Using artificial neural networks to predict riming from Doppler cloud radar observations

Abstract

Abstract. Riming, i.e., the accretion and freezing of supercooled liquid water (SLW) on ice particles in mixed-phase clouds, is an important pathway for precipitation formation. Detecting and quantifying riming using ground-based cloud radar observations is of great interest; however, approaches based on measurements of the mean Doppler velocity (MDV) are unfeasible in convective and orographically influenced cloud systems. Here, we show how artificial neural networks (ANNs) can be used to predict riming using ground-based, zenith-pointing cloud radar variables as input features. ANNs are a versatile means to extract relations from labeled data sets, which contain input features along with the expected target values. Training data are extracted from a data set acquired during winter 2014 in Finland, containing both Ka- and W-band cloud radar and in situ observations of snowfall by a Precipitation Imaging Package from which the rime mass fraction (FRPIP) is retrieved. ANNs are trained separately either on the Ka-band radar or the W-band radar data set to predict the rime fraction FRANN. We focus on two configurations of input variables. ANN 1 uses the equivalent radar reflectivity factor (Ze), MDV, the width from left to right edge of the spectrum above the noise floor (spectrum edge width – SEW), and the skewness as input features. ANN 2 only uses Ze, SEW, and skewness. The application of these two ANN configurations to case studies from different data sets demonstrates that both are able to predict strong riming (FRANN > 0.7) and yield low values (FRANN ≤ 0.4) for unrimed snow. In general, the predictions of ANN 1 and 2 are very similar, advocating the capability of predicting riming without the use of MDV. The predictions of both ANNs for a wintertime convective cloud fit with coinciding in situ observations extremely well, suggesting the possibility to predict riming even within convective systems. Application of ANN 2 to an orographic case yields high FRANN values coinciding with observations of solid graupel particles at the ground.

Cloud and Precipitation Particle Identification Using Cloud Radar and Lidar Measurements: Retrieval Technique and Validation

“…Machine learning techniques can also be used for classifying hydrometeor types. Neural networks in particular have been used for particle characterization (Liu & Chandrasekar, 2000), or detection of supercooled liquid (Luke et al, 2010) and detection of riming (Vogl et al, 2022) in mixed-phase clouds. Roberto et al (2017) used a Support Vector Machine (SVM) to classify particles in dual-polarization radar observations, which was trained and evaluated with classifications from a fuzzy logic algorithm.…”

“…Machine learning techniques can also be used for classifying hydrometeor types. Neural networks in particular have been used for particle characterization (Liu & Chandrasekar, 2000), or detection of supercooled liquid (Luke et al., 2010) and detection of riming (Vogl et al., 2022) in mixed‐phase clouds. Roberto et al.…”

Doppler spectra from DWD's operational C-band radar birdbath scan: sampling strategy, spectral postprocessing, and multimodal analysis for the retrieval of precipitation processes

“…Estimating the typical degree of riming for individual modes of frozen precipitation allows a high-resolution analysis of the principal particle growth mechanisms in precipitating clouds above the melting layer. Future efforts could also include testing another recently introduced approach for retrieving the rime mass fraction from radar Doppler spectra that is less impacted by vertical air movements than the method employed in this study (Vogl et al, 2022). These riming retrievals can then be combined with scanning polarimetric radar measurements or atmospheric wind and temperature fields derived from atmospheric models to provide a more complete picture of precipitation microphysics and atmospheric thermodynamics (Oue et al, 2018;Trömel et al, 2021).…”

Identifying cloud droplets beyond lidar attenuation from vertically pointing cloud radar observations using artificial neural networks

“…These applications are particularly powerful due to their ability to perform part of the data pre-processing themselves. Vogl et al (2022) showed that ANNs can be used to predict riming using ground-based zenith-pointing cloud radar variables radar reflectivity, spectrum width, and skewness. An earlier approach from Luke et al (2010) transfers the features of Doppler spectra into particle backscatter and volume depolarization of a high-spectral-resolution lidar (HSRL) using a multi-layer perceptron model and was further validated by Kalesse-Los et al (2022) by applying the pretrained machine learning model to data from the Analysis of the Composition of Clouds with Extended Polarization Techniques (ACCEPT) campaign (Myagkov et al, 2016a;Myagkov et al, 2016b).…”