Long-term monitoring of a hyacinth macaw Anodorhynchus hyacinthinus (Psittacidae) roost in the Pantanal, Brazil

The hyacinth macaw Anodorhynchus hyacinthinus is the world’s largest parrot species and is currently listed as Vulnerable by the IUCN. This species commonly flocks in collective roost sites. We monitored a group of hyacinth macaws that has used a single roost site on a ranch with traditional, extensive cattle management in the Pantanal, Brazil, for over 50 yr. We analyzed 15 yr of monitoring data on the use of this roost site. We used simultaneous counts (n = 37), where individuals flying to the roost site were counted, and also counted pairs in nests. The results indicate that in 2001 there were 234 (mean no. of ind.; 95% CI = 55) macaws in the roost; 15 yr later 708 (95% CI = 142) were registered. The highest number of macaws registered was in the years of 2013 and 2015, when >1000 ind. were observed at the ranch. The model showed an increase rate of 26 ind. yr−1 over the 15 yr monitoring period. Temporal and seasonal fluctuations were ob served, with the highest number of macaws recorded during rainy and non-breeding seasons. We discuss these results with an emphasis on the type of ranch management that favors the maintenance of hyacinth macaw food resources. The type of traditional cattle management used at the study site benefits both cattle production and macaw conservation due to positive interactions between cattle feeding habits and landscape-level management practices that preserve macaw habitat.


