Home / Wiki / M / Model Output Statistics (MOS)

Model Output Statistics (MOS)

Trading and SpeculationUpdated 16 Nov 2026

Model output statistics (MOS) is a technique for improving weather forecasts at specific locations. The raw output of a numerical weather prediction model describes average conditions over large grid boxes and carries systematic errors. MOS compares several years of a model's past forecasts with what was observed at a station and fits equations, usually by regression, that turn new model output into a forecast for that station. The method was developed at the US National Weather Service in the early 1970s, and the agency has long issued MOS guidance for temperature, wind and the chance of rain.

Station forecasts of this kind are the starting point for pricing a weather derivative or a temperature event contract, which settle on readings at named stations. Traders judge each source by its mean absolute error. In the United States, classic MOS products are giving way to the National Blend of Models.

Last updated 16 Nov 2026