Thursday, 15 January 2004: 4:15 PM
Spatial bias errors in the operational NCEP Eta model
Both 0000 and 1200 UTC runs of the operational Eta model are archived at the National Severe Storms Laboratory along with local runs of the operational Eta model using the Kain-Fritsch convective parameterization scheme. Verification statistics over a period of 14 months (26 January 2001 through 31 March 2002) are generated and analyzed for these data. The forecast errors are defined by the difference between the 24 h forecast field and the verifying initial analysis field. Forecast errors are available for any of the archived fields, which include heights, and winds for 850, 700, 500 and 250 hPa. By applying basic statistical tools, easily applied to any gridded forecast field of continuous random variables, appropriate corrections are made for temporal and spatial degrees of freedom, yielding the spatial distribution of 95% confidence intervals about the mean error at each gridpoint. Mean bias errors are defined for the year and the four seasons. Some notable differences between the spatial bias error structure of the 0000 UTC forecast cycle and the 1200 UTC forecast cycle, as well as the operational Eta and Eta using the Kain-Fritsch scheme are also observed.
Hovm÷eller analyses show that sometimes the bias error associated with a propagating system is coherent across the entire domain. At other times, this is not the case and the bias errors are essentially removed. A more detailed examination shows that coherent spatial bias errors appear to be associated more with errors in ridges, than in troughs. Specific examples are shown, and some plausible causes for the nature of the bias errors are proposed.