) dependiendo de la altura del posadero, estimada a partir de un modelo de regresión logística (las líneas y bandas translúcidas representan los errores estándar y los puntos representan las observaciones individuales). La influencia de la altura del posadero en la presencia de las aves fue significativa solamente para Chloroceryle amazona (véase información más detallada al respecto en el texto).
Probability of presence of these fish-eating birds.Like kingfishers, raptors also use perches for hunting, feeding, and resting (Reinert, 1984).Perching reduces the energy required for hunting, feeding, and resting compared to flying and hovering (Collopy and Koplin, 1983).The fact that our kingfishers did not select or avoid perches regarding their type (artificial or natural) suggests that their energetic balance of costs and benefits would be similar.Moreover, artificial perches add to the natural perches available for kingfishers, increasing the available places to hunt or rest, and could even lead birds to reach some places that would be inaccessible otherwise.Other birds have been found to use artificial perches to save energy in their feeding behavior (Sheffield et al., 2001).There are also cases when the introduction of artificial structures modifies the behavior of birds, as with the introduction of artificial lights (Aubrecht et al., 2010;Kempenaers et al., 2010), or noisy elements (Polak, 2014).We think that the fact that the studied kingfishers did not select or avoid perches regarding their type (natural vs artificial) may reflect that both perch types offer similar access to food, minimizing the energy necessary for fishing (Resende, 2008).Furthermore, as the region is rather isolated and well-preserved (Junk et al., 2006), we consider the few artificial structures represent only a low level of alteration or disturbance in the ecology of kingfishers.
The literature reports M. torquata perching at variable or high perch height, C. amazona at low height and exposed perches and C. americana always at low perches (Remsen, 1991;Gwynne et al., 2010;Bitterman, 2012;Chodacki and Skipper, 2019).Our results show that only C. amazona selects for height, perching at highest sites available.For two species of kingfishers studied in Japan, differences in perching height, stream flow rate, and prey sizes were found (Kasahara and Katoh, 2008).Niche partition and a positive correlation between large body size of kingfishers with higher perches has been found in many studies (Willard, 1985;Bonnington et al., 2008;Kasahara and Katoh, 2008;Chodacki and Skipper, 2019).In our study, however, the mid-size kingfisher species (C.amazona) selected perches that were higher than those chosen by the largest kingfisher species (M.torquata).Individuals of C. amazona selected higher perches, while birds of the other two species did not appear to select or avoid perches based on height.Our results on perch selection do not therefore support a correlation with species' body size.Because of its concentration and abundance of wildlife, the Pantanal provides kingfishers with a diversity and abundance of prey, especially fish (Swarts, 2000;Novakowski et al., 2008;Resende, 2008).Despite this, C. amazona and M. torquata have similar diets and choose fish of similar size (Willard, 1985).Thus, it is possible that perch height segregation may relax competition between these two species.This would be the case, at least, during the dry season when our study took place.Of course, perch selection of kingfishers may change in the rainy season, when ponds are larger and deeper, and fewer prey are available.Future studies should evaluate the degree of kingfishers' perch fidelity or shift between seasons and how flooding dynamics may drive their habitat selection.
The thirteen small isolated ponds in our study are similar in depth, stream flow, turbidity, and width.Sampling was conducted in the dry season, when ponds are shallow and stream flow is low, forming rounded little ponds close to wooden bridges of the path.At these ponds, both predation risk and food availability would be similarly high for the available perches during the dry season (Holbrook and Schmitt, 1988).A perch without vegetation cover and close to the water may enhance fishing success, Table 2. Perch selection of the three studied kingfisher species inhabiting the Pantanal of Brazil.Results of the model selection analysis of the six competing conditional logistic regression models proposed to test the different study hypotheses.Bird presence (1, presence; 0, absence) is the response variable in all models and all of them are conditioned by the individual ID.
Tabla 2. Selección de posaderos en las tres especies de martín pescador estudiadas en el Pantanal de Brasil.Resultados del análisis de selección de los seis modelos de regresión condicional logística propuestos para confirmar o rechazar las diferentes hipótesis de estudio.La presencia de aves (1, presencia; 0, ausencia) es la variable de respuesta en todos los modelos y todos ellos están condicionados por el ID del individuo.while increasing predation risk.On the contrary, using a perch hidden among leaves may reduce predation risk but lower fishing success.In an environment of high availability of fishes and high predation risk such as that at the Pantanal, we would expect kingfishers to select perches with high vegetation cover.Our results, on the contrary, showed that only perch height, and not its vegetation cover or distance to the water, were selected.Further research quantifying prey availability and predation pressure could clarify why these birds neither select nor avoid perches based on vegetation cover and distance to the water, at least in the dry season.

