currently I’m working on the automatic extraction of training sets for my radar spectrum based target classifier.
Using the Cloudnetpy Detection Status and Target Categorization, the training set is selected as follows:
Use all datapoints with detection status = “Good radar & lidar echos” .
Add all pixels containing the classes = “Ice & supercooled liquid” and “Cloud droplets only”.
Then remove all pixel with detection status = “Lidar only” (induced by adding “Cloud droplets only”)
This work ok’ish, but I was wondering if we get better (more) “Good radar & Lidar signals” pixel when processing Cloudnetpy with PollyXT data rather than ceilometer. Would this be true and is there a PollyXT reader implemented jet?
Also, would PollyXT reduce the "missclassification" regarding the vertical lines at 18UTC between 3km - 4km altitude? The classification using 94GHz radar (without gaps) seems to be more vulnerable compared to the 35GHz radar (with gaps). Or is there an easier way to filter out these datapoints?