Blenkinsop, S., Fowler, H.J., Harpham, C., Burton, A., and Kilsby, C.G. Modelling transient climate change. Part 2: Weather Generator and projected changes. Submitted to Journal of Hydrology, December 2008.

Abstract

The coarse resolution of climate models is especially problematical for hydrological applications and creates the need for future scenarios which are “downscaled” to an appropriate spatial scale. Two of the key issues which remain in this field are the incorporation of uncertainty into future scenarios and the production of scenarios for the near-future as well as for the end of the century. We describe a new procedure which addresses these issues by producing a large multi-model ensemble of transient climate change scenarios.

In Part 1 of this paper (Burton et al., submitted), we describe a new hybrid dynamical-statistical transient downscaling method which is coupled to a stochastic rainfall model. This is used for the generation of an ensemble of transient future scenarios of daily rainfall for the Brévilles catchment in northern France derived from changes projected by 13 regional climate models (RCMs). Here, we describe a new and complementary procedure to provide associated transient scenarios of daily temperature and potential evapotranspiration (PET).

Monthly change factors (CFs) for mean temperature from the same 13 RCM experiments from the PRUDENCE ensemble are used to project future change for the end of the century. These are pattern scaled to produce monthly CFs for each year for the period 1997 to 2085. In conjunction with the previously produced rainfall scenarios these CFs are used to condition a stochastic weather generator calibrated on current climate conditions to produce transient scenarios of daily maximum, mean and minimum temperature. The weather generator successfully reproduces the main features of observed temperature and is then applied with a new perturbation technique in which a different set of CFs are applied to each year of the simulation rather than applying the weather generator with one set of CFs as is the case for stationary simulations. Finally, a regression-based method is developed to derive a consistent transient series of daily PET based only on predictors of temperature and the time of year.

The utility of transient temperature scenarios at a daily resolution is demonstrated by evaluating simulated time series of several temperature-based indices. The Brévilles spring is projected to experience an annual decrease of ~3 frost days per decade and a shortening of the frost season of ~5 days per decade. Conversely, an increase in the annual frequency of hot days of ~6 days per decade with a lengthening of the growing season by ~3 days per decade is projected.

By examining a large ensemble of climate change scenarios we are able to indicate the range of uncertainty in these projections. Additionally, by producing an ensemble of transient scenarios we are able to provide additional information which could aid planners in assessing the likely timescales of specific impacts of climate change, as demonstrated by an example management question.