Subgrid-scale model tests using petascale computers

Large-eddy simulation (LES) has become an effective research tool for the atmospheric bound¬ary layer (ABL) and probably the most in the atmospheric surface layer LES suffers from inherent underresolution and poor subgrid-scale (SGS) model performance. To develop improved SGS models a methodology for testing model perfor¬mance is needed. Previous tests rely on comparing the modeled and measured/DNS SGS stress (apriori tests) or comparing LES and measured profiles (a posteriori tests). Such test methods have severe limitations. From apriori tests it is difficult to predict the effects of model behavior on LES results, e.g., the correlation between the modeled and measured SGS stress components provides little information about model performance in simulations. From a posteriori tests it is difficult to relate deficiencies of LES results (e.g., the mean velocity and Reynolds stress profiles) to specific aspects of the model behavior. The two types of tests are disconnected as they deal with different variables (instantaneous SGS stress and LES statistics), respectively. Therefore, apriori and a posteriori test results cannot be directly compared to further evaluate model performance.