Standardized Anomaly Model Output Statistics Over Complex Terrain
Probabilistic weather forecasts are typically provided by ensemble prediction systems. Due to necessary approximations and simplifications these forecasts often show systematic errors and are known to be underdispersive. This is especially true over complex terrain where a large proportion of the terrain-induced small-scale effects are not (yet) resolved by the numerical model. To alleviate these errors, statistical postprocessing methods are often applied to calibrate the forecasts. The statistical models allow to correct for both, bias and underdispersion.
This seminar talk gives an introduction to the standardized anomaly model output statistics (SAMOS) approach. SAMOS is a statistical postprocessing procedure which allows for high-resolution spatio-temporal corrections of ensemble forecasts. Observations and ensemble forecasts are transformed into standardized anomalies by removing the long-term climatological mean and dividing by the corresponding climatological standard deviation. This removes all site and season dependent characteristics from the data and allows to estimate one single regression model for all stations at once on the standardized anomaly scale. As the regression coefficients do not depend on a specific location SAMOS allows to create spatial corrected probabilistic forecasts for any arbitrary location within the area of interest.
Beside the introduction to the SAMOS concept two applications are presented. The first application demonstrates the use of SAMOS to create high-resolution probabilistic snow forecasts for an alpine area in Austria using ensemble forecasts and hindcasts from the European Centre for Medium-Range Weather Forecasts (ECMWF). The second application shows very preliminary results on an application using SAMOS to correct sub-seasonal temperature and precipitation predictions over the western United States.