hymetry in
| models of
‘y mapping
| resolution
10menon at
arting from
veloped for
>loped, was
nt test sites
ed to study
ints within
nalysis lies
1 developed
'HM
n suffer an
"sorption of
ic light, the
size of the
(extinction
in literature
from RS
one (DOP)
(1)
layers
related to
| for small
iter column
e there is a
ranging Eq.
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004
(1) lead to the classical DOP equation for the water depth
determination
N In(L, ) N In(L, ), |
pe ied al 2)
where N= number of spectral bands
In practice, to guarantee homogeneity the DOP model assumes
constant coefficients of absorption, and, as will be discussed
later, this is the main cause of the failure of the DOP algorithm
in the lagoon of Venice, where the spatial lack of homogeneity
is very high.
22 A new stratified genetic algorithm for batimetric
measures in the lagoon
The high spatial variability of the seabed and the presence of
suspended sediments in the water is the principal causes that
make inapplicable the DOP model in the lagoon of Venice. The
high variability of which the coefficients of absorption assume
in the lagoon of Venice can be interpreted as a lack of
homogeneity in the water column, circumstance that does come
less one of the conditions imposed to the Jupp bathymetry
mapping model. Since the condition of uniformity of the
seabed in the lagoon of Venice is not verified, the Jupp model
has been modified in order to compute an accurate description
of the real bathymetry into a complex environment such as the
lagoon of Venice. Looking at Eq. (2), the term
N In(L )
S 6
Lu —2k;
iz]
is interpretable as a corrective term that normalizes the
geometric media of Ly. It can be noticed that Eq. (2) doesn't
perform a complete regression on the radiometric input data,
but uses only the deriving information from the coefficient of
absorption (the slope) to achieve a first bathymetric estimation,
and, subsequently, compensates the error by esteeming
separately the term (3) in shallow water areas. In our new
model, called stratified genetic algorithm (SGA), the term (3)
has been eliminated from Eq. (2) and a new parameter (Y;) has
been introduced to perform a complete regression on the
multispectral dataset.
zy (4)
where m = number of layers
The new algorithm proposed uses Eq. (4) to build a statified
genetic algorithm. Exploiting the different penetration of the
electromagnetic radiation in the water, Eq. (4) is computed for
intervals of increasing depths. The SGA algorithm divides the
water column into elementary volumes of increasing depth, and
for each of them computes the coefficients of absorption (kj),
the parameters (Y;) and the estimated bathymetry (z). This
95
procedure is repeated for every spectral bands available, until
all the image is processed. In practise, for every elementary
water volume built, the SGA computes a batimetric estimation
for every spectral band available and puts in competition
results. Only those outputs (coefficients of absorption,
coefficients of regression and the estimated bathymetry) with
the highest correlation coefficient are assumed as representative
for the whole elementary water volume. The final bathymetry
map is built incrementally, using for every step the best
correlated parameters found.
3. BATHYMETRY MAPPING IN THE LAGOON OF
VENICE
3.4 Study area
The lagoon of Venice is a particularly complex territory, where
land and water intersect to form systems and subsystems
regulated by delicate equilibriums and where ecosystems of
incomparable beauty coexist with a considerable industrial and
agricultural activity. The lagoon of Venice is extended on a
surface of about 550 km?, among the terminal course of the
Brenta river (mouth of Brondolo) and the final flow of the Sile
river to north (mouth of old Piave). From the hydrographical
point of view, the lagoon is divided into three portions: the
mouth of Chioggia, the mouth of Malamocco and the mouth of
Lido. These subtend and determine real lagoon basins, that are
clearly identifiable and have easily recognizable limits. The
78% of the lagoon surface is characterized by wide expanses of
water crossed by a dense net of canals with different depth.
The communication of the lagoon with the sea determines its
brackish character, that guarantees the survival of the peculiar
biological characteristics, and the daily sea ingression-
regression across the mouths constantly models the physical
conformation of the lagoon.
3.2 Dataset
The SGA algorithm proposed has been used to evaluate the
bathymetry in the lagoon of Venice, using the multispectral data
acquired with the QuickBird sensor on May 16, 2002 (10:05:40
GMT). Figure 1 show the study area (from 45°33°28” N and
1221710" E to from 45?23'53" N and 12?30'33" E) The
spectral and spatial characteristics of the QuickBird images are:
Pancromatic band with a GSD of 0.64:
= (.450-0.900 um spectral range.
Multispectral bands with a GSD of 2.56 m:
= Blue (0.450-0.520 um);
= Green (0.520-0.600 um);
= Red (0.630-0.690 pm);
= Near infrared (760 -900 pm).
QuickBird data have been converted to at-sensor radiance and
atmospherically corrected using the 6S (Second Simulation of
the Satellite Signal in the Solar Spectrum) radiative transfer
code (Vermote ef al., 1997) and the optical thickness obtained
from the Aeronet network (http://www.aeronet.gsfc.nasa.gov).
Satellite data was subsequently geometrically corrected and
georeferenced.
A sounding bathymetry chart of the entire lagoon at scale
1:5,000 was used as batimetric reference data. Depth sounding
points were extracted from the chart and they have been
interpolated and resampled with a TIN (Triangular Irregular
Network) algorithm, in order to obtain a regular sampled raster
