Improving Droplet Size Distribution Measurements from Grayscale Optical Array Probes Using a Shape Factor Method

Abstract:

Optical array probes (OAPs) are the de-facto standard for making in-situ measurements of particle size distributions in clouds. These instruments are flown on aircraft and image the shadow of individual droplets as they pass through a laser. Commercially available OAPs have image resolutions between 10 and 150 µm, generate images with widths of 64 or 128 pixels and have a bit depth of either 1 (black/white) or 2 (4 levels of grayscale).

It is well understood that images from OAPs are degraded by 1) diffraction, 2) pixelization, 3) thresholding to the instrument bit depth 4) electronic response time, 5) clipping when a shadow is partially imaged. Smaller particles are particularly affected by 1-4, and the impact can be large enough that the image breaks up or is not recorded at all. Larger particles are mostly affected by 5.

Heymsfield & Parrish (1978) and Korolev (2007) provided corrections for clipping and diffraction for 1 bit images under certain circumstances, so in many cases, 2-bit data is downgraded to 1-bit to use these methods. More recently, O’Shea (2019) utilised 2 bit data, for a specific size range with no clipping, to improve diffraction corrections.

Here we present a new shape factor method for utilizing the greyscale information in OAPs to more accurately size droplets. The method utilises the radial distribution of the shadow depth and the deviation of the shadow from circular symmetry. Together these two parameters are sensitive to the effects of both diffraction and partially imaged shadows. The method interpolates a synthetic data set to determine droplet diameter and is tested using a second synthetic data set. Because the method relies on two simple-to-calculate parameters, it can be applied to any size particle and to instruments with any bit depth, resolution or image width.

Synthetic data for a Cloud Imaging Probe-15 OAP (15 µm resolution, 62 pixel wide image, 2-bit) indicates accurate sizing can be performed with little bias for particle diameters from 20 µm to 3 mm (1.3 pixels to 3 times the full image width). By comparison, the standard 1 bit analysis requires droplet sizes above ~50 µm and shows an oversizing bias up to ~5%.

References

Heymsfield & Parrish (1978) doi: 10.1175/1520-0450(1978)017<1566:ACTFIT>2.0.CO;2

Korolev (2007) doi: 10.1175/JTECH1980.1

O’Shea et al (2019) doi: 10.5194/amt-12-3067-2019