Datensatz

Zusammenfassung

Cloud radar Doppler spectra contain vertically highly resolved information about the hydrometeors present in a cloud. A mixture of different hydrometeor types can lead to several peaks in the spectrum due to their different fall speeds, giving a hint about the size/ density/ number of the respective particles. Here we present the effort of joining two methods to separate and interpret peaks in cloud radar Doppler spectra (Kalesse et al., 2019; Radenz et al., 2019). The overarching goal is to make them insensitive to instrument type and settings, and applicable to all cloud radars which are part of the ACTRIS CloudNet network. A supervised machine learning peak detection algorithm (PEAKO, Kalesse et al., 2019) is used to derive the optimal parameters to detect peaks in cloud radar Doppler spectra for each set of instrument settings. In the next step, these parameters are used by peakTree (Radenz et al., 2019), which is a tool for converting multi-peaked Doppler spectra into a binary tree structure. peakTree yields the (polarimetric) radar moments of each detected (sub-)peak and can thus be used to classify hydrometeor types. This allows us to analyze Doppler spectra with respect to, e.g. the occurrence of supercooled liquid water or ice needles with high linear depolarisation ratio (LDR). The new toolkit is evaluated using observations from two field campaigns, the DACAPO-PESO experiment and the NASCENT campaign. For DACAPO-PESO, two different radar systems are compared, and for NASCENT, detected peaks are validated against in-situ cloud observations collected with a holographic imager.