How it's made: Model Output Statistics (MOS): What is it & How it Contributes to Your Forecast
Welcome to the third part in the How It’s Made series - Category Three: Statistical Post Processing.
The How It’s Made series takes us under the hood of how the weather experts create your forecast. It covers the Five Categories that are part of the weather forecasting methodology.
In this article, we’re focusing on Category Three: Statistical post processing, and, in particular, we’re exploring the Model Output Statistics at the heart of accurate forecasts.
Sign-up for updates today, to make sure you don’t miss out
Model Output Statistics (MOS) is a real showpiece at the heart of creating your forecasts. The weather experts use it to add value to purchased raw weather data, by correcting model data for local influences and get closer to the actual situation.
The MOS takes account of two years of historical data and compares local observations with the model data issued. Each location has its own precise local characteristics which will not be fully reflected in a coarse-scale model. The MOS realigns the deviation and ensures that the appropriate weighting and characteristics are applied from the three main models. As such, a specific individual MOS forecast is generated for every weather parameter for every observation point.
"Our MOS system is the core of our worldwide forecasting system. Not only because it adds a lot of quality to raw-model data, also MeteoGroup’s MOS system provides the customer with specific derived weather parameters that strongly increase the usage options"
- Wim van den Berg
Senior Meteorological Consultant
The Six Step Process to Create the MOS
Post-processing is done by the weather experts to correct the quite coarse-scale nature of raw model output; these corrections are necessary to ensure that local effects are taken into consideration. The six-step process outlined below delivers the high-quality forecast of MeteoGroups’ MOS.
Step 1: Annual MOS update using historical model and observation data
There are separate MOS equations relating model output with observed weather elements (temperature, wind etc.) for every observation station and forecast step. These equations are updated every year using up to 2 years’ worth of historical (model and observation) data.
Step 2: The MOS prepares the ‘main’ and ‘derived’ MOS elements
Using the historical data it has been trained with, as well as live model data and live observations from weather stations, radars and satellites, it runs the MOS equations, the results of which are referred to as the ‘main’ MOS elements. It also creates ‘derived’ MOS elements; for example, where there are no observations available, it derives the forecast from related data. Derived data also includes probability forecasts (risk of extreme events) and a translation of the forecast into the most significant weather type.
Step 3: The MOS main and derived elements are combined to create the forecast
Every hour, the MOS forecast is updated for the coming 48 hours and 4 times per day the MG MOS updates the forecast for the coming 15 days. In total, MOS does billions of calculations per hour.
Step 4: The MOS forecast undergoes a quality check
It’s important that a quality check is performed to ensure consistency between elements. Example: when non-zero precipitation is forecasted, there must also be some clouds.
Step 5: The MOS forecast is prepared for delivery
The combined output of the main and derived elements is “packaged” for shipping to internal or customer systems.
Step 6: Forecasters make any necessary edits in MeteoBase
Experienced meteorologists adjust the MOS forecast to allow for the more extreme weather conditions. This adjusted version of the MOS forecast is called the MeteoBase.
How the MOS adds value to weather data to create a more accurate forecast
Besides by experienced meteorologists, the MOS is influenced by many different factors, including real-time observations, radar, and satellite images. It is by combining all data sources together that weather experts ensure the forecasts are in line with the current situation. This approach has two key benefits, namely:
#1 Delivering one of the lowest mean absolute errors in the market is achieved by:
- Using a unique approach, incorporating 4 weather models (for Europe, ECMWF, GFS, EURO4, and ECMWF-EPS)
- Measuring the structural mistakes in the forecast and correcting the MOS by incorporating a dynamic bias correction, using observations from the previous 5-25 days
- Input from meteorological experts with the option to edit forecasts in MeteoBase.
#2 Enabling nowcasting and dynamic adaption to the latest observations, which means:
- For weather stations, mistakes can be measured in the short-term forecast (up to 6 hours) and corrected accordingly
- Using precipitation radar data, it’s possible to include the movement of precipitation over the coming 3 hours
- Using satellites, the expected cloud pattern for the next 3 hours can be corrected.
In addition, the MOS and MeteoBase also provide the probability of weather events like precipitation, snow, and thunder. The MOS can also estimate uncertainties in the MOS forecast itself, providing a forecast of the average error in the predicted temperature and wind speed.
How do the weather experts keep the MOS accurate?
The weather experts play a vital role in putting the finishing touch to the ultimate weather forecasts to be issued. In addition to the technical and physical setup, as highlighted in Step 6 in the process to create MOS, experienced forecasters are essential to finessing the MOS via the MeteoBase system.
Additionally, over time, the purchased weather model data gradually get out of sync with the statistical calculations that are unleashed on it. Formulas and algorithms designed two or three years ago and have applied since might not be entirely sufficient today.
For example, when using a specific weather model mix, the MOS may always add two degrees to a specific area to arrive at the correct maximum temperature. However, if the purchased models are adapted to more closely approximate the correct maximum value, adding those two degrees as a correction factor will result in the temperature forecast being too high.
These fluctuations mean experts regularly need to make adjustments to ensure the MOS is still accurate. And, at times, can mean a complete overhaul to ensure it remains accurate and reliable.
At its core, the MOS and MeteoBase data are used to create your weather forecast and help identify situations where the weather poses a risk.
Data from the MOS underpins other statistical methodologies like Road & Route models, for road surface temperatures. Equally, the MOS can be applied to create custom solutions for specific weather-related problems. For example, de-icing operations and dynamic line rating.
Sign up for the weekly How It’s Made Update. Stay informed and get weekly updates, delivered straight to your inbox.
Our MOS system is the core of our worldwide forecasting system. Not only because it adds a lot of quality to raw-model data, also MeteoGroup’s MOS system provides the customer with specific derived weather parameters that strongly increase the usage options.