Optical calibration, bathymetry, water column correction and bottom typing of shallow marine areas, using passive remote sensing imageries
busy? see 4SM slides

Breaking news:
using a panchromatic band
Using the Panchromatic band for water column correction
to derive water depth and spectral bottom signature:

Landsat 8 OLIP bandset used for this work

Purple=1Blue=2Green=3PAN=4Red=5NIR=6 and SWIR1=7


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
Waoh: pansharpened Landsat 8 for shallow water work
The PAN solution
Calibrating the PAN solution
The advantages of the PAN solution

refer to PANCHRO at Caicos Bank
The "PANchromatic solution"
for shallow water column correction and depth retrieval

The "PAN solution" uses the PANCHRO band,
or a synthetic hyperspectral PANCHRO band,
along with regular multi/hyper-spectral bands

OLI bands 1, 2, 3, PAN, ...6
        become     bands 1, 2, 3, 4, ...7
WV2 bands 1, 2, 3, PAN, ...8      
      become     bands 1, 2, 3, 4, ...9
State of the art is (march 2015):
  • WV2 PAN and MULTI are not properly co-registered (time delay). This hurts for deglinting of PAN band.
  • The PAN solution is far more desirable than any MULTI/HYPER solution for 4SM water column correction

Using PAN-sharpened images
  • YES: shallow water work benefits a lot of using a pansharpened image
    • pan-sharpening is expected to preserve the spectral properties of multispectral bands intact, while reducing the footprint of the pixel
  • BUT: this needs to be thoroughly investigated
    • which PAN response curve
    • which pan-sharpening method: FuzeGo is what we want
    • co-registration must be impeccable
    • system noise: why 8 bits?
    • what of the NIR band
    • deglinting is a concern
    • GoogleEarth-derived imageries: waoh!
  • This is illustrated at
  • Pleiades should be the next candidate

SPOT response curves

Landsat 7 ETM response curves


IKONOS response curves


Landsdat 8 OLIP  response curves


QUICKBIRD response curves


WV2 response curves


the advantages of using the PAN solution
  • Practical: using the PAN solution is feasible and profitable:
    • on WV2 and Landsat 8, it only uses Coastal, Blue, and Green bands against the PAN band for retrieving depth
    • on hyperspectral image, a synthetic PAN band is created to cover most of the visible range
  • Generic: based on the optical calibration for the Multi/Hyperspectral bands, the optical calibration of the PAN band is derived, using the specific PAN response curve and Jerlov's data.
    • this is a now generic and automatic process, hard coded in 4SM, which requires minimal manual control
    • it may be used for any brand of imageries, only requires the specific PAN response curve
  • Detailed shallow:
    • unlike Stumpf et al's method, it yields very detailed results in the shallow range 0-5 m
    • although a well adapted response curve for the PAN band should help a lot.
    • in this regard, the QUICKBIRD panchro response curve appears to me to have the best potential, as it cover generously the NIR range and also part of the BLUE range as well: this needs to be investigated and passed on to sensor designers.
  • PAN solution is safer than the RED solution in case of local turbidity
  • Improved depth retrieval for  OIII and all Coastal water types:
    • Ki/Kj>0.8: when Kblue~=Kgreen , i.e. when the ratio Kblue/Kgreen exceeds ~0.8, there is poor color separation
      • therefore the conditions are not favorable for quality depth retrieval and water column correction.
      • this is the case of OIII and all Coastal water types.
    • Ki/Kpan<0.8: this ratio remains well below 0.8 even in the case of OII and all Coastal water types.
  • Reasonably smooth: it yields reasonably smooth results:
    • the level of system noise is quite low in the PAN band. This is particularly true for the synthetic hyperspectral PAN band in the 0-5 m depth range where no smoothing is required.
    • the computed optical model for the PAN band is a continuous smooth curve, which gets quite noisy at depth though
  • No more gaps: no more of these anoying gaps that show up when using the MULTI solutions in 4SM: removal of the water column effect is much improved
  • Homogeneous bottom typing: this allows for richer and more reliable bottom typing results
    • see Bottyom typing at Princess Cays
    • on a WV2 or Landsat 8 image, bottom typing is achieved using only the water column corrected Coastal, Blue and Green bands: this ensure a most homogeneous process over the whole depth range
  • PAN sharpened: paves the way for water column correction of PAN-sharpened images
    • this should appeal to users of copyright-free Landsat 8 images
    • this should appeal to the military for REA
    • this should appeal to engineers in charge of designing new sensors for the shallow coastal environment
  • No need for
    • formal atmospheric correction
    • conversion to calibrated reflectance
    • any field data for optical calibration
    • spectral libraries of bottom type endmembers
    • costly proprietary software: 4SM does it all
    • costly computing setup: a consumer's laptop running LINUX does it all
  • BUT time delay is a real pain
    • MULTI and PAN bands should be acquired exactly at the same time, without any time delay
    • and co-registration should be perfect
    • see WV2 at Oahu and OLI at Andros Island


