Lossless Coding and Compression of Radar Reflectivity Data

The coding and compression of weather imagery can be improved by following a domain-specific approach in which the characteristics of the data are taken into account in devising the encoding scheme. We show that by predicting that the next valid data element is identical to the current one, a dramatic reduction in source entropy (information content) is achieved. We also show that encoding runs of missing values in combination with such a prediction scheme makes further compression unnecessary.

A study was carried out on seventy volumes of radar reflectivity data collected by the WSR-88D in Fort Worth, Texas on May 5, 1995. These volumes provide an indication of the distribution of reflectivity data on a typical storm day. Since the amount of valid data is higher on a storm case, these volumes also serve as a worst case scenario from a compression stand-point.