GISdevelopment.net ---> AARS ---> ACRS 1989 ---> Oceanography

Water depth determination from satellite data

Dr. Mohd Ibrahim, Seeni Mohd
Centre for Remote Sensing Faculty of Surveying
University of Technology Malaysia, 80990 Johor Bahru .


Abstract
The coastal waters of importance in navigation. However, inspection of hydrographic charts of these waters reveal a large number of doubtful soundings. The task of updating hydrographic charts using conventional techniques alone is costly and time-consuming. Remote sensing from satellites appears to hold promise for obtaining depth information in shallow coastal waters.

In this paper the physical principles, an optical model and algorithm for water depth determination from remote sensing techniques are given. The results from a study carried out in the coastal waters of Penning island in Malaysia using the Landsat -3 Multispectral Scanner data are also presented.

Introduction
According to the International hydrographic Bureau's estimates, sufficiently adequate sounding to determine sea floor topography only exist for about 16% of the area covered by the world's oceans. Another 22% of the area has data sufficient only for the determination of major sea floor features; while for the remaining 62% there is not enough data for deterinining sea floor topography (Kapoor 1976). Watson (19860 notes that the situation is still very much the same since bathymetric surveying by conventional shlpborne sounding techniques is slow. Hazardous and expensive. Interest has been generated in the application of remote sensing techniques at least in the critical shallow areas, which are frequently used, by ships approaching or leaving ports or harbours.

The possibility of using remote sensing technology was addressed as early as the late 1960s (Brown et at, 1971). These studies led to the NASA /Cousteau Ocean Bathymetry Experiment in 1975 which demonstrated the feasibility of using landsat high-gain multispectral scanner (MSS) data in bathymetry (Polcyn 1976). Since then, a number of studies have been carted out

Physical principles, optical model and algorithm for depth determination from remote sensing.
  1. Physical principles

    The radiation reaching a satellite or aircraft sensor is a function of some radiative transfer processes in the ocean-atmosphere system. These processes are described by Sorensen (1980) as follows (see Figure 1):

    1. single or multiple scattering of solar photons by the atmosphere into the field of view of the sensor.

    2. Reflection of unscattered or scatiered photons from the surface and subsequent scattering into the field of view of the sensor

    3. Reflection of unscattered photons by the instantaneous field of view (IFOV) under examination and subsequent propagation (with losses) to the sensor

    4. Reflection of single-and multiple-scattered solar photons ()with losses to the sensor

    5. Scattering of single-and multiple-scattered or unscattered solar photons (which have penetrated the sea surface) by water and suspended particulate and subsequent propagation (with losses ) to the sensor.

    The total radiance seen by the remote sensor is the sum of the radiance's due to processes 1 and 2 (the path radiance. Rp) plus the sun of the radlances due to processes 3,4 and 5 (the surface radiance, Rs) after accounting for the transmission losses, l.e.


    Figure. 1: Processes contributiong to radiance at the sensor(after Sorensen 1980)

    R = Rs.ta+Rp -----------------------(1)

    Where ta is the appropriate transmittance through the atmosphere. The su surface radiance Rs is the sum of the reflected radiance Rr and the ocean-leaving radiance Ro, i.e.

    Rs=Rr + Ro -------------------------(2)

    In applications over land areas, Ro is absent and Rr is the quantity which is desired Conversely. In most applications over water. Information concerning sub-surface conditions is sought . in requited. Thus. For land viewing it is necessary to remove Rp and multiply the remaining signal by ta -1. in the ocean viewing it is necessary to remove Rp+ta. Rr and multiply the remaining signal by ta-1. In the ocean case the remove cont :.ibution from process 3 is usually called sun glitter (or glint). Also, the radiance Ro can be derived from the upwelling radiance lust beneath the sea surface. Rw. Through

    -----------------(3)

    In which p(v) is the reflection from beneath the rough ocean surface (in the case of a smooth surface, p(v) is the Fresnel reflectivity of the interface) and n is the refractive index of water. Only 5-20% of the signal received over water by a sensor on board a satellite or a high-flying aircraft stems from the atmosphere and the surface, all of which must be quantified and subtracted over eater by the total signal (Sorensen 1980). The interesting part of the radiance value measured over water by the remote sensor is therefore the total signal minus that contains the oceanographic information. Will at most comprise values from three sources the bottom suspended and dissolved, matters in the water and the water itself. For bathemetry. It is the signal from the bottom that is important.

  2. Optical Model

    A number of models can bc used. However, The model of polcyunnd lycu (1975) is considered here. This model is as follows:

    R = Rp + ta Rg + ta Ri rb -------(1/n2)exp {-a(secd + Sec ds)|z|-----------------(4)

    where
    R= radiance at the detector
    Rp= atmospheric path radiance
    R1= radiance on the sea surface
    Rg= radiance from the sea surface
    N1= telleetance of the scabed
    ta =transmittance of the column of atmosphere below the satellile
    a= allennation coffielent of light in water
    n= refraetive index of water (approx. 1.33 ).
    d'= apparent angle of observation under the water
    d's= apparent solar zenith water
    Z= depth of water
Case study - penang lsland
A study was carried out on the coastal waters of Penang island using the landsat 3 Multispectral Seanner (MSS) data. The study area is shown in Figure 2. The turage was acqured on 10 January 1979 at about 02:52 hours GMT by the landsat 3 MSS when the hright of the tide was 1.6 in above lowest astronomical tide. Geometrical recttication. Depth used and some results that were abtatned are presented.

  1. Geometrical rcctification and depth algorithms

    1. Geometrical rectification

      The sub-scene used in the study was geometrically rectified to enable quantllative comparison to be made between the remotely -sensed image and existing maps and charts. The following relationship was used for this pumpose. Where E is the Easting (Longitude), N is the Northing (Lattude). S is the scantiar number and I is the column or plxcl number on the CCT.

