How complex should a model be? Inferences from a long-term study of an island reintroduction

This blog post is provided by Doug P. Armstrong and tells the #StoryBehindThePaper for the paper “Capturing the dynamics of small populations: A retrospective assessment using long-term data for an island reintroduction”, which was recently published in the Journal of Animal Ecology.

The one factor common to long-term studies is that they are never planned. Instead, they evolve serendipitously as opportunities present themselves.

Start of a long-term project

In January 1992 I arrived in Auckland to start a postdoc with John Craig at the University of Auckland, New Zealand. The idea was to conduct two bird reintroductions where we experimentally manipulated the composition of founder groups to test the effects of prior familiarity on post-release aggression, survival and reproduction. This was supposed to be done and dusted within two years, after which I was to head back to Australia.

It all started innocently enough. By April 1992 we had completed the reintroductions, of tieke (saddlebacks) to Mokoia Island and toutouwai (North Island robins) to Tiritiri Matangi Island (I now look back wistfully on the days when it was possible to plan and conduct a reintroduction within three months…). I then monitored the experiments as planned and wrote up the results. However, within a year I had acquired a new partner (Wendy) who became a new wife, had a baby (Katerina) on the way, and not independently had taken on a job at a New Zealand University (Massey) and bought a house. So the two-year projects were first extended to five years, with a new focus on longer-term population dynamics rather than post-release establishment. The toutouwai project then had several subsequent extensions until we finally stopped monitoring in January 2018, just short of 26 years since the reintroduction.

Doug Armstrong, Wendy French and 14 toutouwai heading to Tiritiri Matangi Island via helicopter, June 1993. The first 44 birds were translocated to Tiritiri Matangi in April 1993. Photo: Tony Monks.
Factors leading to long-term monitoring

So why did this project lend itself to a long-term study? It was an interesting scenario from a conservation perspective as the species was initially confined to 15 ha of fragmented forest remnants, with this forest area expected to slowly expand to about 150 ha as the island regenerated (it had reached about 33 ha by the end of the study). It was therefore a great opportunity to study the factors affecting small populations in regenerating habitats, with the advantage that we had monitored it from the time it was founded. However, the main reasons the project kept continuing were that the monitoring was fun, relatively easy, and cheap.

While toutouwai look like birds they behave like small grey dogs. Like dogs they respond violently to conspecifics that invade their territories, making them easily detected using playback, but are much friendlier toward humans. Consequently, they can be trained to fetch mealworms and quickly become addicted to these delicious bags of fat. They then generally turn up to meet the mealworm gods when we enter their territories and obligingly show us their nest sites by carrying the mealworms to the chicks. While the nests were up in trees, most could be accessed with a ladder due to the small scale of the island and fragmented nature of the forest, allowing us to band most of the population as chicks in the nest. The relative ease of this monitoring (and less repressive regulatory environment at the time) meant that previously inexperienced students could be quickly trained up then left to continue the breeding monitoring over summer, providing an invaluable learning experience – four of the co-authors on the paper (John Ewen, Wendy Dimond, Åsa Berggren, Barabara Egli) initially came to the island as such students. We also initially received free accommodation from the New Zealand Department of Conservation and free accommodation from fishermen. While costs increased over time they remained low compared to most fieldwork, and the later advent of a commercial ferry service meant the island became easily accessible.

WM-BY, one of the founding females of the Tiritiri Matangi toutouwai population, June 1993. Photo: Doug Armstrong.
Benefits of long-term monitoring

While long-term studies usually arise opportunistically, they allow inferences that are otherwise impossible. We have elsewhere (Armstrong et al., 2021a) given five reasons to consider long-term monitoring of reintroduced populations: 1) informing ongoing management decisions for the population; 2) predicting future viability of the population; 3) informing site selection for further reintroductions, 4) predicting population dynamics at other sites, and 5) improving our understanding of the dynamics of reintroduced populations. We have used our long-term data for Tiritiri Matangi toutouwai to address future viability, in particular the likely impact of inbreeding depression over the next 150 years (Armstrong et al., 2021b). However, the focus of our most recent paper (Armstrong et al., 2021c) was to improve our understanding of population dynamics – specifically to work out how complex a population model should be.

Optimal complexity for capturing population dynamics

It is widely recognized that the art of population modelling is to incorporate factors essential for capturing a population’s dynamics but otherwise keep the model as simple as possible. However, there is no simple recipe for determining this optimal complexity. Our approach was to first create a full model incorporating all factors affecting survival and reproduction, then progressively reduce the model and determine what complexity gave the highest likelihood of retrospectively replicating the observed dynamics. As expected, some degree of simplification was useful for improving that likelihood due to improved precision of predictions. However, there were two additional key results, both of which are good news. The first is that the cost of additional complexity was minor as long as the data analysis and population projections were done in an integrated framework using modern Bayesian methods. The second is that the factors that turned out to be essential (post-release depression of survival and reproduction, density dependence in juvenile survival) had been detected within the first five years of monitoring, suggesting the dynamics of reintroduced populations can be understood fairly quickly if good monitoring is put in place.

Doug Armstrong and Katerina Armstrong French monitoring toutouwai on Tiritiri Matangi Island, December 1993. Photo: Wendy French.

Armstrong, D.P., Parlato, E.H., & Ewen, J.G. (2021a). Five reasons to consider long-term monitoring – case studies from bird reintroductions on Tiritiri Matangi Island. In M. Gaywood, P. Hollingsworth, J.G. Ewen, & A. Moehrenschlager (Eds.), Conservation Translocations. Cambridge University Press.

Armstrong, D.P., Parlato, E.H., Egli, B., Dimond, W.J., Kwikkel, R., Berggren, Å., McCready, M., Parker, K.A., Ewen, J.G. (2021b). Using long-term data for a reintroduced population to empirically estimate future consequences of inbreeding. Conservation Biology, 35: 859–869.

Armstrong, D.P., Parlato, E.H., Egli, B., Dimond, W.J., Berggren, Å., McCready, M., Parker, K.A., Ewen, J.G. (2021c). Capturing the dynamics of small populations: a retrospective assessment using long-term data for an island reintroduction. Journal of Animal Ecology.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s