Page 1 of 1
Remote Sensing Reflectance
Posted: Sun Jul 19, 2020 8:41 am America/New_York
Recently, I'm doing some researches on retrieving inherent optical properties, and have downloaded some in-situ data, but there some questions when I used them.
I need the remote sensing reflectance(Rrs) for input, but there are many cruises data doesn't include it, so I try to calculate it.
1. Rrs is defined as the water-leaving radiance(Lw) divided by irradiance just above the surface(Ed0+), but in the NOMAD(Version 2.a), there are only Lw and Es, does the Ed0+ equal Es? But in some cruise, like CLT2005-2007, they seem not equal(Ed0+ from ASD file, and Es from HPro file).
In the same cruise, I also extrapolate the radiance(Lu from HPro file) to the radiance just beneath the water(Lu0-) and then calculate the Lw, and they are far different from the Lw in the ASD file, which quite bothers me.
2. In some cruises just have the irradiance just beneath the surface(Ed0-), and I want to extrapolate it to the Ed0+. I found that they have the relation below:
Ed (0-) = t+×Ed (0+)/(1- ?R)
but I don't know the values of quasi-constant t+ and ??and the Reflectance also doesn't include in the cruise or can be calculated. So I want to know can Ed0+ be extrapolated from Ed0+ directly through other relations without additional measurements.
3. In the FWTIC2010 cruise, there are two Rrs, and they are quite different, especially in the ultraviolet and visible waveband, they could differ several times, and I don't know which should I choose. So does this also mean that the other cruise data include Rrs may have the reliability problems?
Remote Sensing Reflectance
Posted: Mon Jul 20, 2020 1:16 pm America/New_York
1. For NOMAD and SeaBASS validation, SeaBASS staff have typically calculated Rrs from in-water profiles (see: Werdell and Bailey, 2005, and Bailey and Werdell, 2006).
Yes, Ed(0+) and Es are synonyms, and both mean spectral downwelling surface irradiance. In NOMAD it is named Es, though there are 2 possible sources. Radiometric profilers typically measure Ed(z) and are often paired with an above-water Es measurement from the deck of the vessel or nearby buoy. Per protocol recommendations, in most cases, the Es in NOMAD comes from the above-water measurement, but in some cases it had to be extrapolated from profiles of Ed(z). NOMAD’s binary “flag” field includes which data points had above-water Es measurements, although it does not explicitly label whether the reported Es was extrapolated or from the surface.
Regarding your other comment about achieving closure between different types of Rrs measurements: one issue is that surface extrapolations can vary, depending on software and settings used, especially in more complex waters. As just one example, the depth range(s) selected for the extrapolation can cause substantial differences. A separate issue is that data from above-water hand-held instruments such as an ASD can be useful but are generally less reliable or consistent compared to other methods of measurement; historically we have opted not to include data from hand-held above-water radiometers in NOMAD. The SeaBASS team was recently asked to add a new feature to our search engine to allow more differentiation of measurements to make it easier for users to select whether they want types of above-water measurements included in search results, so I expect we’ll add that later this year.
2. If you have Ed(0-), you can propagate across the sea-air interface to Ed(0+) via this simple equation:
Ed(0+,lambda) = Ed(0-,lambda) / 0.96
It can be challenging to extrapolate Ed(z) to Ed(0+) accurately in the case of a float or buoy without depth-resolved measurements to help account for rapid sub-surface fluctuations in the light field, and waves can complicate assigning an exact z to measurements just centimeters below the surface. Some additional uncertainties will be introduced if trying to work with an Ed(z) measurement if z is a single fixed depth below the surface.
3. Some cruises have AOPs measured from multiple instrument sources, in which case their use must be evaluated case-by-case. Near-IR and/or UV portions of spectra can be difficult to extrapolate well due to rapid attenuation and low signals. As a side note, ocean color protocols recommend “multi-cast” measurements where profilers repeatedly measure the near-surface depths several times in succession, or in other cases entire deeper profiles are measured repeatedly in succession; all such measurements may be evaluated. So, even if a cruise doesn’t include AOP measurements from different types of instruments, a strategy is typically needed for downsampling original data that were measured nearby in space and time. The NOMAD dataset and Validation Search results have already had that performed.
Just FYI, the FWTIC2010 cruise data you mentioned are especially complicated because they involved measurements in optically shallow conditions where bottom-reflectance was likely present. Such measurements can be complex and aren’t suitable for all types of analyses (e.g., they are not suitable for standard algorithm validation or in NOMAD). Since SeaBASS was originally designed around validation, its File Search engine does not return these files unless you manually enable the “Optically Shallow” option under Data Warnings on the search form. Additionally, such files contain a relevant “/data_use_warning” in their headers.
Remote Sensing Reflectance
Posted: Tue Jul 21, 2020 9:04 am America/New_York
Thanks for your patient reply, Chris! It helps me a lot, but there are some questions more I want to ask.
