Biophysical
impact models range in complexity from simple monotonic relationships
established between a single climate variable and a single type of
response, through to complex simulation models where developers have
attempted to incorporate all of the processes thought to be of
importance in determining system responses. Examples of the latter
include dynamic vegetation models and basin-scale hydrological models.
All
biophysical models rely on empirical relationships between driving
variables and system responses at some scale of analysis, but the level
of empiricism varies enormously. In process-based models, many of the
equations describing physical or biological processes are well
established theoretically and have been verified empirically (e.g.
photosynthetic processes in plants or water flow in soils). Other
processes may be less well established, and are subject to greater
uncertainty. Taken together, the description of interacting processes
allows for a deeper understanding of the behaviour of different
components of a complex system and hence a better appreciation of the
reasons for a given response of a system. However, such models tend to
be very demanding of data, expertise and time for model testing and
application, which may limit their use in different regions. Model
estimates of future impacts increasingly rely not only on climate
projections, but also on scenarios of other conditions that either
affect impacts directly (e.g. changes in atmospheric composition or sea
level) or precondition sensitivity to impacts (e.g. population, income,
land use and land cover change, technology). To assist users,
process-based models with a potentially wide application are being
packaged in user-friendly decision support systems, where users are
able to tailor the impact model to the needs of their own assessment,
being provided with detailed guidance on data collection and procedures
for model calibration and testing, as well as advice and built-in
graphical and statistical tools for the analysis and interpretation of
model outputs.
Table 3 4 identifies examples of
decision support tools| Sector | Examples |
| Agriculture |
APSIM,
the agricultural production systems simulator
DSSAT,
Decision Support System for Agrotechnology Transfer
GRAZPLAN,
four models to support decisions for grazing systems |
| Water Resources |
WEAP,
a water evaluation and planning system
RiverWare,
a general river and reservoir modeling tool
WaterGap, Water - a Global
Analysis and Prognosis |
| Biodiversity | GLOBIO3,
a global biodiversity assessment model
LPJmL, Lund-Potsdam-Jena managed
Land Dynamic Global Vegetation and Water Balance Model |
| Coastal/Marine | DIVA, Dynamic Interactive
Vulnerability Assessment, is an integrated model for assessing
consequences of sea-level rise
Roadmap, Roadmap for Adapting to
Coastal Risk |
| Multi-sector | SimClim,
the Simulator of Climate Change Risks and Adaptation
Initiatives
CLIMSAVE
IA, Integrated Assessment Platform for impacts, adaptation
and
vulnerability in Europe
CIAS,
Community Integrated Assessment System, a system of linked energy,
climate, impacts and economic models |
In contrast, at the other end of
the spectrum are simple empirical-statistical models that are based on
a statistical association between the overall response of an exposure
unit and a set of climatic predictors, without consideration of the
intermediate process that might have produced a given response. Here,
statistical associations are sought between responses to climatic
variations observed over long time periods or across geographical or
altitudinal climatic gradients (cf. the section on
impact attribution).
Impacts of future climate change are estimated by applying the same
statistical relationships observed in the past and assuming they can be
extrapolated to future conditions represented using climate scenarios.
The advantages of such models include minimal data requirements
(usually only observations and scenarios of readily accessible climate
variables) and speed of application. However, there can be major
pitfalls in relying on extrapolation of statistical relationships to
represent responses under future conditions. Consider, for example, the
effects of climate warming on wheat yield in central Europe. Simple
regression of wheat yield and temperature might reveal a negative
association between wheat yield and temperature (decreased yields in
warmer years and higher yields in cooler years). Applying such a
statistical relationship with scenarios of future warming would hence
predict reduced crop yields. However, use of a process-based model that
incorporated not only the negative effects of increased temperature on
yield, but also positive effects of future CO2-fertilization, as well
as effects of changes in soil moisture, might produce yield responses
that are quite different for a scenario with the same warming but also
increased CO2 concentration and precipitation changes.
To
conclude, analysts wishing to apply biophysical models in projecting
future impacts, whether process-based or statistical, need to consider
carefully the outcomes required from the modelling exercise. This
involves weighing their confidence in the capability of a model to
provide a reliable representation of responses to changed future
conditions alongside the simplicity of its application and possible
limitations imposed by data, expertise and computing capacity.
