4SM tutorial
on a 500*500 Landsat TM
4 bands image (1985)
of Negril Shores, Jamaica,

UTM georeferenced with sea-truth dataset

Courtesy of Tomy Lindell, University of Upsala, Sweden.    

Also refer to work done in 1998: Erim Marine 98

work in progress

back to 4SM tutorial        next to Prepare your image           
    Quite a difficult image
  • too small, not enough shallow pixels
  • 30 m pixels are very large; only 70 pixels from beach to deep waters
  • fairly dark TM2 band: low dynamic range
  • striping
  • faithful bright sandy pixels fairly scarce
  • 0-5 m depth range poorly represented
  • possibly turbid waters along the beach
  • heterogeneous open waters
  • no swell-modulated glint
  • non-DGPS sea-truth data




First let's take a look at the results I finally obtaned
through an intnsive use of seatruth field data



Sea truth correlation: r 2=0.73 RMS=1.40 m
This improves markedly on results obtained in 1998.

neg85 mapZ   is the depth contour map obtained in 1998.


Three depth profile

Red: computed depth
Black: Sea-truth depth
Then we shall try to see, step by step, whether it is possible,
 to reach there without using that field dataset.

Several key points have to be addressed:

        • Which pair of bands for measuring the ratio KiKj?
        • What wavelengths?
        • What Lsw values?
        • What La values?
        • Is the water body homogeneous? 
        • What smoothing scheme?
        • What about the electronic noise?
        • SIG or not SIG?
Flow chart
Bottom typing was the ultimate goal of this venture.
Whether it can be of any help for coastal management remains to be seen.




Créer un site
Créer un site