Tuesday, 5 October 2004
During the winter of 2002-2003, an evaluation of the Doppler on Wheels (DOW) mobile radar as a surrogate for in situ aircraft measurements of turbulence was performed in the vicinity of Juneau, Alaska. DOW radar data was collected that was approximately coincident in space and time with the collection of wind data by the University of Wyoming King Air research aircraft. In general, repetitive targeted DOW PPI surveillance scans sampled the atmospheric volume measured by the aircraft within three minutes of aircraft data collection. Estimates of turbulence were derived from the aircraft measurements of along-track, cross-track and vertical wind measurements every second. In theory, these three values should be similar, at least for fully developed isotropic turbulence. In practice, however, disagreement is not unusual. For the DOW, six algorithms were developed to convert measured radial velocity and/or spectral width to estimates of turbulence. Data confidence was assigned to these estimates using a combination of reflectivity and normalized coherent power. Several target regions for the comparison were defined on the basis of prior observations of elevated levels of turbulence. For each aircraft crossing of a targeted region, or "hazard box," the turbulence time series data were reduced to a single value by applying either a median-filter, 75% filter, or a 90% filter. Similarly, the two-dimensional turbulence field derived from the spatially and temporally coincident (approximately) DOW scan was reduced to a single value by applying the same filters to the spatially distributed radar-derived data set. The three wind component-based estimates of aircraft turbulence combined with the three filters applied to the time series data result in 9 representative values of aircraft-based turbulence estimates. The six radar turbulence algorithms combined with the three spatial filters result in 18 values of DOW-based turbulence estimates. Attempts to correlate pairs of aircraft and DOW estimates without discriminating between target areas were unremarkable. The greatest correlation coefficient associated with any of the 162 potential combinations of aircraft- and DOW-based estimates of turbulence was 0.4, corresponding to an R-squared value of 0.16. However, restricting each statistical correlation on the basis of target area and DOW location, produced a substantial improvement in skill. One such combination of target area and DOW location produced a correlation coefficient of 0.88 (R-squared 0.78). This improvement suggests a site-specific turbulence generation mechanism and/or a dependency upon look-angle in our analysis. This presentation will review the analysis procedure and results, and discuss the potential for turbulence measurements from scanning radar measurements.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner