Model Output Statistics
Model Output Statistics ( MOS) is a statistical technique in modern weather forecasting, which has been developed in the 1960er/1970er years in the United States. It often involves multi-linear regression equations, which are applied to numerical weather models. Nowadays MOS process find worldwide use and serve as a tool in the case of the local weather forecast.
Background
Numerical weather models such as the U.S. Global Forecast System (GFS ) or the prediction model of the European Centre for Medium-Range Weather Forecasts (ECMWF / ECMWF ) in Reading, England calculate the future state of the atmosphere at certain times by Is the determined on the basis of measurements and soundings state predicted by solving the equations for later time points. This is done to so-called grid points, with which the earth is coated reticulated. Depending on how far these grid points apart, the quality of the forecast varies especially in soil parameters: a large distance due to local inaccuracies, a tight mesh provides greater regionality.
The GFS, for example, works with a mesh size of 35 km, the ECMWF / ECMWF grid points is at a distance of 25 km. This is sufficient as a rule to predict the state of the upper atmosphere very well. The prognosis of soil parameters such as maximum temperature, minimum temperature or surface wind for a particular location, however, is not accurate enough for daily use. This is especially true when considering a spatially highly variable meteorological parameter considered ( Low in highly differentiated terrain on a cloudless night).
On the other hand, there are purely statistical weather models, but hardly provide for the period beyond 6 hours still useful results. To close this gap, MOS methods have been developed.
Operation
In MOS process is multi-linear regression equations, which serve the direct forecast results from numerical weather models ( Direct Model Output, DMO) for a particular location based on station measurements that must be present for the place to verify or change. Thus, the advantages of a numerical model with the measured values are combined on site to obtain the most accurate forecast for the respective location.
In this case, those parameters by regression predictors are called filtered out of the DMO, which have a significant influence on other meteorological parameters that describe eg the surface weather. These parameters are called predictand. In principle, any numerical weather model can be coupled with a MOS process; also several slightly different MOS methods are applied to the same model. However, a prerequisite is that the station for which a local forecast is to be created, has a minimum 18-month series of measurements.
Representation
The results of a MOS - run are usually displayed either in chart form (MOS - diagram ), which is widely used for example in the U.S. legally, or used for generation of weather symbols, as one often finds, for example, in Germany. A program of the MOS parameters generates the desired and required symbols such as cloud / sun / precipitation, wind arrow, risk of rain etc.