www.ccsenet.org/cis Computer and Information Science Vol. 3, No. 3; August 2010 Published by Canadian Center of Science and Education 217 Evaluation of a Stochastic Weather Generator in Different Climates Behnam Ababaei (Corresponding author) PhD Student, Irrigation and Drainage Engineering Member of Young Researchers Club of Islamic Azad University Science and Research Branch of Tehran, Iran Tel: 98-21-4413-1066 E-mail: Behnam.ab@gmail.com Teymour Sohrabi Professor, Department of Irrigation and Reclamation Faculty of Agricultural Engineering and Technology, University of Tehran, Karaj, Iran E-mail: Tmsohrabi@yahoo.com Farhad Mirzaei Assistant Professor, Department of Irrigation and Reclamation Faculty of Agricultural Engineering and Technology University of Tehran, Karaj, Iran E-mail: Fmirzaei@ut.ac.ir Bakhtiar Karimi PhD Student, Irrigation and Drainage Engineering Member of Young Researchers Club of Islamic Azad University Branch of Kermanshah, Iran E-mail: Bakhtiar.Karimi@gmail.com Abstract Stochastic weather generators are used in different studies which often require long series of daily weather data for risk assessment. They can produce synthetic daily time series of any length. Any generator should be tested to ensure that the synthetic data is proper for the purposes for which it is to be used. The main objective of this paper is to test a stochastic weather generator, LARS-WG, at 65 sites in Iran chosen to represent different climates. Statistical tests were carried out to compare characteristics of the observed and synthetic weather data such as, the lengths of wet and dry series, the distribution of precipitation and the lengths of frost periods. The LARS-WG generator uses complex semi-empirical distributions for weather variables and tended to match the observed data well, especially in terms of the daily distributions and the mean monthly values, although there are certain characteristics of the data that the generator could not reproduce accurately, for example the monthly standard deviations. LARS-WG model showed different performance in different climates and stations. Therefore, evaluation is strongly recommended if it is going to be used in different climates and stations. Keywords: LARS-WG, Weather generator, Evaluation, Different climates, Iran 1. Introduction Weather generators are models that replicate the statistical attributes of local climate variables but they don’t reproduce observed sequences of events (Wilks et al., 1999; Wilby et al., 2004). There are several reasons for the development of stochastic weather generators and for the use of synthetic weather data instead of observed. The first one is to generate weather data time series long enough to be used in a risk assessment in hydrological and agricultural applications. Observed daily weather is one of the major inputs into mathematical and agrohydrological models, but the length of the time series is often insufficient to evaluate the probability of extreme events. Moreover, observed time series represent only one ‘realization’ of the climate, whereas a weather generator can simulate many ‘realizations’ and then, a wider range of feasible situations. The second reason is to provide the means for extending the simulation of weather to locations where observed weather data is not available by interpolating the parameters of a weather generator between sites using an interpolation