Verification of the time evolution of precipitation systems in numerical weather forecasts

Precipitating weather systems have life cycles involving formation, evolution, and decay. These systems can translate across spatial regions spanning 1000's of km. Climate scientists, numerical modelers, and operational hydro-meteorologists are especially interested in understanding how the characteristics of precipitation systems change over space and time. For example, a topic of current interest is whether global climate change can be observed in the changing characteristics of precipitating weather systems. In addition, there are many outstanding questions regarding the ability of numerical weather prediction systems to predict realistic weather events that translate and evolve over time and space that could be addressed with appropriate forecast evaluation methods. For example, do state-of-the-art numerical weather prediction models accurately predict temporal changes in the morphological characteristics of convective precipitation systems? In order to answer these types of questions, automated techniques of identifying and tracking individual weather systems must be developed and applied to forecast and observed data, and methods of comparing forecasts and observations of time-varying features must be developed. This paper will focus on the problem of developing a framework for the comparison of meteorological features that evolve with time. Since weather systems typically evolve and translate in space and time, explicit analysis of their characteristics requires a feature-based or "object-oriented" approach. Such an approach involves identification of the precipitating weather system of interest and measurement of appropriate characteristics of each system. Individual systems within the forecast and observed data must be tracked over time, which involves issues of determining the beginning and ending of specific events. Issues related to comparing predicted and observed weather events must also be addressed, as well as the problem of comparing forecast and observed events with unequal lifetimes.