Simulation of extreme temperature events by a stochastic weather generator: effects of interdiurnal and interannual variability reproduction



A WGEN-like four-variate (maximum and minimum temperature, precipitation and solar radiation) stochastic daily weather generator Met&Roll is used to provide synthetic weather series for models simulating crop growth and hydrological regime in present and changed climate conditions. Since impacts of climate change will be largely affected by changes in climate variability and extreme events, present climate models should satisfactorily reproduce both interdiurnal and interannual variability of the weather series and the occurrence of extremes. Three improvements of Met&Roll aiming at a better reproduction of interdiurnal and interannual variability have been introduced: (i) lag-0 and lag-1 correlations among solar radiation and daily extreme temperatures are allowed to vary during a year; (ii) a Markov chain of the third order (instead of the first order) is used to model precipitation occurrence; (iii) the synthetic daily weather series is adjusted to fit the series of the monthly means, which is generated using the four-variate first-order autoregressive (AR) model.

Model performance regarding the simulation of extreme temperature events is evaluated against observations at 83 stations covering most of Europe. None of the improvements of the generator lead to a generally better reproduction of extreme high and low temperatures (1 day extremes); the basic version of the generator performs best for annual maxima, whereas the inclusion of the annual cycle of correlations slightly enhances the simulation of annual minima. For multiday extremes, the incorporation of the monthly generator tends to improve heat- and cold-wave characteristics, mainly the chaining of hot and cold days in spells; in western and central Europe, it improves (worsens) the simulation of heat waves (cold waves). Bad reproduction of most extreme event characteristics out of western and central Europe indicates a limited applicability of a generator based on the first-order AR model for temperature outside the mild climate zone and in areas with a high degree of continentality. The validity of assumptions of the generator should be tested using weather series related to a given area before its application in impact studies, and, if necessary, adjustments must be made. Copyright © 2005 Royal Meteorological Society