Weather Systems
Many systems built by RGA Labs require direct interaction with atmospheric weather forecasts. RGA Labs maintains extensive experience with a number of weather forecast providers, including Foresight Weather, who provide the state-of-the-art meso-scale and ensemble numerical forecasts; conventional local atmospheric forecasters; and NOAA.
The Metropolis™ forecasting system makes strong use of weather, both actual and forecasted. Usually, hourly forecasts from 0 to 168 hours forword are employed. Typically, the systems cycle through all of the forecast and analysis hourly, thereby providing a new 168-hour forecast and error estimate profiles each hour.
It is not uncommon for a Metropolis™ client to employ weather forecasts from several sources. Metropolis™ is designed to capture each of these forecasts and, similar to how multiple load forecast models are consolidated, the weather forecasts are “stacked” by a non-linear neural model(s) to provide a coordinated weather forecast, as well as a real assessment of the uncertainty in the weather forecast. The weather uncertainty can be converted to a process forecast uncertainty. Additionally, the consolidated forecast is more accurate than any single forecast.
In the case where only one weather forecast is used, it is recommended that a high performance, fully continental, meso-scale forecast be employed. Companies such as Foresight Weather (www.foresightwx.com) provide an excellent forecast for this purpose.
The advantage of using a meso-scale numerical surface forecast is the increased accuracy and resolution, as well as the ability to use a larger weather variable domain in the process forecast models.
In addition to process forecasting, Metropolis™ provides complete historical statistical performance analysis to fully evaluate different weather forecast sources for accuracy.
For more information on our forecasting technology, check out: