The flight calls of Vinaceous-breasted Parrots (Amazona vinacea) lingered in the chill air of the austral winter dawn in southern Brazil. However, no parrots were seen, because the dense early morning mist had not dispersed by the time the parrots left their night-time roosts in search of food. Normally it would not matter, but in this instance a team of experienced parrot observers, assembled to make a formal coordinated count of the parrots, were thwarted in their task that morning.
This is just one of a variety of circumstances (known as sources of uncertainty) that can negatively affect the estimate of population size, not only of the Vinaceous-breasted Parrot, but also of other species of parrots and other animals. The challenge is to find the best possible way around each circumstance, to arrive with confidence at an estimate which is closest to the real population size. Due to differences in ecology and behaviour, some species of parrots can be easier to count than others. Species like the Vinaceous-breasted Parrot, which fly conspicuously to and from night-time roosts, may be easier to count but are still affected by sources of uncertainty. A recently published study* addresses these issues in relation to the total count of Vinaceous-breasted Parrots in their three home countries of Argentina, Brazil (where the majority occur) and Paraguay. These counts constitute one of various actions supported by the Loro Parque Fundación for the conservation of this species in Brazil, undertaken by the Friends of Environment Association (Associação Amigos do Meio Ambiente – AMA), and in Paraguay by the BirdLife International partner, Guyra Paraguay.
The counting of parrots is valuable because, together with size of geographical distribution, size of population is probably the most important variable in population biology as a key predictor of extinction risk. It is central to the assessment of any population management strategy and is directly implicated in three of the five IUCN (International Union for the Conservation of Nature) criteria for including species in threat categories, the Vinaceous-breasted Parrot being assigned to the ‘Endangered’ category. Vinaceous-breasted Parrots seem to be associated with the Parana pine (Araucaria angustifolia) but are also frequently observed foraging and nesting in other trees.
Furthermore, their somewhat uncertain geographical distribution extends well beyond the current extent of Parana pine forests. Unpredictable movements of Vinaceous-breasted Parrots related to food availability make it difficult to anticipate where the birds are, or whether individuals observed in different places are the same or different. The IUCN gives the existing geographical distribution of the Vinaceous-breasted Parrot as approximately 145,700 km2 but has recently updated the distribution map with a larger, ‘possibly extant’ boundary that encloses all of the existing patches.
Despite the difficulties to locate roosts and count the number of individuals, roost counts are at present the most effective way of assessing the population size and delimiting the distribution of the Vinaceous-breasted Parrot. Through repeated roost counts conducted every month, the field teams have identified the month (May) with highest attendance at the roosts and therefore best indicated for a total count. Even so, roost counts suffer from five key sources of uncertainty that need to be minimized in monitoring efforts.
Two sources are related to locating roosts, the first being the uncertainty about the extent of the Vinaceous-breasted Parrot’s distribution. When does a gap in the distribution map represent true absence of the species or absence of observations? This problem is highlighted by the difference between the IUCN ‘Extant’ (existing) and ‘Possibly Extant’ distributions. The second source is uncertainty about density of roosts at the local to regional scale. At what point should the field teams stop spending resources on finding more roosts, versus dedicating time to studying the known roosts in detail?
The third source of uncertainty refers to the possible movement of individuals between roosts. If individuals do move between roosts, then such movements somehow must be accounted for, or counts must be simultaneous which is the method used for the Vinaceous-breasted Parrot. The fourth and fifth sources of uncertainty relate to the counting technique itself, and respectively cover false negative and false positive observations of individuals. A false negative occurs when a parrot that is present at a site is not counted because it was not seen, i.e. imperfect detection. A false positive happens when a parrot is counted twice by mistake.