      P=C1+C2E+C3N+C4E2+C5EN +C6N2 -----------------(5)

      P=C7+C8E+C9N+C10E2+C11 EN+C12N2 ---------------(6)

      The coordinates of ground control points (gcps) which are well delined on the image were used to determine the values of the coefficients c1 .C2……C12 in these equations between 10 and 30 points were used to perform a least - squares fit to obialn the best values of the coafficients .

    2. Method of depth determination and depth algorithms

      For the remotely -sensed data used in the study. pixel intensittes (digital numbers) were extracted at some points of known depth (calibration depths) in order to algonthus relating plxcl intensitles to the dept. The pixel intensities at these alienation depths were obtained by transforming the geographical positions of these points into scanline and pixcl number (column number) by using equations (5) and (6) and subsequently using computer programs to read the inteusllies in telesant hands at these values of scanline and pixel number, A least squanes mintulsation was carried out relating plxcl intensllies to depth to obtain the best valuers of the unknown coefficients in the algorithms used. Having determined the coefficients. The depth at any point on the particular hange can be obtained by using the values of these covert ion and the plxcl intensely values at this point from the relevant bands.

      Ashnple algortlun based on the model described by equreton (1) was used.

      1 = A1+A2 exp A3Z -------------------(7)

      Equation (7) can usefully he expressed in the allernative form as below

      Z = (1/A3[In A2 - In (A1)]-------------------(8)

      These equations express the expected exponential relationship between pixel intenslly, I and depth, Z A1 A2 and A 3 are the unknown ceficients.


  2. Results

    Pixel intensities on bands 4(0.5 - 0.6 mm) and 5(0.6-0.7mm)these equations express the expected exponential relationship pixel intensity. I and depth A1, A2 and A3 New the unknown coemcients. 3.2 th Results pixel intensities on bands 4(0.5-0.6 mm) and 5(0.6-0.7um)were obtained at points of known depth. Plots of pixel intensity versus depth give an exponential relationship on both bands. A least-squares minimization was performed formed using eq. (8)to calculate the best values of the coemcients A1.A2 and A3 by using 100 points The depths of another 123 points that were not used in the least-squares minimisation were then calculated by using the values of the calculated depths are 4.4 m respectively for band 4. And 5.0m and 3.9 m for band 5using depths up to about 38 m. penetration was calculated by determining the depth zm such that the intensity I exceeded the deep-water signal A1by one grey level that is.
    Zm = (InA2) A3 -------------------(9)

    The calculated penetration depth was about 37 mfor band 4 and about 35mfor band 5. The original band 4 image was resample using the nearest-neighbor method ftp, the least-squares fit the intensity resample for various depth ranges were obtained to enable density slicing to be performed and be perfoprmed and a bathymetric map was derived from the resampled image (see Figute 3.) A bathymetric map was also derived from the 1975 and 1977 Admiralty hydrographic charts of the area in order to compare with the bathymetric map derives from the satellite data (see Figure 2.)
Discussion and conclusions
The plots of pixel intensity versus dept give the expected exponential relationship in the bands that were studied. High values for he calculated penetration depth were obtained. The RMSDEV values of the calculated depths are about 10% of the water depth. The depth accuracy requirements are 30 cm for depth up to 30m. 1 m for depth from 30m to 100m and 1% of the depth for deeper than 100m according to the accurecy standards recommended for phydrographic surveys by the International Hydrographic Organization. The results obtained in this study indicate that these accuracy requirements are difficult to achieve by remote sensing techniques However, the bathymetric map delved from the density sliced image of the band 4 Landsat MSS data show many similarities with the corresponding bathymetric map derived from the admiralty hydrographic charts. This shows that in areas where the water clarity is good. Satellite data can be used to obtain some general idea on the depth contours. A significant amount of bathymetric information can be observed on these images. Satellites cannot entirely replace conventional ship borne surveys but they do provide a source of broad scale, synoptic information of medium quality at a very low cost. For example satellites can provide an extremely effective means of carrying out preliminary surveys over wide areas. Especially in Remote Regions. Ships need then be used only in those areas where closer investigation is indicated and in this way the sending of ships on unproductive surveys may frequently be avoided. In the preparation of bathymetric charts satellite data may be used to fill in contours between lines of ship soundings and may reduce the number of soundings required and hence the cost. Furthermore, because of the frequent over flights saltiest provide an effective means of monitoring changes to the coast and seabed.

References
  1. Brown, W. L. Polcyn, F.C. and Stewart, S.R. 1971 A method for calculating water depth, atienuation coefficients, and bottom reflectance characteristics. Proceeding of the 7th International Symposlum on Remote Sensing of Environment ERIM . ann arbor, Michigan, 663,682

  2. Kapoor, D.C., 1976 International cooperation in hydrography. International Hydrographic Review, 53, 7-15.

  3. Polcyn, F.C. 1976 final report on NASA /cousteau ocean bathymetry Expertment : Remote bathymetry using high-gain landsat data NASA -CR-ERIM Report no 118500-1-F

  4. Polcyn,f,c. and byzenga D.R. 1975 Remote bathymetry and shoal detection with ERTS :ERIS water depth. ERIM report No. 193300-51-F.

  5. Sorencen. B.M., 1980 Important conslderations regarding remote sensing of lmarine water quality. In Coastal and Marine. Applications of Remote\Sensing.edtted by A.P; Cracknell, Remote Sensing Society, 77,83.

  6. Watson, FF 1986 Ocean charting. International Iiydrographic Review, G3,119 151

Figure 2: Ballymetric map derived from Advirally hydrogaphic clearts (depths are relative to lovest astronomical ude)