Optical calibration, bathymetry, water column correction and bottom typing of shallow marine areas, using passive remote sensing imageries
Busy? 4SM in 10 lines
4SM: the way ahead : automatic 4SM

There is substance here for good applied science.

At the age of 68, with plenty of time as a retired geologist: we need to talk.

VNC : let's work together live!

1 - NO NEED for field data, nor for atmospheric correction
2 - this is demonstrated in this website, using a variety of hyper/multi spectral data
Requirements are
1 - homogeneous water body and atmosphere
2 - some coverage of optically deep water
3 - some coverage of dry land
Problems are
1 - the precision on estimated depth is found wanting, because the noise-equivalent change in radiance  of accessible data is too high for shallow water column correction work 
2 - radiance data should be preprocessed by the provider at level 1 in order to improve S/N ratio
3 - exponential decay: the deeper/darker the bottom, the poorer the performances
I keep digging
until suitable data
become available

Your comments are invited                            See you on 4SM blog        


I see a potential for developing 4SM into an automatic process.
I want to contribute to this effort. I also want to develop a graphic interface.
After many years in isolation,
I might as well release 4SM under some flavor of GPL license. 
I just need some encouragements and support.

we need to talk 
My wishlist (outdated)
accumulate experience and burn it into the 4SM  project
  • Some money/resources: your commitment as SuperUser requires that I am connected and supported
    • Transfer: hopefuly someday I'll need to engage into a business plan with a business partner
    • if I seem to be unable to meet your expectation/requirement
    • if I get bogged down with 4SM code or my computer setup
    • Computing: I need access to a computer proficient person as/when required
    • for big upload or download, I need to go to a nearby internet lounge or Wifi spot
    • from wherever I happen to be roaming, which includes France, Tahiti and New Zealand
    • internet VOIP telephony,  
    • updating my WIFEO website on a regumar basis , 
    • Communication:  I need 3G mobile internet on my laptop, for email, 
  • Sea truth: that's what is most wanted.
  • Your cooperation, commitment, resilience as a SuperUser: that comes next!
  • Exposure: links from my website to your own reporting of climbing the 4SM learning curve on your own website.
  • Worldwide: develop experience over a wide variety of shallow water environments worldwide.
  • Band setting: develop experience over a wide variety of multispectral and hyperspectral imageries.
  • Wavelengths: consolidate a database of specific operational wavelength for each waveband of all possible multispectral imageries.
  • Reflectance: calibrate spectral LM=LsM-La in units of reflectance through offset and gain values, and vice versa.
  • Spectral Soil Line database: map variations of operational spectral shallow water LM along the coasts worldwide, in units of reflectance. This shall take the form of a series of LMWL values in units of refelctance, to be attached to a particular location on site, or to a particular generic type of bright beach material (coral sand, etc). This should be enough to specify the local Soil Line for any spectral image down the line. 
  • Spectral Knir database:  this exercise should  yield a fairly secure set of  Knir values for all wavebands involved.
  • Bottom typing: this is the ultimate goal of 4SM, of course
    • if we can prepare a bottom typing map ahead of any field work, sure the planning of that field work shall become much more reasonable and rational.
    • provided the set of Ki/Kj ratios is secured from the image itself, then the bottom typing does NOT depend on any seatruth on computed depths.
    • all that's needed then is to work out the labels for bottom types identified, and reduce/bin the list of arguably distinct types.
ToDo list
  • 4SM
    • develop bi-dimensional histograms under AOI polygon
    • work at a printable version of 4SM website
    • 32/64 bits computing
    • compile 4SM in 64 bits
    • someone to compile all 4SM calls and OpenEV in 64 bits
  • OpenEV: someone to contribute to the project in order to 
    • improve some aspects of OpenEV that are frustrating, like
    • relative path in .opf projects
    • random order in imaqge/compose
    • etc
    • add functionalities that we need for 4SM
Automatize the 4SM process
  • by all means, avoid images captured with oscillating mirror sensors, or face the consequences
  • Lsw: develop a way to estimate Lsw automatically from the image. This is easiest in the red-nir wavebands.
  • La
    • assuming atmospheric adjacency effect does not come into play
    • assuming spectral LsM is available, 
    • assuming spectral Lsw is available, 
    • assuming La~=Lsw in the red-nir part of the spectrum, 
      • sampling the spectral Soil Line may be done automatically for the pair of red and nir bands of the image. 
      • this is what is done at present in 4SM under supervision: Soil Line by mSE mask value set manually at mSOIL by the practitioner : soil line  
      • for 4SM to do it on its own, it would need algorithms to:
        • read applicable local spectral LM values for all  wavebands of the image from the database of spectral soil line in units of reflectance.
        • convert these values into units of DN of the image.
        • derive the value of slope LMred/LMnir which applies to the image. 
        • this defines the Soil Line in a scatter plot for this pair of bands in the image. This soil line is the straight line which starts from spectral La (~=the TOA radiance of a black body at null depth) and reaches through spectral LsM=LM+Lsw (~=the radiance of the brightest shallow bottom type at null depth at the top of the atmosphere) with a slope value of  LMred/LMnir.
        • apply a tolerance of ~1.18 to the value of this slope: slope*1.18 and slope/1.18; in other words allow for a certain level of fuzziness of the soil line in the image, as it is well known that the system noise increases as the radiance level increases.
        • select all spectral pixels of the image which plot inside this tolerance in a scatter plot of Lred=Lsred-Lswred versus Lnir=Lsnir-Lswnir. 
        • Note that all shallow water pixels plot on the red side of the soil line: 
        • this provides a mean for segmenting the image into dry lands, whether bare or vegetated, and marine areas
        • we also need a similar way to segment clouds out.
        • work out the average spectral soil line for all wavebands
    • in the scatter plot of Li versus Lnir, determine spectral Lai for all bands i through a simple geometric construction.
      • this is usually fairly straightforward provided the image exhibits some bare land, from dark to bright. 
      • in case the image is totally under water, chances are that the Soil Line may be set to be parallel to the Darkest Pixels Line:   Davies Reef.
  • Lw: derive spectral Lw=Lsw-La.
  • Kgreen/Kblue and spectral K
    • select the Brightest Pixels Line for all pairs of bands from cloud free marine areas of the image.
    • this does not need the image data to be deglinted, through the LsNirMaxBPL parameter of the -Extract command line argument. 
    • linearize the Brightest Pixels Line data for all pairs of bands.
    • work out the best fit of  a straight BPL line with the observed BPL pixels for the Blue/Green pair of bands: its slope if the ratio Kblue/Kgreen.
    • use Jerlov's data to interpolate the intermediate marine water type which applies to this slope.
    • use this slope to derive spectral K for all visible wavebands. 
    • check  that these values are acceptable.
  • Automatic deglinting:
    • provided there are some glinted areas over optically deep waters, it should be possible to ascertain the Glint parameters through some automated process.
  • Modeling: 4SM should now be ready to
    • apply the above spectral calibration parameters : WL, Lsw, La, LM, spectral K
    • to deglint the current pixel
    • and operate the inverse simplified RTE   LBWL=LwWL+(LsWL-LswWL)*exp(KWL*Z) 
    • by increasing Z
    • until the spectral water column corrected soil line LBWL
    • is deemed to fit reasonably well onto the spectral Soil Line LMWL for that image
    • for each shallow water pixel of the image.

Busy? 4SM in 10 lines