Paper Review : Spectral Reflectance of Marine Macroplastics in the VNIR and SWIR measured in a Controlled Environment
Paper review (clap clap clap), damn, I feel like Pewdiepie typing this out. Well, I have recently begun work on Hyperspectral Imagery, and I came across this paper that works precisely on what I need to. Let us play some music and go, I guess! I am listening to 1Take (Naija to London). Oh, I almost forgot the link to the paper.
"While at least 8 million... the spectral properties of wet as well as submerged plastics in natural marine environments are still largely unknown... Finally, we demonstrate in an example how to use the open-access data set driven from this research as a reference for the development of marine litter detection algorithms."
I have included this line from the short excerpt given before we ride into the paper because, firstly, this is the first time I am reading up on plastics. This means I did not really, before reading the paper, know that we do not have enough data on the spectral properties of wet and submerged plastics in the marine environment. Also, I intend to work on hyper/spectral imagery of plastics, so open-access data is a gold mine to me.
Before we understand any paper, it is nice to understand its motivation. We all know plastics pose a problem, and so their mapping is necessary. Why use a satellite and imagery for it? Well, let us look at the old techniques. First up, visual surveys from ships are long, time-consuming and expensive. Okay, what about plankton neuston net trawls? Wait a minute. What does this even mean? Well, a simple google search revealed that it is just net trapping marine life in the upper 10cm high water column and studying the organisms. But it requires a lot of sampling in order to get true data. So, in comes remote sensing. Alas, you can't have your cake and eat it too. Remote sensing methods have their limitations. It is limited to monitoring of aggregated marine plastics only and should be used complementary to the ship, net trawl and visual survey methods, as I vaguely quote from the paper.
"The chemical composition of the plastic polymers is derived from the spectral measurements in the SWIR (1000–2500 nm) to MWIR (2500–5000 nm) spectrum as well as LWIR (6000–14,500 nm)... The water is absorbing the light strongly in the NIR and SWIR and other water constituents such as suspended particulate matter alter the refected signal. Moreover, there is also a contribution from the atmosphere which also alters the signal received at the sensor. Finally, water and small individual plastics can be in the same pixel complicating the detection."
For those who might be casual readers, SWIR stands for short-wave infrared, MWIR is medium-wave infrared, NIR is near-infrared and LWIR is long-wave infrared.
Before we move ahead with the paper, let us look at why infrared? And, go through the very basic principles of remote sensing.
A profile to understand what the terms mean.
I while try and very quickly, simplify stuff to explain how basic remote sensing works. So, the sun sends in mostly the visible spectrum. Other stuff also comes but the energy peaks in at around visible and infrared.
This energy reaches the earth's surface, now some does get absorbed, and some gets reflected. The reflected part maintains characteristics, but the earth only emits back in longwave.
Therefore, in accordance with the surface temperature of the earth, you also now get some longwave radiation. This radiation also gets altered numerous times as it passes the layers of the atmosphere. And this poses us a new situation in our plastics case.
Let us move back to the paper now. See, not only does the signature of a plastic piece get altered by marine stuff such as whitecaps, seafoam and bubbles. We also need to account for medium alterations and water body flow.
Well, we do have a saviour.
"Garaba and Dierssen performed some experiments on harvested plastic samples taken to the laboratory (items such as buoys, bottle caps, containers, ropes and nets) and on virgin pellets (PVC, PA 6.6 and PA 6, LDPE, PET, PP, PS, FEP, ABS, Merlon, PMMA). The study revealed the presence of diagnostic absorption features in the NIR-SWIR spectrum centred around 931 nm, 1045 nm, 1215 nm, 1417 nm, 1537 nm, 1732 nm, 2046 nm and 2313 nm. They reported a decrease in reflectance magnitude after the dry plastics were dampened."
Next, we have some value work that does not really need much discussion. It is of immense use while making an algorithm.
Something important from the discussion is, "Several absorption features are found in the SWIR region, as in the literature, but the exact position of the features can be slightly different. For instance, Garaba and Dierssen, mentioned features centred at 931, 1215, 1417 and 1732 nm, whereas we have found them centred often at 1192, 1413, 1730 nm. The small shifts in the wavelength should be examined further. Possible explanations could be the specific type of plastic with coatings and additives or the lab conditions."
This is quite important. I can take up some random plastic bits and pieces and test them in a lab in India but using that signature set to determine plastics on the coasts of England is liable to give me wrong values, as the types of plastic, the industrial coating levels and lab conditions are very different. This needs to be accounted for. And a possible method could be an extensive open-source data set with data from every corner of the world. At least, that is what my naive scientific acumen suggests as of now.
Overall, the paper did give me a good idea of what I am looking ahead to and would have to work on. It was my first paper reading on plastics and non-weather related remote sensing, and I did enjoy it. I would have loved to be able to understand the paper at some points better, and I guess I will as the time comes.
What I am happy about is this,
Also, before I end this piece, I found their mechanism of experiment quite interesting, and I wanted to share it here.
Expect to read a lot on imagery here, as I move into the field and properly set foot. Gracias!