INTRODUCTION
Historically, high levels of endemism, high biodiversity, and restricted distribution of species, at both the biome and niche scale, have been used to define priority areas for wildlife conservation (Vane-Wright et al. 1991, Kerr 1997, Watson et al. 2014).In addition to these parameters, refuges (defined as an 'area of land dedicated to the pro-tection and maintenance of biological diversity, and of natural and associated cultural resources, and managed through legal or other effective means' by Chape et al. 2003, p. 2) have been established to protect a vulnerable or threatened population from extinction.However, the effectiveness of these refuges in the conservation of minimumviable populations may be questionable because, over time, they undergo degradation within and outside of the conservation area, mainly due to the effects of fragmentation, fire, soil degradation, and human encroachment (Scherer-Neto & Toledo 2007, Taylor et al. 2012, Geldmann et al. 2013).As a result, the remaining populations may experience local extinction (Barlow & Peres 2004, Recher et al. 2009).In regions where events that result in habitat degradation (such as fires) occur frequently, local populations seek out safe areas that provide a reliable supply of food and breeding sites (Robinson et al. 2014).
Gregarious behavior in birds leads to increased protection against predators and more efficient food acquisition (Ward & Zahavi 1973, Beauchamp 1999).According to Munshi-South & Wilkinson (2006), gregarious behavior and foraging may lead to a longer life span, especially for species that aggregate in roosts (Beauchamp 1999).Some juvenile and nonbreeding Psittacidae continue to use communal roosts during the non-breeding season.During the breeding season, breeding pairs of some species may remain in cavity nests in trees and ravines, e.g.Cape parrots Poicephalus robustus in South Africa (Wirminghaus et al. 2000), red-tailed parrots Amazona brasiliensis (Cougill & Marsden 2004) and Lear's macaw Anodorhyncus leari in Brazil (Pacífico et al. 2014), and red-lored parrot Amazona autumnalis in Ecuador (Berg & Angel 2006).This behavior makes it possible to monitor groups that are representative for the whole population, which is especially important for endangered species.Our assumption is that, in regions where factors affecting habitat degradation, such as fire, agriculture/livestock, and human occupation are intense, habitual overnight roosts can become important refuges for endangered bird species, and consequently such roosts become important monitoring areas for these species.
The hyacinth macaw Anodorhynchus hyacinthinus is the largest flying parrot species in the world, measuring up to 1 m of wingspan, with an average weight of 1300 g (Forshaw 1978, Sick 1997, Guedes 2009).Hyacinth macaws are a gregarious species generally found in pairs, families, or groups that vary from 3 to 100s of individuals, mainly in feeding and roost sites, and hyacinth macaws exhibit strong nesting, feeding, and roost-site fidelity (Guedes & Harper 1995, Guedes 2009, Guedes & Candisani 2011).Because of its characteristics and behavior, the hyacinth macaw is a charismatic and iconic species for biodiversity conservation (Guedes & Candisani 2011).It has a wide geographic distribution, and over 95% of the population occurs in Brazil.The population is distributed in 3 non-contiguous areas: (1) the Pantanal: Mato Grosso do Sul and Mato Grosso (see Fig. 1); (2) the northeastern region: Maranhão, Bahia, Piauí, Goiás, and Tocantins; (3) the Amazon region, in the states of Pará and Amapá (Guedes et al. 2008, Presti et al. 2015).Recently, this species has also been reported in the state of Amazonas (Barreiros & Gomes 2010).Despite their wide distribution, Guedes et al. (2008) estimated the total wild population of hyacinth macaws at approximately 6500 individuals.
Due to large-scale captures for the wildlife tradeattractive for size, color and tolerance to human presence -and habitat degradation, including deforestation and fragmentation, the hyacinth macaw was included in the Convention on International Trade in Endangered Species (CITES) Appendices I and II, and has been listed as Vulnerable by the International Union for the Conservation of Nature since 1988 (IUCN 2014).In the last 25 years, there has been a significant increase in scientific knowledge about the genetics, nests, eggs, and chick management of this species, as a result of studies carried out by the Hyacinth Macaw Institute (Guedes & Harper 1995, Faria et al. 2008, Allgayer et al. 2009, Guedes & Candisani 2011, Marchesi et al. 2015, Presti et al. 2015).However, little is known about population fluctuations of hyacinth macaws and dynamics of the use of overnight roosts.
In July 1999, the World Wildlife Fund of Brazil (WWF-Brazil) carried out a workshop in the State of Mato Grosso, focused exclusively on hyacinth macaw conservation strategies, which included a visit to the São Francisco do Perigara ranch located in the Barão de Melgaço sub-region, in the northern Pantanal, Brazil.This ranch was chosen because it is the site of a traditional roost which has been used by hyacinth macaws for over 50 yr, and because of the abundance of the species in the area.According to censuses performed by the Hyacinth Macaw Institute, approximately 15% of the total population and 20% of the Pantanal population frequents the ranch (Guedes & Candisani 2011).The presence of macaws at the ranch, and the finding by specialists that it could be one of the largest roost sites ever recorded, gave rise to the present research.Our hypothesis was that sustainable ranch management, which ensures food availability and protection against anthropogenic impacts, can result in an increase in the number of hyacinth macaws.Therefore, we aimed to quantify and study temporal fluctuations and seasonal changes through long-term monitoring of the number of hyacinth macaws, in order to understand the dynamics of roost-site use by the species.
Average cumulative rainfall in the area ranges from 1000 to 1500 mm.during the rainy season (November− April), and is less than 200 mm during the dry season (May−October).The average temperature in the region ranges from 21°C in the coldest month (July) to 29°C in the hottest month (January) (Alvares et al. 2013).During the rainy season, approximately onethird of the ranch is flooded, and thus inaccessible, due to overflow from the São Lourenço River.The other two-thirds of the ranch, including the ranch facilities area, is located on higher ground unaffected by flooding.
The vegetation in this region consists predominantly of savanna with dense, tall woodland (cerradão) and riparian forests along the Piquiri, São Lourenço, and Cuiabá rivers (Allem & Valls 1987).The ranch landscape consists of a mosaic of forests and associated environments containing several types of palm trees, especially the acuri palm Scheelea phalerata and the bocaiuva palm Acrocomia aculeata, as well as natural grasslands in the 'Guatós indigenous' area, cultivated pastures con sisting of exotic species, including signal grass Brachiaria decumbens, and natural grasslands degraded by human land use.The latter 2 landscape types occupy less than 2 % of the ranch area.In sandy areas, there are some mono-specific vegetation types, predominantly consisting of canjiqueira Byrsonima orbygniana.The floodplain area is densely vegetated, and includes cambará Vochysia divergens and acuri and bocaiuva palm trees.In summary, nearly 32% of the study area consists of forests (cerradão, seasonal forest, and gallery forest), 35% of the area consists of open savanna, and the remaining area consists of open grasslands (wet and dry) and watercourses (Silva et al. 2000).Areas 1, 2, and 3 are hyacinth macaw roosting sites, and area 1 was the most commonly used by the macaws.Wide arrows: directions from which the macaws came to the roosting site, with the most common directions indicated by numbers 4, 5, 6, and 7; thin arrows: macaws' movements before they settled at the roosting site for the night.Asterisks: point-count stations from which the observers carried out the macaw counts.The photo (lower left) shows the hyacinth macaws feeding on fruits of bocaiuva and acuri palm trees consumed and partially digested by cattle.Photo by Luciano Candisani

