Exploring dual-lidar mean and turbulence measurements over complex terrain

Abstract. To assess the accuracy of lidars in measuring mean wind speed and turbulence at large distances above the ground as an alternative to tall and expensive meteorological towers, we evaluated three dual-lidar measurements in virtual mast (VM) mode over the complex terrain of the Perdigão-2017 campaign. The VMs were obtained by overlapping two coordinated Range Height Indicator scans, prioritising continuous vertical measurements at multiple heights at the expense of high temporal and spatial synchronisation. Forty-six days of results from three VMs (VM1 on the SW ridge, VM2 in the valley, and VM3 on the NE ridge) were compared against sonic readings (at 80 m and 100 m a.g.l.) in terms of 10 min means and variances, to assess accuracy and the influence of atmospheric stability, vertical velocity, and sampling rate on VM measurements. For mean flow quantities–wind speed (V_h), and u and v velocity components–, the r² values were close to 1 at all VMs, with the lowest equal to 0.987; whereas in the case of turbulence measurements (u′u′ and v′v′), the lowest was 0.869.  Concerning differences between ridge and valley measurements, the average RMSE for the wind variances was 0.295 m² s⁻² at the VMs on the ridges. In the valley, under a more complex and turbulent flow, smaller between-beam angle, and lower lidars’ synchronisation, VM2 presented the highest variance RMSE, 0.600 m² s⁻² for u′u′. The impact of atmospheric stability on VM measurements also varied by location, especially for the turbulence variables. VM1 and VM3 exhibited better statistical metrics of the mean and turbulent wind under stable conditions, whereas, at VM2, the better results with a stable atmosphere were restricted to the wind variances. We suspect that with a stable and less turbulent atmosphere, the scan synchronisation in the dual-lidar systems had a lower impact on the measurement accuracy. No correlation was found between VM measurement errors and the vertical wind speed measured by the anemometers, confirming the validity of the VM results and the zero vertical velocity assumption. Lastly, the VMs’ low sampling rate contributed to 33 % of the overall RMSE for mean quantities and 74 % for variances, under the assumption of a linear influence of the sampling rate on the dual-lidar error. Overall, the VM results showed the ability of this measurement methodology to capture mean and turbulent wind characteristics under different flow conditions and over mountainous terrain. Upon appraisal of the VM accuracy based on sonic anemometer measurements at 80 and 100 m a.g.l., we obtained vertical wind profiles up to 430 m a.g.l. To ensure dual-lidar measurement reliability, we recommend a 90° angle between beams and a sampling rate of at least 0.05 Hz for mean and 0.2 Hz for turbulent flow variables.