|The authors have acknowledged that they are lacking critical information about the internal optical configuration, so they have restructured the paper to focus on trying to improve the interpretation of the measurements and then to use derived corrections to improve comparisons of total particle count with an OPC, and LWP with microwave retrievals and model reanalysis.|
I applaud their attempts to characterize the CPS so that other scientists will wish to use it; however, there are several severe problems that they fail to address and since the manufacturer is not part of this study, I don't see any actual, further improvement in interpreting the data. At the moment, it is really nothing more than a particle counter with very little size or shape information.
Here are the primary issues:
1) Corrections to the number concentration using pulse width analysis.
2) Poor statistics on particle size and shape.
The first issue lies in the assumptions that go into doing a coincidence correction based on the transit time. The primary assumption assumes that differences in the actual pulse width (PW), compared to the expected, are mainly a result of flow velocity variations or multiple particles in the beam. Hence, the authors go to a lot of effort to model the flow around the sonde and in the particle deliver system but never do they acknowledge that the large variations in PW are like a result of the non-uniform beam intensity and non-uniform beam geometry. Without taking these factors into account, something requires inforrmation and cooperation from the manufacturer, the correction factors that are derived are meaningless. The probability of coincident can be calculated quite directly since the sample volume of the sonde seems to be known. Why isn't this done?
The second issues, that of poor statistics on particle size and shape, concerns how effective radius, LWC and shape are derived. The data transmission rate, according to the authors and manufacturer, is 25 bytes/second. The manufacturer has chosen to use this to transmit size and shape information for only the first six detected particles each second. According to the drawings, the sensitive sample area presented to particles in the inlet is 1 cm2. Since the flow velocity is approximate 5 ms-1, this means that the CPD will be detecting approximately 500 cm3 per second. Even if the cloud concentrations are very small, for example 10 cm-3, this will be 5000 particles/second. If the CPD can only transmit size information on the first six out of these 5000 particles, this is only 0.12% of the particles. Statistically speaking, the probability that these 6 particles represent the parent population is less than 1%. This by itself makes the CPD of limited use, but what is even more unfortunate is that the size that is derived from these six particles will be heavily biased toward smaller sizes, give the generally log normal size distribution of droplets in cloud where smaller droplets dominate the number concentration, i.e. the first 6 particles samples will mostly likely be in the smallest droplets. Likewise, in a mixed phase cloud, the liquid phase will predominate so that the polarizatio ratio will be biased toward the droplets rather than ice.
What is puzzling is why the manufacturer chose to waste the limited data transmission by sending individual article information. They could have instead, compiled a size distribution of 5 or 6 channnels, with increasing width. 25 bytes is 200 bits. 12 bits is 4096 counts, so that a size distribution, percentage of ice and some housekeeping could have been encoded much more efficiently thatn the current configuration.
Summarizing, although I have recommended a major revision, and am willing to review another revision, until the manufacturer teams with these or other investigators to address the problems that I have raised, I don't think there will be many CPD users in the future,