Model
We conclude that Pantanal kingfishers did not select or avoid perches regarding artificial/natural origin, distance to water, height above ground and vegetation cover.The exception was C. amazona, which selected higher perches, probably to avoid interspecific competition with M. torquata.The high quantity of food available and the similarity in predation risk, also high at the Pantanal, may result in similar quality (balance of costs and benefits) of the available perches for kingfishers.We think the isolation and good preservation of the Pantanal (with few artificial structures and still high numbers of quality natural perches available) may have prevented artificial structures, such as power lines, from modifying perch selection for the moment.However, bushfires are increasing dramatically in the area, fueled by climate change, and it is predicted they will worsen in future years if policies and land use remain the same.More than 14,000 km 2 of native vegetation are already predicted to be lost in the area by 2050 (Guerra et al., 2020).We consider that C. amazona, which selects for higher perches, could be particularly affected by the loss of native vegetation related to bushfires.We recommend using habitat selection to estimate the vulnerability of species and to prioritize conservation areas.(Brazil).This plot illustrates the interpretation of the two dimensions: higher values in dimension 1 mean taller perches that are more probable to be artificial and show lower vegetation cover; higher values at dimension 2 mean a higher vegetation cover.

Fig. 1 .
Fig. 1.Hypotheses on perch selection of three species of kingfishers (Megaceryle torquata, Chloroceryle amazona, and C. americana) studied at the Pantanal of Brazil: birds of the three species will select artificial over natural perches (H1), perch height (H2), distance to the water (H3), and vegetation cover above the animal (H4) will affect perch selection, probably in a different way for the different species.

(
x2 -test = 2.015, p = 0.365).In addition, perches used by the three species were at similar distances to water (Kruskal-Wallis' test, x2 = 0.307, p = 0.858) and were covered by a similar amount of vegetation (Kruskal-Wallis' test, x2 = 1.936, p = 0.380).However, in the height of perches used by the three species differed (Kruskal-Wallis' test, x2 = 7.827, p = 0.020).Pairwise post-hoc comparisons revealed significant differences in perches used by M. torquata, these being taller than those used by C. americana (p = 0.02).While the other two species used a broader range of perch heights, C. americana always used perches below 5 m (fig.3).For the multivariate approach (FAMD), we retained two dimensions that explained 49.41 % of the variance in perch use of Pantanal kingfishers (fig.1sinsupplementary material).The first axis explained 29.15 % of the variance and was mainly

Fig. 2 .
Fig. 2. Multivariate analysis (Factor Analysis of Mixed Data) of the habitat use of the three species of kingfishers studied at the Pantanal of Brazil (Megaceryle torquata in blue, Chloroceryle amazona in red, and C. americana in green).The first axis was mainly related to perch origin (natural vs artificial) and the second to bird species and vegetation cover.

Fig. 3 .
Fig. 3. Probability of presence depending on perch height for the three studied species (Megaceryle torquata in blue, Chloroceryle amazona in red, and C. americana in green), estimated by a logistic regression model (lines and semitransparent bands, depicting the SE; points represent the observed individual values).The influence of perch height on the presence of birds was only significant for Chloroceryle amazona (see more details in the text).

Fig. 1s .
Fig. 1s.Scree plot for the multivariate exploration of perch use (Factor Analysis of Mixed Data) for the three species of kingfishers (Megaceryle torquata, Chlroroceryle amazona, and C. americana) studied at the Pantanal of Mato Grosso do Sul (Brazil).We retained two dimensions, that explained 49.41 % of the variance in perch use of the three species of Pantanal kingfishers regarding the four studied variables: 1, perch nature (artificial vs natural); 2, perch height (m); 3, distance to the water (m); and 4, perch vegetation cover (%).

Table 3 .
Perch selection of the three kingfisher species studied at the Pantanal of Brazil.Results from the model that best explains habitat selection of Megaceryle torquata, Chloroceryle amazona and C. americana, a conditional logistic regression model of bird presence explained by the perch height and the interaction between perch height and kingfisher species, and conditioned by the individual ID.The reference category for the species was C. amazona.