1. Considering the hand-held instrument may contain much more uncertainty in measurement, so if there are both hand-held measurements and others, like in-water profiles, what is their priority and which should I choose. As you mentioned that the NOMAD typically calculated Rrs from in-water profiles, but it seems there are some cruises in NOMAD(like rb-01-02) that choose the surface measurements rather than in-water profiles to calculate the Rrs, though I found there is only a small difference between in-water profiles and surface measurements.
However, for cruises CLT2005-2007, there are both in-water profiles and surface measurements for Lu(z)\Lw and Ed(z)\Ed(0+), and I found even for the good extrapolate Lw is still quite different from the Lw from ASD(like station 2005051806). By the way, does the Lw calculated by Lt-?Lsky is more robust and reliable than the extrapolate Lw?
In the same cruises?no matter the Es or the extrapolate Ed(0+) from in-water profiles(from HPro) is quite different from the Ed in ASD file, some 2 times larger(like station 2005051802), and some 2 times smaller(like station 2005051803), so I wonder the Ed in ASD means Ed(0+) or not, though it was combined with the Lw to calculate the Rrs.
In the Standardized Fields and Units page, Rrs seems to be defined as the remote sensing reflectance just above the water(Lw/Ed), but in the cruise AMMA-RB-06, Rrs is calculated by Lu(z)/Ed(z), and in cruise I8SI9N, Rrs in 0 depth represents the remote sensing reflectance just above the water and seems to be calculated by Lw(extrapolate from Lu(0-)) divided by Ed(0+)(extrapolate from Ed(0-), though there is Es), and the rest are calculated by Lu(z)/Ed(z), but there are no comments given for the same field symbol but different definition in both two cruises, which might mislead the user.
2. I have downloaded some cruise contains Ed(0-) and Es and done some tests. The conversion coefficient 0.96 won't cause much error for the wavelength between 400nm and 600nm, but for the shorter and longer wavelength, it becomes unstable and seems to have some dependence on wavelength, so should I abandon them?
But I fonud the equation won't work in cruise AMMA-RB-06, the conversion coefficient change a lot for different wavelength, especially in UV portions of spectra, the Ed(0-,325) is about 30 times larger than Es(325), this phenomenon is quite strange. Is it possible that there is something wrong with the measuring instrument or data generation? But there is not any description in the document or header.
3. I am a little confused about the definition of optically shallow water, which typically refers to waters that the Rrs is influenced by the bottom, but I don't know how to define this degree. Does the Ed(z) reach the bottom decrease to 1/e or 10% or 1% of Ed(0-)? There are some data doesn't have the '/data_use_warning= optically_shallow' metadata headers, but for some wavelengths(green for Case I waters, yellow for Case II waters), their first optical depth are larger than the water depth, so should I use them cautiously?
Remote Sensing Reflectance
Posted: Thu Jul 23, 2020 2:41 am America/New_York
1. There are a variety of instruments, such as profiling radiometers, floating/tethered radiometers, and above-water radiometers (mounted on a fixed platforms or ships, and those that are handheld.) It is difficult to generalize about their quality and priority of use, since it’s possible to collect great measurements with any of them, and there are many factors that can impact the quality of the data. Greater uncertainties are often found with handheld measurements, so such decisions depend on your needs.
Regarding Rrs calculations, good points, especially about AMMA-RB-06 Rrs that don't match the definition. The SeaBASS team performs its own calculations of Rrs from profiles using in-house software for consistency for use in validation and NOMAD, so those "not-real-Rrs" aren't in those processed datasets, but those files are confusing for other users. We'll flag that issue in those few older files for follow-up. SeaBASS has tried to improve its documentation and other reporting requirements over time, but sometimes things have slipped in.
2. My equation notation might have caused confusion. Just to clarify, simply divide Ed just below the surface by 0.96 to estimate Ed just above the surface, AKA Es. I wrote lambda to signal spectral dependence, but it isn’t an active variable in the equation so it should scale all wavelengths appropriately.
Otherwise, such large differences could be caused in a specific file by problems with a sensor, but also note that sometimes it isn’t possible to perform a good extrapolation, e.g., when the signal is very low due to attenuation (which is more likely to happen at UV or NIR wavelengths). I share your concern that specific data or points may need to be flagged or removed from your analysis if there are large differences that can’t be reconciled or explained by comments or metadata.
3. It is reasonable to perform a check for optically shallow water as part of your analysis (referencing the recorded water depth, or a bathymetry source). Much of the radiometric data in SeaBASS were collected with validation in mind, but there are some files that could contain bottom reflectance. The optically shallow warning was added as a label for submitters to provide in their files starting several years ago. Prior to that submitters were discouraged from submitting measurements known to be optically shallow, but there are at least a few in the archive, and no older files from prior to 2016 currently have that warning applied.