Comparisons between multiple in-situ aircraft turbulence measurements and radar in the troposphere

Abstract

Networks of Windprofiler Radars have the capability to make significant contributions to severe weather forecasting (both on the ground and in the air) through the determination of real-time turbulence strengths, but the potential has still not been fully realized. In order to better understand the accuracy of profilers in determination of turbulence strengths, we have compared radar measurements made at the Harrow radar in Canada (located in Southwestern Ontario as part of the O-QNet radar network) with in-situ measurements made by multiple aircraft. These included measurements made both by commercial aircraft and dedicated research aircraft.Research aircraft (instrumented with accelerometers and GPS tracking devices) and radar data were analysed using structure function, spectral and spectral-width methods. Data were also recorded on-board commercial aircraft using accelerometer-based studies, and results were recorded for subsequent analyses. Over 92,000 commercial aircraft measurements, 4000h of radar data, and 15 days of research-aircraft measurements were available for this study, although only a subset of the commercial aircraft data were useable. The radar-based spectral-width method occasionally produced anomalous negative values of the turbulence strength, usually associated with weak turbulence coupled with significant wind variability over scales of tens of kms, but the aircraft data also had limitations. For the commercial aircraft, frequent zeros were common, also associated with weak turbulence. With regard to the research aircraft measurements, it was found through both spectral and structure function analyses that spectral contaminants exist out to scales of many tens of metres (larger than often assumed), but proper allowance for these effects permitted good estimates of turbulence strength. Spatial and temporal variability was large, however, complicating comparisons with the radar.By comparing the in-situ data to the radar data, it has been possible to place stronger limits on the constants used to determine radar turbulence strengths. Subsequent comparisons with the commercial aircraft data over a six-month period have then been used to show that aircraft and radar probability distributions agree in form, but that the aircraft data are three to five times larger. Corrections for this bias lead to good agreement, both in form and absolute values, for all three data sets.