In this study we examine the accuracy of the horizontal winds and their estimated uncertainties as obtained from the 8-beam step-stare PPI scan data using a modified velocity-azimuth-display (VAD) algorithm. The algorithm propagates the radial velocity error through the calculations to produce estimates of the uncertainty in the retrieved wind speed and direction. It is these uncertainty estimates that are the primary focus of this study. Results from a recent field campaign are used to assess the accuracy of the uncertainty estimates.
In March and April of 2015, one of the ARM Doppler lidars (Halo Photonics StreamLine) was deployed to the Boulder Atmospheric Observatory for the eXperimental Planetary boundary-layer Instrument Assessment (XPIA) field campaign. During XPIA, the 300-m tower at the BAO site was instrumented with sonic anemometers at six levels (50, 100, 150, 200, 250, and 300m), and the ARM lidar was run in its standard operational configuration. In this study, we assess the accuracy of the lidar-derived winds and estimated uncertainties by comparison with the sonic anemometers on the BAO tower.
Three wind retrieval trials were conducted using different methods for estimating the uncertainty in the derived wind components. For Trial 1 the radial velocity uncertainties were set to unity and the uncertainties in the retrieved wind components were estimated from the unweighted covariance matrix. For Trial 2 the radial velocity uncertainty for a given beam and range gate was estimated by computing the variance of the radial velocity over three consecutive PPI scans and three neighboring range gates. The uncertainties in the retrieved wind components were then computed from the weighted covariance matrix. This scheme approximates a scanning strategy in which multiple independent radial velocity samples are acquired in each look direction. This provides a means of estimating the uncertainty directly from the observations and avoids the assumption of isotropy inherent in Trial 1. For Trial 3 the radial velocity uncertainty is taken to be equal to the instrumental noise (i.e. radial velocity precision) so that the effects of turbulence are completely ignored. The uncertainties in the retrieved wind components were then computed from the weighted covariance matrix.
Results from these trials show that the lidar wind speeds and directions are essentially unbiased (~2cm s-1 and ~1o, respectively), regardless of the uncertainty estimation scheme used. The uncertainties for Trials 1 and 2 are similar and show a strong diurnal dependence with larger uncertainties occurring during the daytime period. The uncertainties for Trial 2 are slightly larger than Trial 1. By contrast, the uncertainties for Trial 3 are much smaller than either Trial 1 or 2 and do not show a strong diurnal variation.
Root-mean-squared (RMS) wind speed and direction differences between the lidar and the tower were compared to the lidar’s wind speed and direction uncertainty estimates. The results show that the estimated uncertainties are consistently smaller than the observed RMS differences for all trials. Trials 1 and 2 give similar results, but the uncertainty estimates for Trial 2 are closer to the observed RMS differences. By contrast, Trial 3 produced estimates that are far smaller than the observed RMS differences. Furthermore, when the uncertainty estimates are used to filter the measurements we found that Trails 1 and 2 produced substantially better agreement with the tower than Trial 3. These results indicate that when only instrumental errors are considered, the resulting uncertainty estimates in a wind speed and direction are grossly underestimated and poor indicators of data quality.