4 Discussion
An impressive variety of quantitative modelling approaches are being
used to understand and predict wildlife responses to human disturbance
– and even to address the same conservation or management problems, as
exemplified by our red fox case-study. Understandably, it may be
daunting for someone who is not familiar with modelling, or who is only
familiar with a specific family of models, to decide on which modelling
approach(es) to use. We hope that this manuscript can provide guidance
and broaden horizons for those wishing to model the impacts of human
disturbance on wildlife. Broadening our perspective of what constitutes
disturbance can open ecologists up to other research areas and
approaches. For example, some ecologists may not consider disease to be
a human-mediated disturbance, since it is a part of the evolutionary
history of most organisms, but it is also affected by human-mediated
environmental change. Approaches used to model disease impacts, and
their associated considerations, may thus provide insights into the
modelling of other disturbances that may act via similar mechanisms. A
broader perspective can also expose ecologists to some of the ways that
modelling approaches can be integrated to overcome the limitations of
the different singular approaches.
It is interesting how different processes are being pulled into models
to address different disturbances. This was illustrated well for red
foxes, where dispersal and immigration were explicitly incorporated when
assessing the impacts of regulatory management, since these processes
appear to drive the difficulties in regulating red fox abundance.
Similarly, spatial dynamics, movement, and dispersal were key processes
included for predicting disease transmission, because movement of
infected individuals drives disease spread. Yet there remains an
opportunity to incorporate other processes that impact population
responses to human disturbance, including eco-evolutionary feedbacks and
sociological processes. Since evolutionary changes following disturbance
may enable populations to adapt to disturbances, such as increasing
temperatures, eco-evolutionary feedbacks play a crucial role in the
long-term responses of populations to disturbance (Loeuille 2019).
Social dynamics can also play a key role in population dynamics and
although they are sometimes considered for social wildlife (e.g.,
Brandell et al. 2021; Grente et al. 2024), the human
element is often neglected. Not only are human disturbances inherently
driven by human behaviour, but so are the perceptions of management
actions (Bro-Jørgensen et al. 2019). Such interdisciplinary
approaches are already well developed (e.g., Dobson et al. 2019)
but greater uptake in disturbance ecology modelling could improve
conservation and regulatory management outcomes.
We were pleased to see increasing use of energetic modelling
applications across a range of quantitative approaches. Since energetics
act in a summative way and many non-lethal disturbances have impacts on
energy acquisition or use, energetic models can help us better
understand the impacts of multiple stressors on wildlife populations.
Subsequently, the inclusion of energetic mechanisms represents an
important avenue to scale individual-level responses to population-level
impacts. By integrating models, energetics can be incorporated into most
of the quantitative approaches available for modelling population
responses to human disturbance. Perhaps most importantly, energetic
models may offer the only viable way both scientists and decision-makers
can anticipate the impact of demographic responses to complex patterns
of global climate change that will surely continue for the foreseeable
future. If we do not accommodate changes in behaviour, space use, and
demography that will accompany the relentless change in climatic
drivers, even the best of models will only have value for understanding
the past rather than predicting the future.
While DEB theory is not new (Kooijman 2000), it appears to be having a
resurgence to address the growing concerns of human-mediated threats to
wildlife. Given the common challenge of data availability, the DEB
approach may be more attainable than traditional energetic approaches
that require detailed energetic data that may not be readily available.
This is one of the reasons why we explored this avenue in our European
mink case-study. However, traditional energetics approaches can also
utilise data from proxy species and allometric relationships, and may be
more intuitive to some ecologists. The intricate tie between individual
behaviour and energetics makes these a powerful combination for
assessing population responses to disturbance.
Despite the developments and integrations of quantitative modelling
approaches, models can only ever be a simplified representation of
natural systems. All quantitative approaches rely on assumptions,
imperfect data, and simplifications of the processes they aim to
represent. Consequently, there is a great deal of uncertainty associated
with input parameters and data, model structure (Refsgaard et al. 2006), and resulting model predictions (Rounsevell et al. 2021).
Several approaches have been developed to minimise or quantify the level
of data or model uncertainty. For example, sensitivity analysis aims to
determine the influence of uncertain parameters in model outputs, which
can highlight priority areas for data collection or model development
(Cariboni et al. 2007). Alternatively, ensemble modelling aims to
reduce model uncertainty by combining predictions from multiple models.
This approach has become more common for species distribution modelling
(Hao et al. 2020) but has received criticisms due to the
‘smoothing out’ of model outputs. Challenging models with different
assumptions against each other (i.e., robustness analysis (Levins 1966)
or model intercomparison) is another approach to quantifying uncertainty
in model outputs, as well as investigating the influence of different
model structures. With this approach, there is no averaging of model
outputs, but the resulting model predictions may be vastly different.
Quantifying uncertainty brings about its own challenges. When providing
predictions to managers, it is important to highlight the degree of
uncertainty in model predictions, yet uncertainty can make it more
difficult for managers to make decisions. This makes knowledge transfer
crucial, as stakeholders are able to conceptualise uncertainty in model
outputs and make informed decisions when clearly communicated (Mahevas
& Sigrid 2024). Nonetheless, the acceptance of modelling tools as a
decision support tool depends on whether different stakeholders agree
with the representation of the system and their understanding of its
components. For this purpose, participatory modelling is a widely used
approach (Voinov & Bousquet 2010) that aims to increase and share
knowledge of a system and its dynamics under different conditions and to
anticipate the impact of management actions to support decision-making.
However, the involvement of stakeholders is no guarantee for the
appropriation of model results, especially if their participation is
limited.