Model output statistics

Model Output Statistics (MOS)

Model Output Statistics, generally refered to as "MOS", is an omnipresent statistical technique that forms the backbone of modern weather forecasting. The technique pioneered in the 1960's and early 1970's is used to post-process output from numerical weather forecast models. Generally speaking, numerical weather forecast models do an excellent job of forecasting upper air patterns but are too crude to account for local variations in surface weather. Pure statistical models, on the other hand, are excellent at forecasting idiosyncriacies in local weather but are usually worthless beyond about six hours. The MOS technique combines the two, using complex numerical forecasts based on the physics of the atmosphere to forecast the large scale weather patterns and then using regression equations in statistical post-processing to clarify surface weather details. The accuracy is generally far better than either a pure statistical model or a pure numerical model (NWP).

Leading MOS scientists

MOS has a very small field of practicioners. The most often cited pioneers are Glahn, Lowry, and Klein. Currently, most work in the field is done from the Meteorological Development Lab (MDL) [1]. Significant work has been contributed to the field by Murphy, Dallavalle, Mass, Vislocky, Fritsch, Allen, Hughes, and Antolik.

Current Operational Forecasts

MOS is run operationally using millions of equations from the MDL. The latest forecasts from various "parent" models can be found at [2]