How many Vinaceous-breasted Parrots are there in western Santa Catarina? A group of six biologists from the Ferraz Lab and Unochapecó set out to answer this question in a recent field trip to the west of the Brazilian state of Santa Catarina, from December 12th to 16th, 2015. Ours is one deceivingly simple question with far-reaching implications. A. vinacea is a threatened species with geographic distribution ranging East-West from Eastern Paraguay to the south of Bahia state, in Brazil. The species is under pressure from the twin pressures of habitat loss and illegal parrot trade. While agro-industrial expansion claims an increasing part of the landscape previously covered by old-growth Araucaria forests that provided key habitat and resources, nest-poachers steal what appears to be a growing proportion of A. vinacea’s annual reproductive output. Whatever we do to revert these dangerous trends will benefit a great deal from a robust quantitative assessment of the Parrot’s population size and distribution. How many parrots are there? And where?
Imagine the map of the Brazilian state of Santa Catarina as the profile head of a left-looking dog: the dog’s snout—Western Santa Catarina—is sniffing a large expanse of old-growth forest in Missiones, Argentina; while its head and ears, east of the snout, carry a reasonable amount of mature forest. Western Santa Catarina presents a potential connection between the largest expanse of A. vinacea habitat in Brazil, and the Araucaria forests occupied by the largest non-Brazilian population of the parrot. Furthermore, given the intensive agro-industrial activity in Western Santa Catarina, it is surprising that any parrots remain there at all. This part of the species’ geographical distribution is likely a stronghold of A. vinacea resisting widespread habitat loss, and an important connection between relatively less threatened populations. If one wants to develop a counting methodology that will start small and gradually expand over the species range, Western Santa Catarina is a good place to start.
Ours is not the first attempt to count Vinaceous-breasted Parrots. A team of researchers led by Jaime Martinez and Nêmora Prestes, at the University of Passo Fundo, carried out two global counts of A. vinacea in 2014 and 2015. Three of us participated in the global counts and we all look forward to contributing in the future. The global count has been a very worthwhile effort, reaching a population estimate of 3,133 A. vinacea for the entire distribution range in 2015 (N. Prestes, personal communication, May 21, 2015). Our goal with the Western Santa Catarina counts is to develop field and analytical techniques for measuring uncertainty about the population estimates. The number 3,133 is much better than nothing, but we know that this is not a final exact number; there could be more birds or even a few less. In order to infer temporal changes in population size—to monitor the population and the impact of management decisions—we need a quantification of uncertainty about the final number. Western Santa Catarina will provide a case study of the statistical estimation of A. vinacea’s population size.
From our experience during this first count there are two main sources of uncertainty: uncertainty about the number and location of roosts, and uncertainty about each known roost count. What is a roost? A. vinacea breeds approximately between July and December, with reproductive adults living in pairs dispersed throughout the breeding habitat. When the juveniles leave their nests around the end of the calendar year, the parrots start congregating every night in roosts, which they use until the beginning of the next breeding season. Roosts are the place to count parrots because they allow us to see many birds with minimal geographical coverage. As of now, we know of two roosts and suspect the location of another two in Western Santa Catarina. There maybe more, and this is the first source of uncertainty about our counts. So far, our strategy to address this problem is to interact as much as we can with local people in search for clues about the location of roosts that are unknown to us. While we search for more roosts, our estimates refer only to the number of parrots in known roosts.
The second source of uncertainty relates to the count of birds at a given roost. This is at once a topographic and a human problem. We count parrots at a roost by deploying three observers in strategic locations around it. The locations are selected to cover as much as possible of the flyways in and out of the perceived area of parrot concentration. Observers are equipped with radios so they can tip each other about passing birds and double check their observations. Because each roost has its own topography, some roosts are easier to count than others; there are places where it is difficult to cover all the flyways with only three people. The inevitable result is that the best count by a team of three observers can be an underestimate—if some parrots pass unseen—or an overestimate—if some parrots are double (or triple) counted. The human part of the problem is that different individuals may obviously count different numbers of birds when looking at a moving flock distributed over three-dimensional space.
To address the uncertainty about each roost’s count, we deploy two teams of three observers simultaneously. Each counting point will then have two observers (one from each team) with enough distance between them to make sure they do not interact and do not hear each other’s radio communications. Teams meet separately at the end of each count to compare observations and agree on a final result; observers from different teams do not exchange information about their counts before the final team result is agreed upon and written on a spreadsheet. The final results consist of team- and roost-specific counts, with every count refering to a unique morning or evening session, i.e. we do not combine counts from different sessions. In the analysis stage, we will consider each team’s count as a binomial sample from a Poisson distributed roost population size and will model counts with a Binomial mixture model. As a plan B, we also note down each team’s count as a fuzzy number; if, in the end, we do not have enough data to fit a Binomial mixture model, we will use fuzzy algebra to sum across dormitories and obtain a Western Santa Catarina number.
At the latest tally, our December trip returned a count of approximately 250 parrots. We plan to keep sampling monthly until July 2016 to produce a 2016 pre-breeding estimate with a quantification of uncertainty. From our recent experience, it appears that the higher the number of birds in the dormitory, the bigger the difference between teams; but so far, the counts from a given dormitory and session always fall within the same order of magnitude.
If you like this trip report, please send us feedback. This work is part of Viviane Zulian MSc. research project and we are just at the beginning of data collection. We will be very thankful for any comments that help us find mistakes in our reasoning or improving the counts. Also, if you know of an A. vinacea roost in Western Santa Catarina, we would love to hear about it. Viviane even set up a website (in Portuguese) to gather information about roost locations and counts. You can check it out at: http://vivianezulian.azurewebsites.net. Please do keep checking this news page for information about upcoming counts. Your attention and feedback are important to us.