Extended Abstract (112K)8.4
An Analysis of the Spatial Distribution of Eta Surface Temperature Forecasts through the Gridded Forecast Editor
Steven A. Amburn, NOAA/NWS, Tulsa, OK
Many National Weather Service offices (WFOs) are now making gridded forecasts through the Gridded Forecast Editor (GFE), developed by the Forecast Systems Laboratory (FSL) in Boulder, Colorado. These gridded forecasts can provide greater spatial resolution to the customers than current official forecasts. However, greater spatial resolution may not always equate to greater accuracy.
To produce gridded temperature forecasts, a meteorologists may start with a blank grid and draw contours of temperature. It may be preferable that the meteorologist start with an initialized grid field from one of the numerical models. The Eta model provides the greatest spatial resolution of surface temperature, due to the high resolution topography in the model. An additional adjustment, developed by FSL for the GFE, enhances the resolution to 5 km. Meteorologists cannot draw temperature grids with that same resolution. It may therefore be best that a meteorologists use the Eta/GFE grids and adjust the entire temperature field up or down to the optimal, final forecast. Other methods of modifying the initialized grid field will destroy much or all of the spatial resolution provided by the model.
The Eta/GFE initializations provide highly detailed surface temperature forecast grids across eastern Oklahoma and northwest Arkansas by accounting for the minor, but distinct variations in elevation. An analysis was conducted to determine if the spatial distribution of Eta model maximum and minimum temperature forecasts, through the GFE, was superior to the forecast temperatures from Model Output Statistics (MOS) and/or the official Coded Cities Forecast (CCF). Specifically, when GFE grid fields are initialized with Eta model maximum and minimum surface temperatures, do uniform ("global") adjustments to the fields result in the lowest average temperature errors? Statistics are presented to answer this question.
Session 8, Phenomenological Forecasting II
Wednesday, 14 August 2002, 1:30 PM-3:00 PM
Previous paper Next paper