  • Since working in 1996 on a SPOT 1 XSP image of Bora Bora in French Polynesia, I know that the "Pan solution" shall someday find its way for computing shallow water depth and achieving the water column correction of multispectral imageries.
    • I was able to achieve water column correction down to ~10 m in these clear lagoon waters: YES, using only XS and PAN.
  • That was confirmed in 2001, by working on a SPOT 5 PAN-sharpened XSP image of Sanaa in the Red Sea.
    • I was able to achieve water column correction down to ~8 m in this not-so-clear coastal water type
  • Then in 2012, I tried using a WV02 image of Marmion Marine Park, Western Australia: too difficult.
  • Then came Lansdat 8 OLI data in 2013:
  • Then I worked in 2013 on WV02 images of
  • Then I worked in 2014 on a PAN solution for hyperspectral CASI data
  • Then I worked again in 2014 on the PAN solution for WV2
  • Then I worked on a Hyperion hyperspectral image
  • Then in january 2015, I worked on HICO hyperspectral data
  • Then in march 2015, I worked on a PAN-sharpened Quickbird image
  • The process is now under full control, and burned into 4SM code.
  • Water column correction is achieved down to ~30 m in very clear waters.
  • Then in july 2016 I worked on a time series of 18 Landsat 8 OLIP scenes


The PAN solution

in addition to 
the "Green solution", the "Yellow solution", the "Red solution", the "Red-edge solution"

++++++++++Typically in 4SM++++++++++
in the case of a WV02 imagery:
  • the general inversion formula uses BOA radiances in units of uncalibrated DNs
    • LBi=Lwi+ (Li-Lwi)*exp(Ki*Z)
  • a "RedEdge solution" may be be operated for very shallow bottoms.
  • a "RED solution" is operated for pixels that have significant bottom signal in the red band
    • Z is increased until the ratio (LB1+LB2+LB3+LB4)/(4*LBRED) achieves an acceptable fit with the SOIL Line observed in the image data. 
    • This yields  Z, and LB1, LB2,  LB3,  LB4 and  LBRED
  • a "YELLOW solution"  is operated for deeper/darker pixels that have significant bottom signal in the yellow band, using the ratio (LB1+LB2+LB3)/(3*LBYELLOW). This yields
    • This yields Z, and LB1, LB2,  LB3 and  LBYELLOW
  • a "GREEN solution" is operated for  deeper/darker pixels that have significant bottom signal in the green band, using the ratio (LB1+LB2)/(2*LBGREEN). This yields
    • This yields Z, and LB1, LB2,  and  LBGREEN
++++++++++The trouble is++++++++++
  • the calibration is often less than perfect.
  • the deep water radiances tend to vary across the image.
  • the SOIL Line assumption is a sheer simplification of the reality.

  • This results in conspicuous gaps in the raster image of retrieved depth, and ominous discrepancies in the water column corrected images along the seam_line between solutions: see  illustration at Princess Cays.
++++++++++The PAN solution++++++++++
alleviates most of these difficulties, and beyond.
In my view, it is the ultimate solution.
  • The"PAN solution"  is operated for all pixels that have significant bottom signal in the panchromatic band, using the ratio (LB1+LB2+LB3)/(3*LBPAN).  
    • This yields Z, and LB1, LB2,  LB3 and  LBPAN , and other bands as well, depending on the water depth
  • it carries all the benefits of the Red, Yellow and Green solutions, without the attached problems.
  • it covers the whole range of the NIR, Red, Yellow and Green solutions, yielding real water column correction over the whole 0-30 m depth range in clear Bahamas waters:
    • see illustration at Princess Cays.
    • as a result, bottom typing may/shall use all the information carried by the water column corrected Purple, Blue, Green and PAN bands over the whole depths range, which covers approximately half the depth range of the Blue band (subject to PAN response curve).
  • it even helps fine-tune the deep water radiances Lsw and the water volume reflectances Lw.
it should work fine with PAN-sharpened images

refer to PANCHRO at Caicos Bank
refer to PANCHRO at San Lorenzo Channel, Baja California