Video particle profiler (VPP): (from Bochdansky,et al (2017) JMS)

The VPP was similar to that published in Bochdansky et al. (2010). However, instead of 45° angle lighting from both sides, side lighting with two white high-intensity LED lights was used ~7 cm in front of the lens. Some backscatter from transparent exopolymers (TEP), or from small particles embedded in that matrix, was possible using high-intensity light. The light beams were restricted using a slit width of 1 cm; however, as the light intensity dropped exponentially in the front and back of the image beam, only the brightest lit image plane was used for analysis. This method reduced bias caused by overlapping particles, removed motion blur streaks, and provided more accurate particle size estimates. At the focal plane, the field of view was 3.5 cm tall and 4.7 cm wide. The analysis program for the VPP was expanded from that in Bochdansky et al. (2010) to include more variables for particle characterization (including perimeter, volume and porosity). The VPP can record 30 images per second, with image analysis by a Linux-based image analysis program (an adapted Avidemux video editing software) at high speeds (approximately in real time after retrieval). The images were later aligned with depth from the CTD using time as the common variable and by filming a clock displaying UTC at the beginning and the end of each video sequence. In Matlab, CTD data were matched at one-second resolution with the particle data. The raw data consisted of millions of particles with associated CTD data. These raw data allow us to resample particle metrics at all scales. Particle volumes were calculated as shown in Fig. 2. Instead of assuming a specific geometric shape, the projected area of the particle on the screen (sum of white and black pixels within the perimeter of the particle) was converted into a circle that was then converted to volume. This method reduces error in volume calculations greatly because 2-dimensional information rather than 1-dimensional information is used to reconstruct volumes, thus avoiding the bias of assigning disproportionally large volumes to elongated objects. This approach is widely used in image analysis of ocean particles (e.g., Iversen et al., 2010). Total particle volume (pixel3 frame-1) was approximated by multiplying the mean volume of particles with the mean particle number.