Detecting turbulent structures on single Doppler lidar large datasets: an automated classification method for horizontal scans

Abstract

Abstract. Medium-to-large fluctuations and coherent structures (mlf-cs's) can be observed using horizontal scans from single Doppler lidar or radar systems. Despite the ability to detect the structures visually on the images, this method would be time-consuming on large datasets, thus limiting the possibilities to perform studies of the structures properties over more than a few days. In order to overcome this problem, an automated classification method was developed, based on the observations recorded by a scanning Doppler lidar (Leosphere WLS100) installed atop a 75 m tower in Paris's city centre (France) during a 2-month campaign (September–October 2014). The mlf-cs's of the radial wind speed are estimated using the velocity–azimuth display method over 4577 quasi-horizontal scans. Three structure types were identified by visual examination of the wind fields: unaligned thermals, rolls and streaks. A learning ensemble of 150 mlf-cs patterns was classified manually relying on in situ and satellite data. The differences between the three types of structures were highlighted by enhancing the contrast of the images and computing four texture parameters (correlation, contrast, homogeneity and energy) that were provided to the supervised machine-learning algorithm, namely the quadratic discriminant analysis. The algorithm was able to classify successfully about 91 % of the cases based solely on the texture analysis parameters. The algorithm performed best for the streak structures with a classification error equivalent to 3.3 %. The trained algorithm applied to the whole scan ensemble detected structures on 54 % of the scans, among which 34 % were coherent structures (rolls and streaks).

Properties of coherent structures over Paris: a study based on an automated classification method for Doppler lidar observations

“…By subtracting the mean wind speed from the radial wind speed observations, the mlf-cs field is retrieved. A detailed description of this process is presented in Cheliotis et al (2020). A characteristic example for each of the identified mlf-cs fields that we aimed to classify is displayed in Fig.…”

“…In section 2b, we briefly summarized the main points of our classification and indicated the classification results for the training dataset made by the QDA algorithm. The classification error was minimized at approximately 9% for five texture analysis parameters (Cheliotis et al 2020). We thereupon utilized the QDA algorithm using these five texture analysis parameters in order to classify the whole ensemble, consisted of 4577 PPI scans.…”

“…We have developed an automated method with the aim to achieve consistent medium-to-large fluctuations and coherent structures (mlf-cs) classifications for single Doppler wind lidar observations (Cheliotis et al 2020). We focus on the streaks and rolls as presented by Young et al (2002).…”

“…The remote sensing methods can provide the mlf-cs fields, by applying the velocity azimuth display (VAD) method in the radial wind observations (Browning and Wexler 1968;Drobinski et al 2004;Lhermitte 1962). In a previous article (Cheliotis et al 2020), we presented our automated mlf-cs classification method based on statistical texture feature extraction (Haralick parameters: Haralick et al 1973;Chaki and Dey 2020) and the quadratic discriminant analysis (QDA) (Hastie et al 2009). We assessed its performance on a training dataset of 150 samples by means of tenfold cross validation.…”

Dependency of vertical velocity variance on meteorological conditions in the convective boundary layer

“…Sathe et al, 2015;Smalikho and Banakh, 2017), and identify coherent structures (e.g. Ansmann et al, 2010;Cheliotis et al, 2020). Doppler lidars are a reliable method to retrieve the σ 2 w /w 2 * profile, as shown in a comparison between σ 2 w /w 2 * profile derived from a Doppler lidar, Large Eddy Simulations (LES) and the empirical profile (Lenschow et al, 2012).…”