Modelling Catchment-Scale Responses to Climate Change - Climate Change Impacts on Freshwater Ecosystems - page 251

were predicted. Conversely, the simulation of run-off in the neighbouring
headwater Dalby stream draining a 10.5 km 2 sandy, groundwater-dominated
catchment suggested almost no change in monthly run-off during the summer.
Other results from groundwater-fed catchments (e.g. Jackson et al . 2007) also
suggest that the effects of climate change will be apparent more quickly in systems
with impermeable geology or loamy or clay soils where the hydrology is more
responsive than in those catchments underlain by permeable geology or sandy soils
where the groundwater will buffer short-term variations. An increase in flood plain
inundation from an average of 34 to 51 days per year was predicted for the Gjern
catchment. Total N exported from the river basin increased by 7.7% even though
N retention in the river system was forecast to increase. This was more than
counterbalanced by an increase in the transfer of N from land to surface waters.

Linking models

Many specialized ecosystem models are available which give robust and detailed
predictions in a limited scientific field. With climate change, the possible
environmental impacts are diverse and interconnected and involve the responses
of a myriad of processes in whole catchments. Integrated management strategies
will be needed to tackle the problems, and to model responses on a catchment
scale individual models also need to be integrated (e.g. Andersen et al . 2006; Evans
et al . 2006). However, linking models poses a number of challenges. The use of the
output of one model as the input to the next requires that the models operate on
the same spatial and temporal scale and on about the same level of complexity.
For most applications, it is not necessary to programme hard links between the
models so that they operate as a single entity. This is difficult as individual models
were rarely designed with this in mind, and input and output data formats are
unlikely to be compatible. Nevertheless, it is essential for some applications such
as those involving multiple model runs, for example, Monte Carlo analysis (see
“Uncertainty” Section) or the creation of response surfaces. Manual adjustments
of the outputs of one model so that they are suitable as inputs to the next allow
independent models to be linked reasonably easily but are time-consuming and
can involve considerable skill and scientific judgement. Much more work remains
to be done, particularly so that the aspirations of the Water Framework Directive
for integrated resource management over whole river basins can be addressed.
An example of the approach and potential of linking models to simulate processes
at the catchment scale is provided by the study of Kaste et al . (2006) on the
Bjerkreim catchment in Norway.
The Bjerkreim River Basin (685 km 2 ) has an average run-off of 2430 mm yr −1
and discharges into an estuarine area (58°28

′ ′Ε

) near Egersund in south-
western Norway. The land cover is dominated by non-forested, mountainous
areas (~60%) and is typical of the inner south-western parts of Norway. Water
surfaces, peatlands and heathlands make up about 20% of the area, while forests
and agricultural land cover 15% and 5%, respectively (Kaste et al . 1997).
Nitrogen deposition in the Bjerkreim area is the highest in Norway, 15-23 kg N
ha −1 yr −1 (wet + dry) due to both high precipitation rates and relatively high N

N; 5°59