 |
 |
| Home | Poster Submission | Posters | Archives | Contacts |
|
Science Rendezvous > 2009 Posters Probabilistic Evaluations of a Cloud System Resolving Model Using ARM Observations Pete Henderson and Robert Pincus Evaluations of Cloud System Resolving Models (CSRMs) are usually made using case-studies which, by definition, sample a narrow range of atmospheric states over short time periods. Here we demonstrate an alternative, using a 3-year run to sample a wide range of conditions. The large size of this data set allows us to use probabilistic evaluation techniques employing the actively-sensed profiles of cloud properties obtained at the Atmospheric Radiation Measurement (ARM) program's continental mid-latitude site. This approach also makes direct use of instantaneous observations, thereby avoiding temporal averaging of the observations. The CSRM is driven by ARM's observationally constrained forcing data, and modeled thermodynamic fields are kept close to those observed by nudging them toward the soundings; snapshots of the modeled cloud and thermodynamic fields are saved hourly. To compare the model cloud to the observed cloud, we simulate the measurements made by the ARM's radar and lidar, and use their sensitivities to define cloud occurrence in the model. The probability of cloud (PoC) within the CSRM is then evaluated against the combined binary cloud-mask of the radar and lidar, as a function of height, using methods from ensemble forecast verification that compare the statistics of subsets sorted by the model forecasts. Techniques such as reliability diagrams and Brier scores are then used to quantify performance. Results will be shown for a range of CSRM configurations, from high-resolution (500 m) 3D and lower-resolution 2D, typical of the CSRM's implementation in the Multiscale Modeling Framework, including those employing an intermediate prognostic higher-order turbulence closure (IPHOC). |
216 UCB