AN ASSESSMENT OF THE WEATHER GENERATOR (WXGEN) USED IN THE EROSION/PRODUCTIVITY IMPACT CALCULATOR (EPIC)

The erosion/productivity impact calculator (EPIC) is an important planning model that is used to assess the impact of different farming techniques or changing climates on agriculture. A necessary input is a long series of daily weather which can either be observed or stochastic. The primary objective of this study is an assessment of simulated weather produced by the 'weather generator' (WXGEN), a component model of EPIC. Five strategically located Texas first-order stations were chosen for comprehensive testing, ten 30-year sequences of weather being generated for each site using data incorporated into the model as parameters. Statistics were extracted from these simulations, and were compared with similar statistics obtained from observed 1958-1987 records. More than 20 000 statistical tests were carried out, each with a null hypothesis of similarity. For almost every series of tests at the 0.05 level of significance, over 15% fell into the rejection region, indicating that sequences of weather generated by WXGEN are often unrealistic. If stochastic weather is used this possibly affects the accuracy of predictions made by EPIC. Two problem areas were identified, the exclusive use of first-order Markovian procedures, and the independent nature of wind speed and direction simulation. However, in spite of reservations about WXGEN output, it is probably the only currently available stochastic model that can deliver the necessary weather input to operate EPIC and, with some modification, has the potential to satisfy input requiremens of other types of model.