Evaluation of the real-time WRF forecasts during the Mountain Terrain Atmospheric Modeling and Observations (MATERHORN) Program: Performance, comparison with observations, and further implications

One of the primary objectives of the Mountain Terrain Atmospheric Modeling and Observations (MATERHORN) Program is to evaluate model performance in predicting synoptic and local flows over mountainous terrain and thus to improve predictability. In order to achieve this goal, during the fall of 2012 and spring of 2013, two field experiments were conducted over Dugway Proving Ground (DPG), Utah, with comprehensive observations collected of soil states, surface energy budgets, near-surface atmospheric conditions, and profiling measurements from multiple platforms (e.g., balloon, lidar, radiosondes, etc.).

During both field experiments, a real-time forecast was performed at the University of Utah with a mesoscale community Weather Research and Forecasting (WRF) model at high resolution (~1 km horizontally), four times (at 00, 06, 18, and 24 UTC) a day. The purpose of this real-time forecasting was not only to support decision-making during the field program but also to provide a useful database to evaluate the WRF model's performance in predicting synoptic flows over mountainous terrain. During the field program, a series of 48-h forecasts were produced for over 200 forecast leading times and for all Intensive Observational Periods (IOPs) during September–October 2012 and May 2013. After the field experiments, these forecast results were compared with observations collected from the field experiments.

Comparison between the WRF forecasts and observations show notable biases, with diurnal variations in near-surface atmospheric conditions under quiescent cases and flow-dependent patterns in errors associated with transitions and strong synoptic forcing cases. More comparisons are made with profile observations from multiple platforms. The ability of the WRF model to predict synoptic and local flows over complex terrain and error characteristics is further evaluated. It is found that the WRF model is capable of predicting the atmospheric conditions in most of the cases while the accurate forecast of the near surface atmospheric conditions remain a great challenge. Following the results, sources of the errors and deficiencies in model physical parameterizations are being diagnosed. Additional numerical experiments are being conducted with sensitivity studies and data assimilation in order to get additional insights to the predictability limitation in near-surface atmospheric conditions. The potential ways for future improvements are being discussed.