DIRECTIONAL FREQUENCY CORRECTION FOR THE STAR METEOROLOGICAL JOINT FREQUENCY COMPUTER PROGRAM

Abstract

Atmospheric dispersion computer programs are widely used to simulate air and surface exposures from airborne and deposited radionuclides. For longer-term emissions (i.e., occurring for a year or longer), many of these models use joint frequency tables, which summarize the frequency of occurrence of specific ranges of wind speed and direction as a function of stability categories to define the atmospheric dispersion conditions. This paper addresses directional limitations of a computer code, STAR (STability ARray), that has been historically used to generate these joint frequency tables. As part of an effort to provide an updated version of the STAR code, directional limitations were found stemming from the manner in which the reported direction data are stored in joint frequency data tables. The STAR code provides tabular frequency summaries based on sixteen 22.5-degree wind direction sectors. Surface observation data in the CD-144 format from the United States National Oceanographic and Oceanic Administration (NOAA) is read by the STAR program. For 1964 and later years the wind direction has been reported in CD-144 datasets as 10-degree increments instead of the 22.5-degree sector direction codes used in prior years. When 10-degree data are input, the process by which the STAR code puts each wind direction occurrence into the 22.5-degree sectors results in a consistent positive bias for cardinal direction sectors (north, south, east, and west), and a consistent negative bias for all the other sectors. Individual entries in the joint frequencies summaries with high wind frequencies tend to be overestimated by up to about 30% or underestimated down to about -10%. A larger range of changes is seen for entries with lower wind frequencies. To avoid these errors, it is recommended that joint frequencies generated by the STAR program be checked for this directional bias. If the NOAA observational data are for 1964 or later years and generated by the original EPA STAR program, it is likely that the bias will be in the dataset. To be certain, it is best to regenerate the STAR data summary using a revised version of the STAR program (STARR) or an alterative program that better handles the binning of wind directions.