Roost site
The area around the ranch buildings consists mainly of open pastures, with large individual trees and forest fragments containing bocaiuva palm trees and other fast-growing plants recovering from past deforestation (Fig. 1).There are 2 areas in this landscape that serve as roost sites for hyacinth macaws: (1) an irregularly shaped forest fragment, which is a remnant consisting of 90% bocaiuva palm trees (Fig. 1, Area 1) with additional trees such as mangoes Mangifera indica, Indian almond Terminalia catappa, baru Dipteryx alata, and Brazilian copal Hymenaea courbaril.Over the years, this site has changed structurally as hyacinth macaws have destroyed the sprouts and leaves of the palm trees while roosting in the trees overnight and using them during the day to rest and eat.As the palm trees died, the macaws moved to other sites such as isolated trees or groups of bocaiuva palm near the ranch buildings.(2) An area with dozens of 'bocaiuva' palm trees found within 5 m of the farmhouse, which hyacinth macaws and other parrot species use to roost (Fig. 1, Area 2).

Method for estimating the number of hyacinth macaws
We developed survey methods based on the species' habit of flocking to collective roost sites.In the late afternoon, the macaws fly either in pairs or in groups from different areas of the ranch to a previously chosen place (Fig. 1).We counted hyacinth macaws at the roost site over a long period to monitor the number of hyacinth macaws, and to balance long-term temporal changes with seasonal fluctuations.Counts started in 2001 and continued until 2015, totaling 37 field visits, 18 of which were performed in the dry season and 19 in the rainy season.There were 19 visits in the breeding season (August− January) and 17 visits in the non-breeding season (February−July).Each visit had an overall duration of 3 d: 1 d to search for nesting pairs, and 2 d to visit the re treat areas and conduct 3 surveys at the roost during the afternoon.The roost surveys started at 16:00 h and ended shortly after dusk (approximately 19:30 h).
Before the counts, we carried out a preliminary count of hyacinth macaws that remained in wooded places near the ranch during the day.Once the number of hyacinth macaws near the ranch buildings was counted, a pair of observers went to each count station (n = 3).Each observer team was equipped with binoculars (Nikon Action 8×40 mm), spotting scopes (Vollo VL-237), and chronometers.We selected the locations of point-count stations during the first year of sampling: these were located 800 m from the roost site and placed so as to reduce the bias of doublecounting (Fig. 1, Areas 2, 3 and 4).The observers simultaneously recorded the number of individuals (isolated or in groups) and the direction of flight.Many hyacinth macaws engage in short flights before perching in their chosen position at the roost.These short flights or movements (Fig. 1, thin arrows) were also recorded in terms of the number of individuals and direction of flight along trees (see Fig. 1, Areas: 1, 2, 3 and 4).The data were collected and standardized, taking into account entries and exits in order to avoid double-counting.For example, if 12 macaws entered the roost and 2 subsequently exited, the total number of macaws in the roost was counted as 10.
Because the macaws tended to follow the cattle, some individuals remained in areas that were more distant from the roost.For this reason, we also conducted counts in different areas of the ranch, such as retreat spots away from the area surrounding the ranch buildings, and pastures with mineral salt troughs, where cattle gathered, attracting hyacinth macaws to feed (Fig. 1, lower left).We conducted direct counts of individual hyacinth macaws at 4 of these retreat spots.We also counted macaw pairs that remained in nests, although the known number of nests was small (see Table 1).
At the end of each sampling period 3 values were obtained: (1) the mean number of individuals at the roost site registered by the team at the roost site over 3 d, (2) the total number of individuals in areas away from the roost (retreat spots with mineral/salt troughs), and (3) the number of macaw couples in the nests.Thus, the number of macaws on the ranch ('total number') was computed as the number of macaws in the roost + the number of macaws in retreat areas + the number of macaws in nests.