Vinaceous-breasted Parrot abundance was estimated for the years 2016 and 2017, comparing extensive roost counts at 59 sites over the whole distribution of the species with an intensive regional survey (designed to address five sources of uncertainty about parrot abundance) in a rectangle-shaped area of 34,000 km2 in western Santa Catarina state (WSC), Brazil. WSC is remarkable for having a surprisingly high frequency of Vinaceous-breasted Parrot sightings by citizen scientists in an area that is almost entirely (88%) outside the IUCN-defined extant distribution of the species. WSC falls within the Araucaria forest and the Interior forest biogeographic sub-regions of the Atlantic Forest, which have lost, respectively, 87 and 93% of their forest cover since the onset of European colonization.
At the end of each count, the field teams compared notes and agreed on one ‘most reasonable’ (MR) and one ‘highly conservative’ (HC) count result. A dusk MR count includes all individuals observed arriving to a roost, whereas the corresponding HC count would be lower if it includes the possibility that some previously counted individuals had left the roost undetected and returned again within sight. The method addresses the possibility of double-counting. To address imperfect detection as the other source of uncertainty, the field teams replicated their counts by working simultaneously at the same roost and time. The MR and HC regional-scale (WSC) data were analysed with a statistical model of replicated count data which accounts for imperfect detection and is updated each time more data become available. Comparison of the MR and HC results from counts of the same roost and month revealed that MR values were always higher as expected, but also less variable. Accordingly, MR results were adopted for all comparisons.
The whole-distribution counts added up to 3,888 individuals in 2016 and 4,084 in 2017 (logistical constraints prevented counting in 2017 in Argentina), the average of which is about 33% more than the highest total estimate currently given by BirdLife International. The whole-distribution count does not account for imperfect detection, and the results should be interpreted as a lower boundary for population size estimate. Brazil sites accounted for 93% of individuals in 2016, and 99% in 2017. The total number increased by 5% from the first to the second year, even though there were ten fewer sites visited.
However, including only the sites that were visited in both years the total declined by 15.6%, from 2,938 in 2016 to 2,478, in 2017. The highest number of individuals counted at one site over the two years was 364. Santa Catarina had the highest subtotal count, with more than 60% of individuals in both years, followed by Parana, with approximately 20%, and Rio Grande do Sul, with 8-10%.
To connect the regional and entire distribution methods, the researchers extrapolated the estimate of abundance in WSC to three different global estimates of population size and compared the extrapolated result with the whole-distribution counts. The regional scale (WCS) estimates of population size were 945 individuals in 2016 and 1,393 in 2017. These estimates were divided by the WSC area (km2), and the resulting density (assumed to be homogenous) was then extrapolated to three alternative areas that represent a range of possible geographical distributions: the IUCN’s “Extant” and “Possibly Extant” (about 380,000 km2) geographical distributions, and the potential range of South American A. angustifolia forests (about 175,000 km2) (Table 1).
What results is a range of possible total population size estimates, from a minimum of 3,888 to a maximum of 15,529. Evidently, an ongoing challenge is to bring closer together the minimum and maximum estimates. As the authors of the study declare, their estimates of abundance and geographical distribution are larger than those currently reported by the IUCN. However, they recommend that the Vinaceous-breasted Parrot remains in the ‘Endangered’ threat category while further investigation of population trends takes place. Their field protocols and statistical analysis to address sources of uncertainty are designed to better inform assessments of parrot population size. What has not changed is that the observers must have good field experience of the species, and also that adverse weather conditions, such as dense morning mist, will occasionally frustrate the count.
Author: David Waugh, Correspondent, Loro Parque Fundación
Photos: title – R. Tomasi Jr-AMA, 1,3 – AMA; 2 – Guyra Paraguay; 4,5 – J. Martinez-AMA; 6,7 – Zulian et al, 2018.
* Zulian, V., Muller, E.S., Cockle, K.L., Lesterhuis, A., Tomasi, R., Prestes, N.P., Martinez, J. and Ferraz, G. (2018) Addressing multiple sources of uncertainty in the estimation of parrot abundance from roost counts: a case study with the Vinaceous-breasted Parrot (Amazona vinacea). bioRxiv, p. 455774.