Modelling Catchment-Scale Responses to Climate Change - Climate Change Impacts on Freshwater Ecosystems - page 255

1995; van Breemen et al . 1998). This is illustrated in this study by the increase in
the N mineralization rate in response to the MPI and Had scenarios for air
temperature and precipitation in the Bjerkreim River Basin. Decades with elevated
atmospheric N inputs have increased the stores of N in soil, and empirical data have
demonstrated a negative correlation between NO 3 − leaching rates and the C : N
ratio of soil organic matter (Gundersen et al . 1998; MacDonald et al . 2002).
Even given current legislation on reducing N deposition, the MAGIC model
simulated a slight decrease in the C : N ratio of soil organic matter in the Bjerkreim
catchment by 2100. This implies a gradual increase in NO 3 − leaching rates even
with no climate change. With the MPI and the Had climate scenarios, the decrease
in C : N ratios and thus increase in NO 3 − leaching rates were more pronounced,
offsetting the gains from reductions in N deposition. The extent to which this
might happen depends on several uncertain factors. Among these are (i) the size
of the N pool available for mineralization; (ii) the amount of additional carbon
sequestered due to climate change and thereby affecting the soil C : N ratio and
(iii) the actual temperature responses of the various N sink and source processes.
Additionally, there are still large uncertainties associated with future NO 3 −
leaching as a response to decreasing C : N ratios in catchment soils. The empirical
model of Gundersen et al . (1998), which is included in MAGIC, is based on a
large spatial data set, and we currently lack sufficient long-term data on C : N
ratios and NO 3 − leaching to confirm the model on a temporal scale.
Overall, this study has shown that the results of linking different models can
lead to helpful insights about the effects of climate change on water flows and
nitrate leaching in catchments, which are not necessarily clear from running the
models alone. Though the results are uncertain and not to be taken as a firm
forecast, they highlight possible outcomes and also suggest future research that
could be used to improve the model predictions.
Uncertainty
All model predictions are uncertain to some degree, and quantifying and
preferably reducing uncertainty is one of the priorities for climate change
modelling (e.g. Wilby 2005). The meteorological models used to drive climate
change predictions are themselves uncertain, and this can have major impacts on
predictions of water quality. For instance, Whitehead et al . (2006) and Wilby
et al . (2006) used three different GCMs to predict changes in flow in the river
Kennet for 2050: the results ranged from a 19% decrease to an 80% increase. To
add to these, there are the uncertainties associated with water quality modelling
structures, parameters and observations.
Though a good model can be calibrated to observed data, its predictions are
rendered uncertain because of doubts over whether the model structure and
parameters are good representations of the modelled system. It may be possible
to calibrate a large number of different structures and parameters to fit the
observed data equally well, but these differences lead to very different outcomes
when projected into the future. This is known as the 'equifinality problem'
(e.g. Beven 2006). There are various methods of estimating and reducing the