Below there are 3 videos courtesy of my mentor Anouk Ollivier, from VLIZ. I highly recommend you see them briefly, to better understand the coming parts.
WP2 net: Shows how the sampling process works using this tool, as well as the lab work after the campaigns.
VPR: Shows how the VPR works during a real oceanographic campaign.
ZooSCAN: Shows how the samples are scanned, and how their datas are obtained.
As you watched, the WP2 net catches the organisms bigger than the net holes, which usually are around 200 μm. These organisms collected in a collector at the end of the net are transferred to a labeled jar and stored.
Once in the laboratory and as it's shown in the third video, the samples are scanned using the ZooSCAN. Having as result image files with all the organisms present in the sample.
After building a good image library using scanned images of zooplankton, we are ready to use the Plankton Identifier program, which classifies plankton according to its algorithm. However, this algorithm doesn't always work and the organisms are misclassified.
Now talking about the awesome VPR, as you could watch in the second video, samples are not taken, but the camera photographs organisms directly underwater, and then using an algorithm the images are automatically classified . We can say we are scanning the ocean in this way!!! ... Just amazing. Nevertheless, we have the same setback with Plankton Identifier, we have to build a learning set first in order to ensure a proper classification. The more classified images, the more precision the algorithm will have for future classifications.
In the first pages I mentioned something about a method comparison. Can you guess what I meant? Yes, the WP2 and VPR comparison!
The organisms that appeared in the wrong folder have to be transferred to the correct one, so I checked hundreds of folders with thousands of images, looking for infiltrators.
Also, once all these folders are checked, they can be added to the algorithm library to improve it, refeeding its precision to avoid similar errors in future classifications.
Unlike correcting ZooSCAN images, VPR images do not require any program, you just need to move the raw images to the corresponding folder according to the organism.
Here I paste an example of what I saw during the classifications. Can you distinguish anything? I know... it's almost impossible. However, if you invert the colours of your screen you'll be able to see the organisms better, as long as there is an organism, since almost 98% of the images were marine detritus.
In order to compare the two methods, I had to know their sampled water volume and number of organisms. The WP2 net sampled a water column of 21 m with a diameter of 0.58 cm, on the other hand the VPR takes 25 photos per second, being the total sampled volume 36 ml per second and taking into account that the VPR ran for 2:20 hours. The total amounts of images were 35136 for VPR and 3355 for WP2.
Now that you know the main aspects, which is the best method to sample zooplankton in the BPNS? This question was what I had to answer during the last part of my PP, and if you want to know the response I encourage you to come to my presentation session on July 2nd from 12:30 to 14:00 CET. ;-)