The Use of Performance Metrics to Enhance Meteorological Operations
Jakob, C.(a), Pincus, R.(b), Hannay, C.(b), and Xu, K.-M.(c), BMRC (a), NOAA/CIRES CDC (b), NASA Langley (c)
Thirteenth Atmospheric Radiation Measurement (ARM) Science Team Meeting
It is highly desirable to use cloud radar data in the evaluation of model simulations of clouds at various scales. Unfortunately there is an inherent mismatch between the spatial and temporal scales of the models and the observations. Usually this mismatch is overcome by time-averaging the observations and declaring the averages as representative for a given model spatial scale. Here we explore an alternative method of model evaluation that is based on the interpretation of model cloud predictions as probabilistic forecasts at the observation point. First we contrast time-averaging and probabilistic evaluation techniques in an idealized context by extracting pseudo-observations from a cloud system model (CSM) and comparing them to the model forecasts for different domain sizes. We then apply the probabilistic techniques to the evaluation of the CSM simulation against a set of real remotely sensed cloud observations. The study highlights some conceptual advantages of the probabilistic method as a complementary approach to model evaluation using point observations. It also reveals some of the difficulties arising from the use of simple scalar performance measures. It further shows the potential of the method for a range of other applications involving comparisons of point data to area statistics.
Note: This is the poster abstract presented at the meeting; an extended version was not provided by the author(s).