Thursday, 7 August 2003: 4:00 PM
Can't see the forest for the trees: Methods for the analysis and visualization of large radar datasets
In the case study approach of a precipitation event two dimensional images (PPIs, RHIs, etc.) of radar data are often examined frame by frame (as a function of time) in order to understand the evolution and dynamics of the event. When dealing with several weeks, months or even years of data, however, such approaches often end in frustration and little understanding is gained from the massive amounts of radar images. If we are to study the precipitation cycle and understand its relationship to forcing mechanisms (synoptic, density currents, undular bores, diurnal heating, etc.) for a long time series, other approaches that reduce the volume of data to a manageable level can provide more insight. One approach is to reduce the dimensions of the dataset by averaging in one spatial dimension (latitude or longitude) and plotting the data in the remaining spatial dimension as a function of time (time-longitude or time-latitude). Such plots are often referred to as Hovmoller diagrams and have been used in the climatology community for many years. In the Hovmoller diagrams coherent precipitation episodes travelling across the continent appear as streaks, and patterns associated with different types of forcing are revealed. From the streaks, statistics on the propagation speed, span (km) and duration (h) of the precipitation event are easily obtained. Several examples of precipitation episodes forced by various mechanisms will be shown as well as methods for displaying data averaged over several years. This approach is not limited to radar data and when combined with satellite, model analysis and lightning data becomes even more powerful.