AbstractThis thesis represents a computational investigation of ecological community dynamics and
structure. The communities are simulated using an individual-based model, and their
properties are studied using a range of ecological metrics relating to diversity, network
structure, and stability. A significant portion of the thesis focuses on community responses to two
different types of habitat loss: random and contiguous. Particular attention is paid to the strength
of species interactions, which are found to drive variability in population dynamics and also to
mediate the effects of habitat loss. The modelling framework involves several features that make
this a novel treatment of the subject. Specifically, the model is spatially explicit, multi-trophic,
and the behaviour of individuals is constrained by bioenergetic parameters. Furthermore, the
communities consist of two types of interaction: mutualsim and antagonism.
Random habitat loss is found to reduce the temporal variability of population dynamics by
reducing species interaction strengths. At the same time, communities are observed to become
more even, in terms of species abundance and spatial distributions. Under contiguous habitat loss
communities become more variable, which is associated with an increase in interaction strengths.
However, when subject to a high rate of immigration, communities under contiguous habitat
loss do not display significant changes in diversity properties, or network structure. Community
responses to habitat loss are seen to depend on the spatial structure of the landscape, with
random habitat loss providing barriers to the motion of individuals. Immigration also emerges
as a key mechanism in driving community structure and dynamics. At high immigration rates
species do not go extinct.
Community dynamics under variable immigration rates are studied in detail. Closed com-
munities (without immigration) display poor persistence, with most non-basal species going
extinct. High immigration rates are seen to promote community stability in several others ways.
Specifically, high immigration reduces temporal variability in species dynamics; increases the sta-
tionarity of species long-term abundance distributions; and reduces the signature of determinism
associated with oscillatory trophic dynamics. Differences also emerge within single communities,
with high abundance species displaying less stationary but more deterministic dynamics, while
low abundance species display the converse.
Under variable immigration rates many community responses to habitat loss are unchanged.
However, the removal of the rescue effect provided by immigration means that species do go
extinct. Also certain differences are found between mutualistic and antagonistic communities, as
well as further unexpected differences between random and contiguous habitat loss. Mutualistic
communities are seen to be insensitive to immigration rates in terms of total biomass, but display
more extinctions than antagonistic communities. Contrary to previous findings, contiguous
habitat loss produces more extinctions, while random habitat loss is found to result in trophic
collapse of communities.
Finally, a novel method for the inference of species interactions from population dynamics is presented. Interactions are inferred by fitting a generalised Lotka-Volterra model to discretely
sampled time series data. The method is tested against data generated using both ordinary differ-
ential equation models, and using the individual based model. In the case of two species dynamics
the method is shown to accurately recover interaction strengths, and in the case of three and
five species the method shows promise for prediction the demographic rates (including biomass
flows between species). However the reliable identification of interaction network topologies from
systems of more than two interacting species remains unsolved. An application of the method to
60 species systems is presented, reducing the size of the system by aggregating the dynamics
according to functional groups.
|Date of Award||27 Sep 2016|
|Supervisor||Alan R Champneys (Supervisor)|