Weather Input for Nonpoint‐Source Pollution Models

Abstract

Three stochastic weather models are developed for use with nonpoint‐source pollution simulation models that require sequences of daily precipitation and mean daily temperature. Two simple models have parameters that can be estimated from available secondary data. Both generate temperature and precipitation independently. A third model represents a more sophisticated approach and requires primary data for parameter estimation. A model validation analysis was attempted using three U.S. weather stations. The first two models did not reproduce the higher‐order moments or the tail of the wet‐dry precipitation distribution. All models used normal temperature distributions which did not reproduce the higher‐order moments of the observed temperature distributions. Using historical and generated weather data as input to the Cornell Nutrient Simulation model, annual runoff and percolation, and dissolved nitrogen and phosphorus losses were compared. The third model gave output that was not statistically different from that obtained with the historical weather data; the first two models occasionally gave statistically different values. Using the mean annual losses given by historical weather data as a basis, model errors were less than 20%.