Parrots (order Psittaciformes), are key species in tropical ecosystems around the world. They are important determinants of plant species richness (through seed dispersal and predation) and may play a role in pollination of trees. However, many species are critically endangered because of anthropogenic impacts such as habitat loss or degradation, or through the illegal pet trade. Zoos and other captive collections therefore play a vital role as ‘arks’, preserving the species (both its genome and the unique adaptations that are fundamental to its success in the wild), and complementing in situ conservation efforts.
However, parrots are challenging to keep in captivity. In the wild, many species forage over large areas, exploiting ephemeral and diverse food sources, and must learn to identify and process many food types, searching for food items in complex canopy environments. They are generally long-lived species, reproducing slowly, and pair bonding depends on individuals learning appropriate social behaviour as juveniles, and then finding a compatible mate once mature. Breeding success is often low, and with a small pool of potential mates, some individuals never pair bond. Thus, there are two main problems: 1) the challenges and opportunities provided by natural environments are difficult to replicate in captivity; 2) many captive collections lack capacity to collect behavioural data over long time courses. This is necessary to provide the evidence base upon which modifications can be made to enclosures or management practices to enable wild-type behavioural profiles.
This project will consist of two main components addressing these challenges: 1) joining an existing team to extend work already in progress to determine how to enable wild-type behaviour in parrot species, with a specific focus on how to facilitate wild-type social behaviours and social structure in captivity; 2) developing and validating a novel process of AI monitoring of parrot behaviour through video, using neural networks which deliver pose estimation, from which simple behavioural measures will be automatically extracted and monitored. This process will not replace in-person behavioural monitoring but will provide a large evidence-base of activity and social association, enabling zoo resources to be focussed on the most important needs identified.
We expect candidates to have a Merit or Distinction at MSc level in a relevant subject. Some experience (through a formal qualification or self-taught) of programming is essential. Experience of collecting behavioural data on birds would also be an advantage, and field work experience would be desirable. You can find information about our parallel project on great apes here, and a talk on our previous work on parrots here.
Methodology:
- Literature review of the behavioural ecology of the chosen parrot species, focussing particularly on social behaviour and social structure, but also including aspects of cognition and locomotion.
- Field work in zoos to develop methods and quantify how the behaviour of parrots differs from that of wild conspecifics. This will employ observational protocols to produce a baseline dataset and to identify the key indicators which will be used in point 3.
- Development, training and programming of the AI process to deliver automated analysis of video footage of a parrot enclosure and extract simple behavioural measures. This will involve annotation of videos to place markers, training the network, testing it on new videos, and deriving simple quantitative behavioural measures from the location/pose information produced (using Python/R).
- Validation/evaluation of the AI process compared to observational protocols.
- Using data from 2-4, modify enclosures and compare behaviour pre- and post-modification.
Training and skills:
Students will be awarded CENTA2 Training Credits (CTCs) for participation in CENTA2-provided and ‘free choice’ external training. One CTC equates to 1⁄2 day session and students must accrue 100 CTCs across the three years of their PhD.
Training in the specialist methodologies required for behavioural fieldwork will be provided by Chappell and Thorpe. Chappell will provide training in sampling and recording cognitive and social, while Thorpe will provide training recording movement and ecology. Chappell and external training will provide additional intensive training in the specific programming and statistical techniques required for this project (e.g. R and Python programming, linear modelling techniques such as etc). This will be required in order for the student to develop aspects of the project, through modelling the data and programmatically identifying key behavioural elements, as well as developing the AI processes.
Partners and collaboration (including CASE):
This PhD studentship is part of a broader research programme in which we are working with a number of organisations involved in conservation and rehabilitation of a variety of species. This includes zoos in the UK (e.g. Twycross, Chester, and Drayton Manor Zoos), as well as NGOs involved in ape conservation in range countries. Thus, the student will be embedded in an existing network of organisations, with the opportunity to extend the network to NGOs involved in parrot conservation. We do not currently have a CASE partner for this project, but one may develop during the course of the project.
Possible timeline:
Year 1: Literature review of the behavioural ecology of the species selected. Development of observational behavioural data collection to target key wild-type behaviours identified in this review. Training in the protocol at zoo(s). Training in techniques needed to develop AI process, initial piloting on video data collected at zoos.
Year 2: Field work at zoo(s) to collect observational behavioural (baseline) data on parrots. Final development and testing of AI process, and programmatic derivation of simple behavioural measures. Validation of AI process against observational data. Publish on this work.
Year 3: Using both observational and AI data with information from literature review to generate enclosure and management modification recommendations. After modifications, collection of post-modification dataset (observational and AI) and compare to baseline data to determine if wild-type behavioural profile of parrots has improved. Write and publish resulting papers.
Please email potential supervisor Dr Chappell ([Email Address Removed]) for more information.
Continue with Facebook
