Modelling Catchment-Scale Responses to Climate Change - Climate Change Impacts on Freshwater Ecosystems - page 242

Rainfall increases slightly in the 2020s but decreases somewhat by 2050. PET,
however, increases due to rising temperatures, until in the 2050 A2 scenario it is
almost equal to rainfall. Fortunately for river flows, most of the decrease in rainfall
and increase in PET occur in summer: this means that the change in AET is much
smaller, as the soil moisture deficit will limit evapotranspiration. As the Kennet is
a groundwater-dominated river, flow will continue through the summer droughts
as at present, but low flows will become more common by the 2050s.
Though these results appear credible and self-consistent, other attempts to
model the same system have yielded a range of results. Arnell and Reynard (1997)
predicted a decrease in run-off of 21% (range +24% to −37%) in the Lambourn,
the major tributary of the Kennet, by 2050 given climate models then available
and a range of hydrological assumptions. Limbrick et al . (2000) modelled
hydrological changes in the Kennet due to climate change, using earlier versions
of the Hadley Centre models (HadCM1 and HadCM2) as climate drivers and the
hydrological model incorporated in INCA. The average reduction in annual
flows by 2050 was 19%, similar to that described here (15%-17%), as were
many of the seasonality features, such as a reduction in minimum flow of 46%
(51% here). Whitehead et al . (2006) used INCA, and the HadCM3 model as
climate driver statistically downscaled to the Kennet, to explore the implications
for flow and N concentration. In contrast to this study, to Arnell and Reynard
(1997) and to Limbrick et al . (2000), they predicted an increase in mean flow
rates of 2%-5% by 2050. Wilby et al . (2006) used statistically downscaled GCM
data and a more sophisticated hydrological model, CATCHMOD, coupled with
INCA, to study N concentration and flow. For the HadCM3 model, this predicted
a small (c. 5%) decrease in median flow in 2050 for the A2 storyline, and an even
smaller increase (c. 2%) in median flow for the B2 storyline. Both Whitehead
et al . (2006) and Wilby et al . (2006) used three different GCMs to generate
climate drivers, with strongly contrasting results. In particular, the CGM2 model
predicted large increases in mean flow by 2050 of about 35% (Whitehead et al .
2006) or 80% (Wilby et al . 2006), in contrast to the 'dry' Hadley Centre model.
Clearly, the implications of these differences are large.
Statistical models
The quantification of the relationships between the distribution of biota and
environmental conditions is a central theme of ecology, and there are existing
methodologies available for predicting the effects of climate change on species in
freshwaters. One common approach involves modelling the current distribution
of an organism in relation to habitat variables, in particular temperature and flow
regime, predicting changes in the habitat variables due to climate change, and
using the model to predict the change in species distributions. A large number of
studies of this type have been carried out (e.g. Berry et al . 2002; Mohseni et al .
2003), and the techniques being used are rapidly developing (e.g. Rushton et al .
2004). There are, however, problems with the approach. One is how to validate
and test such models (Vaughan & Ormerod 2005). Another is that different
models applied to the same data may predict significantly different impacts (e.g.
Lawler et al . 2006). One technique for evaluating such models is to use ensemble