EOL Seminar: Assessment of the High-Resolution Rapid Refresh Model’s Ability to Predict Large Convective Storms using Object-based Verification

Wednesday, May 28, 2014 - 20:00 to 21:30
Contact Name: 
Steve Oncley
Contact Email: 
Contact Phone: 

James Pinto

Hydrometeorological Applications Program

National Center for Atmospheric Research

An object-based verification technique is utilized to detect large-scale convective storms in both radar observations and in model ensembles.  In this paper, the technique is used to evaluate the skill of the NOAA/GSD High-Resolution Rapid Refresh (HRRR) time-lagged ensemble at predicting both the macroscale properties of Large-scale Convective Storms (LCS) including its ability to predict the timing of large-scale storm initiation (LCS-I).  The macroscale properties of LCSs evaluated in this study include storm frequency, size, aspect ratio, orientation.  An LCS is defined as an area of convection extending over 100 km in at least one dimension the persists for at least 1 hour. Convective areas are identified by finding near-continuous (only small gaps are allowed) areas of vertically-integrated liquid (VIL) exceeding some threshold value. As such, this definition encompasses several different types of convective storms including squall lines, Mesoscale Convective Systems (MCSs) and Mesoscale Convective Complexes (MCCs) and storm clusters.  A key aspect of this evaluation is that mean bias in the modeled VIL field is accounted for by optimizing the VIL threshold used to detect LCS.  It is found that the model has significant skill at predicting the macroscale properties of LCSs; however, the analyses indicate that the HRRR model also exhibited second order biases that were a function of lead time, region and time of day. Generally, the model tended to underforecast the LCS frequency in the southeastern U.S. during the day and overforecast the LCS frequency over the Great Plains at night. Reasons for these second order biases are explored using matched-pair analyses and case studies. Finally, a description of how PECAN observations may be used to improve mesoscale model predictions of LCS is given.


Wednesday, 28 May 2014, 2:00PM

NCAR-Foothills Laboratory

3450 Mitchell Lane

Bldg. 2 Small Seminar Room (Rm 1001)