Daily weather data are needed for many applications such as design of hydraulic structures, studies in watershed hydrology, determination of evaporation, assessment of the fate of pollutants in soils, and execution of weather driven crop simulation models. Many applications require long periods of daily weather data to account for environmental variability. These data usually include total solar radiation, maximum and minimum temperature, rainfall, wind-run, and some measurement of water vapor in the air (Acock and Acock, 1991). In many agricultural areas, such data are either incomplete or not readily available. Records may be of insufficient length, or only monthly summaries may be available. Therefore, it is desirable to generate sythetic daily weather data to meet such needs. Reliable generated daily weather data must have similar statistical characteristics as actual weather data for a given area.
Several models have been proposed that reproduce stochastically generated weather data from long periods of existing daily data (Bond, 1979; Nicks and harp, 1980; Bruhn et al., 1980; Richardson, 1981). These models are based on sound theories supporting the generation techniques, but are sometimes not used because they are not user-friendly or not generally applicable (Richardson and Wright, 1984).
One model for weather generation that has been applied extensively in the United States is WGEN by Richarson and Wright (1984). This model generates estimates of daily precipitation, maximum and minimum temperatures, and solar radiation, and it is designed to preserve interdependence between variables as well as persistence and seasonal characteristics of each variable.
In WGEN, four parameters are required for precipitation generation, which are held constant within each month but are varied from month to month. This approach can introduce inaccuracies in generating precipitation data. The values of these parameters were determined from long records of data in 139 location in the United States. The procedure used for generating solar radiation and temperature is based on the assumption that these are weakly stationary processes (Matalas, 1967). A one-term Fourier series is used to model the seasonal variation in both temperature and solar radiation. The coefficients of the Fourier term were determined throughout the locations tested and it was found that some of the coefficients were strongly location dependent (Richardson and Wright, 1984), thus limiting the application of this model to areas where these coefficients are available.
WGEN requires long records of daily weather data to estimate parameters, limiting its use to regions of the world where sufficient data are available. Monthly summaries of weather data cannot be used to generate daily data. Parameters required for WGEN are currently available only for the continental U.S. Richarson and Wright (1984) also reported that, for some locations, there were differences in mean monthly precipitation and temperatures between actual and generated data. In some cases the cause of the differences was attributed to temporal and spatial smoothing that are inherent in WGEN.
In an effort to address some of the problems stated above, a modified version of WGEN (ClimGen) was developed by Gaylon S. Campbell (Washington State University, 1990). ClimGen generates daily maximum and minimum temperature, and precipitation from either daily weather data, if available, or from monthly summaries.