Evaluating model skill by synoptic pattern allows for an understanding of how model performance varies with particular atmospheric states and will aid forecasters with pattern recognition. To conduct this analysis, a competitive neural network known as the Self-Organizing Map (SOM) will be used. SOMs allow the user to represent atmospheric patterns in an array of nodes that represent a continuum of synoptic categorizations. North American Regional Reanalysis (NARR) data during the warm season (April-September) will be used to perform the synoptic typing over the study domains.
Simulated precipitation will be evaluated against observations provided by the National Centers for Environmental Prediction (NCEP) Stage IV precipitation analysis. To investigate model performance, two types of SOMs will be generated. Climatological SOMs will be created with all available data to determine whether the model has precipitation biases across all patterns. To capture more detail for convective events, a second set of SOMs will be generated using only days with observed precipitation. These more detailed SOMs will provide better insight into mechanisms that may be responsible for model bias. An example SOM for the Southern Great Plains with mean sea-level pressure and 900 mb relative humidity contour is included.