CFAN’s temperature forecast verification statistics for Q3
CFAN’s OmniCast product is now providing real time verification statistics for each city in our 15 day City Temperature Forecast product.
Real time verification statistics will help assess the accuracy of our recent forecasts at each forecast lead time out to 15 days. The statistics help the user compare CFAN’s forecasts with the raw forecasts provided by ECMWF and also NOAA GSFX. Further, the verification statistics can help the user identify locations and periods where the forecasts are biased or are performing exceptionally well.
A recent example of the real time verification statistics of daily maximum and minimum temperatures is provided for Atlanta
CFAN is also providing quarterly statistics for the past year for each region and also the continental U.S., for daily mean temperatures as well as maximum and minimum. Shown below is the average statistics for the entire U.S., for the period Jul/Aug/Sept 2016.
While the accuracy of our temperature forecasts varies seasonally and from location to location, as well as with variations in weather regimes, the most useful comparisons are against the ECMWF ensemble mean and the GSFX.
CFAN’s City Temperature Forecasts consistently outperform the raw ECMWF forecasts and NOAA GSFX at all lead times out to 15 days. On average, CFAN’s forecasts for daily maximum and minimum temperatures consistently beat the ECMWF ensemble mean by 0.5 to 1oF or more. For daily maximum temperatures, CFAN’s forecasts at lead times from10-14 days have comparable errors to the GSFX forecasts at 7 days.
We are continuing to work on new product enhancements for our City Temperature Forecast products. Coming soon are:
Incorporation of higher resolution forecasts from ECMWF
Extension to cities in Europe and Asia
An upgrade to the calibration scheme