Unraveling the lidar-turbulence paradox

The meteorological community, and in particular the wind energy community, have been trying to establish a methodology to correct/convert turbulence measures derived from measurements performed with Doppler wind lidars. The idea with these corrections is that lidar-based turbulence measures become equivalent or comparable to those turbulence measures that we normally retrieve using traditional in-situ instrumentation such as cup anemometers on masts. Part of the reason of this quest is the ability of Doppler wind lidars to be both versatile and affordable, apart from the most important characteristic of these instruments, which is their high accuracy and precision in measuring winds within and beyond the limits of meteorological masts. Particularly, for offshore applications, floating Doppler wind lidars are nowadays the standard for wind resource assessment, since deploying and maintaining instruments on tall meteorological towers offshore is currently too expensive. With regards to versatility, Doppler wind lidars are used for different applications ranging from, e.g., atmospheric modeling, wind turbine wake studies, and wind turbine control, among others. The wind energy community in particular is therefore making efforts in the development of recommended practices and standards to enhance the adoption of measurements from lidars. Corrections for lidar-based turbulence measures have been however investigated for decades. The major difficulties for establishing a turbulence correction are related to two main points: turbulence contamination and turbulence filtering due to probe-volume averaging. These two points lead to what we here refer to as the lidar-turbulence paradox, which basically means that in order to determine the ratio between the lidar turbulence measure to that of a cup anemometer, we need to know the characteristics of atmospheric turbulence; but these characteristics of turbulence are the ones we want to measure with our lidars.

We circumvent the paradox using a physics-based lidar-turbulence model, which serves for the training of a number of neural networks. The measurements we study are from continuous-wave Doppler lidar wind profilers, which were deployed besides a tall 250-m meteorological mast at the Østerild test station for wind turbines in northern Denmark. Metek USA-1 sonic anemometers on the masts match four lidar measurement levels and are used as reference for the turbulence measures. The prediction of the neural networks, which has the lidar measures as inputs, are compared with the measurements from independent Doppler wind lidars.