Diagnosing an Operational Numerical Model Using Q-Vector and Potential Vorticity Concepts

View More View Less
  • 1 National Oceanic and Atmospheric Administration, Forecast Systems Laboratory, Boulder, Colorado
  • | 2 National Oceanic and Atmospheric Administration, National Weather Service, Seattle, Washington
© Get Permissions
Full access

Abstract

A quasigeostrophic (QG) diagnostic model is used to evaluate the nested grid model's (NGM) predictions for a December cyclone whose impact on northeastern Colorado was underpredicted. Although the NGM predicted deepening of the associated 500-mb low, the model was 12 h slow in the onset of deepening and moved the storm too far east too quickly. Synthetic soundings, generated from 12-h predicted data initialized 24 h before cyclogenesis became apparent, were submitted to the same QG diagnostic algorithms used to analyze verifying rawinsonde data. Comparisons reveal that the NGM apparently 1) transported too much potential vorticity, westerly momentum, and cold air into the lower troposphere along the axis of the jet stream; 2) moved the first of two short-wavelength jet streaks too far northeastward and with too much strength; 3) failed to predict the strength of the following jet maximum; and 4) failed to develop an apparent tropopause fold. It is established that these errors were not caused by obvious discrepancies in the model's initialization. Through inference, the errors could have been caused by rapid growth of subtle, undetected initialization errors or by the model's inadequate parameterization of some physical process—perhaps of turbulent dissipation over mountainous terrain. Diagnosis of the model's subsequent initialization (12 h after its first erroneous prediction) indicates that the model did not have available crucial Mexican soundings that might have prevented it from making a similar error in predicting the position and strength of the then-intensifying cyclone. The diagnostic results could have alerted forecasters not only to the presence of the complex jet stream but also to the extent and intensity of its associated tropopause fold. Furthermore, QG diagnostics can alert forecasters to model errors that are not made obvious by conventional model comparisons.

Abstract

A quasigeostrophic (QG) diagnostic model is used to evaluate the nested grid model's (NGM) predictions for a December cyclone whose impact on northeastern Colorado was underpredicted. Although the NGM predicted deepening of the associated 500-mb low, the model was 12 h slow in the onset of deepening and moved the storm too far east too quickly. Synthetic soundings, generated from 12-h predicted data initialized 24 h before cyclogenesis became apparent, were submitted to the same QG diagnostic algorithms used to analyze verifying rawinsonde data. Comparisons reveal that the NGM apparently 1) transported too much potential vorticity, westerly momentum, and cold air into the lower troposphere along the axis of the jet stream; 2) moved the first of two short-wavelength jet streaks too far northeastward and with too much strength; 3) failed to predict the strength of the following jet maximum; and 4) failed to develop an apparent tropopause fold. It is established that these errors were not caused by obvious discrepancies in the model's initialization. Through inference, the errors could have been caused by rapid growth of subtle, undetected initialization errors or by the model's inadequate parameterization of some physical process—perhaps of turbulent dissipation over mountainous terrain. Diagnosis of the model's subsequent initialization (12 h after its first erroneous prediction) indicates that the model did not have available crucial Mexican soundings that might have prevented it from making a similar error in predicting the position and strength of the then-intensifying cyclone. The diagnostic results could have alerted forecasters not only to the presence of the complex jet stream but also to the extent and intensity of its associated tropopause fold. Furthermore, QG diagnostics can alert forecasters to model errors that are not made obvious by conventional model comparisons.

Save