The effect of model parameters on objects in forecast fields

In some forecast fields there exist "objects" with well-defined spatial extent, e.g., convective storms or jet streaks. From an object-oriented verification perspective, it is desirable to know if/how features of such objects are affected by model parameters. In this talk, it is shown that regardless of whether the objects are identified via simple thresholding schemes or sophisticated clustering algorithms, many features of the objects are affected by model parameters. A variety of forecast quantities are examined from two NWP models - COAMPS and WRF.