Abstract: Abstract: Climate change is important for agriculture and the environment. Changing rainfall amounts have positively or negatively impact on plant growth. The reduction in solar radiation can potentially reduce the photosynthesis, growth of plants and potential evaporation. Stochastic weather generators can generate a long series of weather variable statistics, which usually are used as the input of system models to analyze and evaluate the effect of climate on systems. This paper described a stochastic weather generator WGDWS which consisted of dry and wet spell, daily precipitation, solar radiation, and maximum and minimum temperature models. It included two types of models. The first one was a dry and wet spell model in which dry and wet spell lengths were defined as an independent stochastic variable respectively, and it was the principal model. The second one referred to other weather variables whose modeling was dependent on the first one. The generation of weather element values mainly contained two steps: generating a dry or wet spell length based on their empirical distribution in a month, and then generating the daily value of each variable in the related period. The generator could provide any length of time series of daily weather stochastic values as input data for the driving of crop models. The observed 1973 -2003 daily weather data from nine meteorological sites in different geographical region in China were used to determine model parameters of a generator in two types of generator, including WGDWS which was based on dry and wet spell and DWSS which was based on dry and wet days. After generating 100 years of daily weather variables, including total solar radiation, maximum temperature, minimum temperature and precipitation, at the above nine sites with the help of WGDWS and DWSS, monthly statistics of these variables were computed. A T-test showed that there was no significant difference between the generated and observed monthly statistics for different geographical regions at the 1% significant level. The differences of generated and observed maximum temperatures under 0.3℃ accounted for 87%, and the value for minimum temperature was 93%. The absolute errors of the number of rainy days under one day was 92%，and the monthly total precipitation errors under 10mm and 15 mm were 91% and 95% respectively. The absolute errors of the monthly total radiation under 2 kJ/m2 were 89%. The K-S test was conducted to detect differences between the observed and simulated values for dry and wet spells. No significant differences were found at P=0.05. The relative deviations of monthly statistics generated by WGDWS and DWSS were compared. The relative deviations of maximum temperature, minimum temperature, and total radiation for these two kinds of models were similar. There were obvious differences in relative deviations for the number of rainy days and rainfall. In short, WGDWS had similar performance to generate meteorological data to DWSS, and its accuracy to simulate dry and wet spells and number of rainy days was higher than DWSS. The study indicated that the data generated by WGDWS could be used as input for a crop models, especially for studies on the response of crops to persistent drought and continued rainy weather.