Data analysis
Results are reported as mean ± 95% confidence intervals (CI) for samples with n < 30.According to the results of the D'Agostino test for normality, we used a t-test, Mann-Whitney U-test, and Kruskal-Wallis (KW) H-test for small sample sizes (n < 5), following Zwillinger & Kokoska (1999).The statistical tests were used to compare data from the breeding season (August− January) and non-breeding season (February−July), as well as the dry (May−October) and rainy (November−March) seasons.All statistical analyses were conducted using GraphPad Prism v.7 software (GraphPad Software Inc.).
Precipitation is the environmental variable showing the strongest seasonality in the study region.To evaluate the effect of precipitation on the number of hyacinth macaws, we used data from a pluviometer installed at the ranch in 2002.We analyzed the relationship between the number of hyacinth macaws at the roost during each visit and the monthly mean pre cipitation using the Pearson linear correlation method.
To analyze trends regarding the number of hyacinth macaws during the study period, we used a linear regression model.The regression was conducted using the mean total number of individuals from years with > 2 visits.A linear regression model was run using total numbers obtained in the months in which the ranch was most frequently visited, April (n = 7 visits) and September (n = 6 visits), and these months represent breeding and non-breeding periods, and rainy and dry seasons.We used the 'Trends & Indices for Monitoring data' software package (TRIM version 3.53, Pannekoek & Van Strien 2001) to calculate trends over time in the studied macaw group.
Based on roost counts in the months with >1 visit, we found that the lowest mean number of individuals at the hyacinth macaw roost site was recorded in August (n = 3, 116.2 ± 61.3) and the highest in April (n = 7; 357.4 ± 185.2).However, variations in the number of macaws between months were not significant (H = 12.15; p > 0.05).

Temporal and seasonal fluctuations
Rainfall data showed that drier months to be May to August, with between 0 and 20 mm precipitation, and the wetter months to be November to April, with precipitation between 80 and 400 mm (Table 1).The number of hyacinth macaws was different (t-test, t = 2.063; p < 0.05) between the rainy season (average rainfall in rainy months = 191.7 mm; mean number of hyacinth macaws = 354.4)and the dry season (average rainfall in dry months = 36.6mm; mean number of hyacinth macaws = 254.3).This result was corroborated by the comparison between the most surveyed months, i.e.April (rainy) and September (dry) (U = 6.0; p < 0.05).Different patterns in seasonal variations were found in each year.In 2008In , 2011In , and 2015 (p < 0.05) (p < 0.05) the greatest numbers of macaws were counted during the dry period, when mean monthly rainfall was below 100 mm.
Precipitation during the study period was significantly and negatively correlated with the number of individual macaws in the roost (r = −0.64;p < 0.05; R 2 = 0.41).Breeding commences during the dry period (August).There were significantly more macaws on the ranch during the non-breeding season (mean = 319.7 ± 105.7) than in the breeding (mean = 236.9± 136.6) season (U = 80; p < 0.05; Fig. 3).

