Statistical Analysis of Global Ensemble Precipitation Forecasts

Hydrologic models are sensitive to the temporal and spatial variability of precipitation, including both intermittency and variability of non-zero amounts. Marginal ensemble 24-hr precipitation statistics (probability of precipitation, conditional mean of positive precipitation and coefficient of variation of positive precipitation) for different forecast lead times and seasons for the conterminous U. S. have been compared to the statistics of observed precipitation. Techniques to compensate for model biases are considered and applied to the forecasts. Measures of reliability and skill were estimated for the resulting adjusted ensemble precipitation forecasts.