By Adam Grossman on .
The Short-Range Ensemble Forecast (SREF)
This model helps provide more reliable, higher-resolution forecasts for the continental US and much of Mexico and Canada, up to 3 days into the future.
The Real-Time Mesoscale Analysis (RTMA)
This model provides real-time data on near-surface temperature and wind speed at a super high resolution of 2.5 kilometers. This is by far the highest resolution data we now ingest (apart from doppler radar), and it should greatly improve the accuracy of our current conditions in areas that experience localized microclimate effects.
One such place is the San Francisco Bay Area, which can see extreme temperature swings of 10° or 15° in as many miles. Previously, we relied primarily on ground station (METAR) observations for current conditions, but there just aren't enough stations in the Bay Area to account for all the varied microclimates. This new data source should greatly improve matters.
Not only do we use RTMA data directly, but we also use it to bias the ground station observations. In other words, if your location happens to be 5° warmer (on average) than the nearest ground station, we'll adjust the station's temperature upwards by 5° before factoring it into your particular forecast.
(The RTMA model will also be crucial in the future for validating our extended hyperlocal forecasts, but we'll talk more about that at a future date.)
As always, please let us know if you see any weird forecasts. We're constantly working to improve our systems, and that can only happen with your help.