Hyacinth macaw population at the São Francisco do Perigara ranch
According to our results, the total number of hyacinth macaws recorded on the São Francisco do Perigara ranch, in the Barão do Melgaço sub-region of the Pantanal wetland, was approximately 1000 in 2013 and 2015.This corresponds to 15% of the global population of hyacinth macaws (estimated at 6500 ind.; Guedes et al. 2008) and 20% of the Pantanal population (5000 ind.; Instituto Arara Azul 2017).This is the largest concentration of hyacinth macaws observed across the species' range, and the number of macaws on the ranch increased during the 15 yr of monitoring.Our findings at this site show an encouraging and positive trend, unlike that found for many other psittacines.For example, the population of yellow-naped parrots Amazona auropalliata has declined by nearly 50% since 1980 because many of these birds are caught illegally for the pet trade (Dahlin et al. 2018).Marsden & Royle (2015) carried out a survey of density and abundance changes among 356 parrot species.Results showed that protected areas and habitat degradation, mainly conversions from primary forest to anthropogenic habitat, were positively and negatively associated, respectively, with changes in abundance and density of parrots.Studies carried out in different parts of the world show that parrot populations have declined rapidly, mainly due to a lack of protected habitat (IUCN 2014, Birdlife 2015).
In terms of annual variation, the results indicate that during periods of heavy rain, macaws need to forage more broadly due to habitat flooding.Accord-ing to Guedes (2009), during flooding in the Pantanal, there is a decrease in feeding areas for hyacinth macaws, which forage on the ground, forcing them to seek other sites in non-flooded areas on higher ground.This pattern of habitat use is further supported by the correlation between the number of individuals at the roost and precipitation data, especially during the drier years 2010−2013, during which we recorded greater numbers of hyacinth macaws.Guedes & Harper (1995) observed the same trend when monitoring a small roost in the southern Pantanal.
The hyacinth macaw breeding season occurs predominantly between the months of August and January, although breeding may occur earlier or later in some years (Guedes 2009).Our results indicate that months with lower hyacinth macaw numbers coincided with the breeding season.However, there were few natural active nests in the study area (i.e.nests containing eggs and/or chicks), which could explain the variations observed within and among years.Our results indicate that hyacinth macaws can breed in areas that are outside the limits of the ranch.A study performed using radio telemetry showed that a juvenile macaw left the nest located 50 km away from the study area and flew to the Perigara roost along with other 90 macaws (Antas et al. 2010).The low number of active nests can likely be explained by scarcity of manduvi trees with suitable nesting cavities, which are used by hyacinth macaws for breeding, and where 95% of their nests are found (Guedes 1993, Pinho & Nogueira 2003).Before the current owner acquired the ranch, the area was subject to fire and selective logging of trees.Currently, there are few manduvi trees in the area with natural cavities, and most of the available cavities are quickly occupied during the breeding season.The success of artificial nests corroborates our assumption that availability of suitable nest cavities has become a limiting factor for hyacinth macaw breeding in the study area (Guedes 1993, Martin et al. 2004, Aitken & Martin 2008).
Our assumption was that the study area was predominantly used for feeding.According to Beauchamp (1999) roosting aggregations can increase food acquisition efficiency.This idea was further supported when the vegetation characteristics of areas surrounding the ranch buildings were observed.Acuri and bocaiuva palm trees are the most common tree species in the study area.Both species make up the hyacinth macaw's diet in the Pantanal (Guedes & Harper 1995).These tree species are characterized as pioneer trees from secondary succession, and benefit from fire and degradation (Bicalho et al. 2016), which stimulate their growth and fruit production.The fruits of acuri and bocaiuva palm trees are consumed and digested by cattle, facilitating hyacinth macaw feeding (Pott & Pott 1994, Guedes & Harper 1995, Bicalho et al. 2016).

