5.3 Reflection About the Role of Modelling
in Conservation Management
Models are by definition simplified descriptions of reality. They have to neglect
information and processes that are considered irrelevant for the modelling purpose.
It is critical to realise “that model output is not the same as empirical data, and that
modelled projections of the future contain significant uncertainties” (Hansen and
Hoffmann 2011). These uncertainties derive from lacking or unsuitable data,
measurement errors, or systematic mistakes in the data acquisition (Price and
Neville 2003; Willis and Birks 2006). But even the simplified structure of models
may result in uncertainties of results that have to be illustrated by the modellers.
Modelling the complexity of biological systems and their interaction with management is a challenging task (McKenzie et al. 2004). Natural systems are
characterised by system inherent variation. This makes it hard to identify signals
of relevant effects from background noise of usual fluctuations (Hakonson 2003).
An additional source of uncertainty derives from subjective interpretations. In an
impressive selection of models developed and applied for environmental management, Pilkey and Pilkey-Jarvis (2007) prove how the selection of input data and
interpretation of thresholds, system behaviour, as well as modelling results corrupts
their usability in management. In this context transparency of modelling work is a
prerequisite for the use of results in the decision-making context.
In consequence, modelling results can include a wide range of sources of uncertainty. This is especially true when results are built on a cascade of coupled models
in a ‘model chain’ since all models add their own uncertainty to the overall results.
An important approach to handling uncertainty in model output is the quantification of uncertainty levels in results (see Ayala 1996; Bugmann 2003; Oreskes
et al. 1994; Sarewitz and Pielke 1999). Yet, practical work with modelling results
shows that quantification of uncertainty is no easy task. This is even more so as the
validation of modelling results for complex systems, particularly with regard to
predictions about long term future developments, involves major theoretical problems (Harris et al. 2003; Oreskes et al. 1994; Oreskes 2003; Sarewitz and Pielke
1999). Even if models have successfully simulated past or present changes this does
not guarantee that they are also able to predict future changes, e.g. if the earth
climate system and its biodiversity are pushed into unprecedented conditions in the
context of climate change (Hansen and Hoffmann 2011).
As a first consequence, a cautious use of modelling results for decision-making
is recommended (Millner 2012). Uncertainty must be considered when using
modelling results for management decisions.
The gap between modelling and management issues is attributed to the different
objectives of (natural) sciences and decision-oriented management. According to
Opdam et al., the analytical and reductionist approach of scientific work is able to
provide clues on driving forces and key elements, but lacks the ability to provide
solutions and foster decisions if they are not especially tailored to do so (Koomen
et al. 2012; Opdam et al. 2009). Further problems arise from the usual procedures of
70
K. Vohland et al.
in Conservation Management
Models are by definition simplified descriptions of reality. They have to neglect
information and processes that are considered irrelevant for the modelling purpose.
It is critical to realise “that model output is not the same as empirical data, and that
modelled projections of the future contain significant uncertainties” (Hansen and
Hoffmann 2011). These uncertainties derive from lacking or unsuitable data,
measurement errors, or systematic mistakes in the data acquisition (Price and
Neville 2003; Willis and Birks 2006). But even the simplified structure of models
may result in uncertainties of results that have to be illustrated by the modellers.
Modelling the complexity of biological systems and their interaction with management is a challenging task (McKenzie et al. 2004). Natural systems are
characterised by system inherent variation. This makes it hard to identify signals
of relevant effects from background noise of usual fluctuations (Hakonson 2003).
An additional source of uncertainty derives from subjective interpretations. In an
impressive selection of models developed and applied for environmental management, Pilkey and Pilkey-Jarvis (2007) prove how the selection of input data and
interpretation of thresholds, system behaviour, as well as modelling results corrupts
their usability in management. In this context transparency of modelling work is a
prerequisite for the use of results in the decision-making context.
In consequence, modelling results can include a wide range of sources of uncertainty. This is especially true when results are built on a cascade of coupled models
in a ‘model chain’ since all models add their own uncertainty to the overall results.
An important approach to handling uncertainty in model output is the quantification of uncertainty levels in results (see Ayala 1996; Bugmann 2003; Oreskes
et al. 1994; Sarewitz and Pielke 1999). Yet, practical work with modelling results
shows that quantification of uncertainty is no easy task. This is even more so as the
validation of modelling results for complex systems, particularly with regard to
predictions about long term future developments, involves major theoretical problems (Harris et al. 2003; Oreskes et al. 1994; Oreskes 2003; Sarewitz and Pielke
1999). Even if models have successfully simulated past or present changes this does
not guarantee that they are also able to predict future changes, e.g. if the earth
climate system and its biodiversity are pushed into unprecedented conditions in the
context of climate change (Hansen and Hoffmann 2011).
As a first consequence, a cautious use of modelling results for decision-making
is recommended (Millner 2012). Uncertainty must be considered when using
modelling results for management decisions.
The gap between modelling and management issues is attributed to the different
objectives of (natural) sciences and decision-oriented management. According to
Opdam et al., the analytical and reductionist approach of scientific work is able to
provide clues on driving forces and key elements, but lacks the ability to provide
solutions and foster decisions if they are not especially tailored to do so (Koomen
et al. 2012; Opdam et al. 2009). Further problems arise from the usual procedures of
70
K. Vohland et al.