In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B
465
geometrical correction was done with an “image-to-image” ap
proach using a one-meter Ikonos image as basis (which was geo
metrically adjusted using control points from a geodetic GPS sur
vey). The atmospheric and radiometric correction were applied
using an in-house program build for that purpose: Corat-Lands at.
The program takes as input a table containing 1) the name of
the image file, 2) the DN value for the dark object substraction
(Chavez Jr., 1988) for bands 1, 2, 3, 4, 5 and 7, 3) the sun ele
vation angle and 4) the sun-earth distance in astronomical units.
The output is a 16 bit reflectance image (reflectance values were
redistributed between 0 and 10 000).
Meteorological data - Water balance. The calculation of the
AW was first proposed by Thomthwaite in 1948 and improved in
1955 (Thomthwaite and Mather, 1955). The main objective of
the methodology is to determine the hydraulic characteristics of a
given region without direct measurements on the ground (Pereira,
2005). The water balance is the simple budget between input and
output of water within a watershed:
AS = I P + Gi n - Q + ET + G ou t I (1)
Inflow ) \ Outflow
where P is the precipitation, Gi n and G ou t represents the ground
water flow, Q is the runoff water and ET is the évapotranspira
tion.
The Thomthwaite procedure simplifies the AW calculation by es
timating all its components from only two input parameters: av
erage daily temperature and precipitation:
/ IP \
AW t = AW t - 1 ex P {-- w ±)
(2)
where AW t is the available water at time t, AW t -1 is the avail
able water at time t — 1 (in our case we set set t to be every
ten days), PETt is the potential évapotranspiration at time t and
AWC is the soil’s water holding capacity. The water balance can
be summarized in three situations.
• AP < 0; net precipitation (precipitation - potential évapo
transpiration) is less than zero: the soil is drying.
• AP > 0 but AP 4- AW t -1 < AWC; net precipitation is
more than zero but net precipitation plus the available water
from time i — 1 is less or equal than the soil’s water holding
capacity: soil is wetting.
• AP > 0 but AP + AW t ~l > AWC; net precipitation
is more than zero and net precipitation plus the available
water from time i — 1 is more than the soil’s water holding
capacity: soil is wetting above capacity and water goes to
runoff.
2.3 Extraction of lake contours
Water in liquid form is usually well contrasted from its surround
ing dry(er) land unless it is overshadowed by vegetation cover
like mangroves, flooded forests or aquatic plants (Caloz and Puech,
1996) . In many cases, a simple threshold in an infrared image
histogram can reliably separate water from the other land covers
with a relatively good rate of success and investigators have de
veloped simple techniques for doing so in a systematical manner
(Bryant and Rainey, 2002; Jain et al., 2005). Histograms of near
infrared images containing a fair amount of open water surfaces
are usually bimodal with the first peak directly related to water.
Yet, when one looks closer, the water-land limit is often blurred
by a varying width occupied by aquatic plants that can fluctuate
over various time scales (yearly or seasonally). Using a sequence
of historical Landsat images for which we had no validation data,
we needed to have a very strict definition of the water-land inter
face. We defined the lake ’’water-margin” as the point at which
water overwhelmingly dominates the surface and estimated that
point to correspond to 70-80%.
2.3.1 Resampling through interpolation: Another source of
error comes from the mere sampling resolution of 30 meters used
by the TM and ETM sensors. Although not considered an issue
when measuring an ocean of a large lake, it rapidly becomes a
problem when studying very small lakes such as the ones found
in the VPSP that range from just over ten hectares to just under
one hectare. In these small lakes, the number of mixed pixels can
represent a large proportion of the total lake pixels (up to about
35% in the case of the smallest lake). A half pixel shift in image
registration could signify an important difference in water pixel
count.
Scale (or spatial resolution) can have various effects on image
classification accuracy. A finer resolution can usually decrease
the proportion of pixels falling on the border of objects (hence
less mixed pixels) which can result in less classification confu
sion. Conversely, a finer resolution will generally increase the
spectral variation of objects that can, in turn increase classifica
tion confusion (Markham and Townshend, 1981). Fortunately,
water (especially clear and deep) is spectrally a relatively smooth
surface for which a finer resolution will bring more benefit (less
border pixels) than disadvantage (spectral variation). Based on
the fact that water is spectrally smooth and that it strongly con
trasts with dry land, we argue that artificially increasing the reso
lution of an image containing water surfaces can generate a bet
ter definition of the water-land limit. To do so, a number of tests
were prepared to define an appropriate interpolation method to
resample the images.
Amongst the various interpolation methods we opted for the min
imum curvature interpolation (a variation of bi-cubic spline) with
tension as described in Smith and Wessel (1990). This inter
polation method has the advantage of being able to generate a
smooth surface without generating undesirable fluctuations (arti
fact peaks or dips) by using a tension parameter. This interpo
lation proved better than ’’inverse distance weighted” that tends
to produce artifact dips between sampling points (Maune et al.,
2001). The minimum curvature worked well and fast and gener
ated smooth ramps while keeping a sharp water-land edge. Fig
ure 2 illustrates the effect of interpolating the Landsat data to 5 m
on the lake extraction processing.
(a) 30 m
(b) 5 m
Figure 2: Comparison of the lake extraction methods using the
original 30 m Landsat data (a) and the 5 m interpolated data (b).
2.3.2 Calssification: Because of the nature of our study, un
supervised and automatic segmentation methods were discarded.
These approach are most suited with single date image applica
tions when terrain validation is possible. Two supervised methods