Study area management and hyacinth macaw conservation
The ranch's current management approach, with a very large area devoted to vegetation recovery, has helped to maintain food resources for the hyacinth macaws, which have used the property as a traditional feeding and roost site.In addition, with the support of Embrapa Pantanal and the Brazilian Association of Organic Cattle Ranching (ABPO), the ranch has implemented traditional extensive cattle management with a rotational grazing system in several paddocks.During the rainy season, the cattle are concentrated in a non-flooded, fenced area that favors the concentration of partially digested and excreted palm nuts that macaws feed on.As a result, the macaws gather during the flood season, following the cattle (Guedes 1993, Yamashita 1997).Hence, the availability of food resources contributes to an increase in the number of hyacinth macaws, and the cattle management dynamic helps to explain the variation observed between dry and rainy periods.
Safety was likely another important factor for this choice of roost site.The hyacinth macaws were not disturbed on the ranch because there were few employees and limited activity.There were no freeranging pets such as cats and dogs, which could disturb the macaws when they land to feed on nuts.In the Pantanal, the association between wildlife conservation and cattle ranching is evident.Ranches play a key role in biodiversity conservation by keeping areas safe from illegal hunters.Native species benefit from this protection, and many ranches have become refuges for large wildlife populations, in many cases becoming more effective than some state or federal protected areas (Silva & Strahl 1991, Hoogesteijn & Hoogesteijn 2010).For over 50 yr, the São Francisco do Perigara ranch has been protected from forest fires and deforestation, thus helping to maintain resources for hyacinth macaws, such as food and roost sites.These resources are important factors in preserving the macaws in the region.
It is important to emphasize that the São Francisco do Perigara ranch has a management philosophy that encourages the macaws to live in relative harmony with cattle ranching, and the macaw population appears to be growing, either through reproduction, immigration, or a combination of both.The traditional management practiced on the ranch takes into account the care and conservation of important areas for wild fauna (for instance, the ranch includes a fenced forest fragment that has served as a roost site for hyacinth macaws for over half a century), the protection of specific tree species, as well as the rotational grazing system over several paddocks.This strategy is consistent with the natural pasture cycle due to the rainy and dry seasons, and provides a refuge habitat for the hyacinth macaws and other Pantanal fauna species, which are also threatened with extinction, such as the spotted jaguar Panthera onca and the maned wolf Chrysocyon brachyurus.Silva and Solange Latenek.We appreciate the support and presence of Anna Croukamp (Parque das Aves).We also thank Dr. Luciano Sabóia and the Hyacinth Macaw Institute, who provided the installation of artificial nests for the hyacinth macaws on the São Francisco ranch.The ITA thanks Carlos Cézar Corrêa, Grace Ferreira da Silva, Edson Diniz Lino Pereira, Fernanda Fontoura, Luciana Ferreira, Cynthia Mazzi and Fundação Toyota.We thank Leonie Seabrook, PhD, for editing a draft of this manuscript.We appreciate reviews of this paper by Thomas H. White Jr., PhD, of the US

Fig. 1 .
Fig. 1.São Francisco do Perigara ranch, located in the southeast of the State of Mato Grosso, Brazil.(Lower right) São Francisco do Perigara ranch area, in the Pantanal, Barão de Melgaço sub-region.Areas 1, 2, and 3 are hyacinth macaw roosting sites, and area 1 was the most commonly used by the macaws.Wide arrows: directions from which the macaws came to the roosting site, with the most common directions indicated by numbers 4, 5, 6, and 7; thin arrows: macaws' movements before they settled at the roosting site for the night.Asterisks: point-count stations from which the observers carried out the macaw counts.The photo (lower left) shows the hyacinth macaws feeding on fruits of bocaiuva and acuri palm trees consumed and partially digested by cattle.Photo by Luciano Candisani

Fig. 2 .
Fig. 2. Regression models between the number (mean ± SD) of hyacinth macaws on the São Francisco do Perigara ranch and the sampling years, from 2001 to 2015.The solid line shows the distribution of the mean number of macaws at the roost site by year.The dashed and dotted lines show the regression model in April and September, respectively

Fig. 3 .
Fig. 3. Mean and 95% CI of the number of hyacinth macaws in the roost during the non-breeding and breeding seasons on the São Francisco do Perigara ranch, Brazil Fish and Wildlife Service's Puerto Rican Parrot Recovery Program, Paul R. Reillo, PhD, of the Rare Species Conservatory Foundation and Walfrido M. Tomas, PhD, Embrapa Pantanal.LITERATURE CITED Aitken KEH, Martin K (2008) Resource selection plasticity and community responses to experimental reduction of a critical resource.Ecology 89: 971−980 Allem AC, Valls JFM (1987) Recursos forrageiros do Pantanal mato-grossense.Documentos da Embrapa 8. Embrapa, Brasília

Table 1 .
Number of hyacinth macaws registered in roost, retreat areas, and nests during the surveys on the São Francisco do Perigara ranch, Pantanal wetland, Mato Grosso, Brazil, from 2001 to 2015, based on total counts.Individuals (no. of nests in parentheses) refers to active nest with eggs, or chicks, or hyacinth macaw couples defending cavities.−: only one value, no mean available (Guedes et al. 2014)t was not possible to conduct the surveys.bInstallation of artificial nests by the Arara Azul Institute (seeGuedes & Silva 2010).cSignificant differences (KW-test, H = 26.3;p= 0.015) between 2007 and 2013 (Dunn's test = -28.33)and2007 and 2013 (Dunn's test = -29.55).dSurveys conducted only of nests installed by the Arara Azul Institute(Guedes et al. 2014)