In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
i 
467 
(using Chavez’s DOS method) and transformed in reflectance val 
ues. The images were also interpolated to a 5 m resolution us 
ing the minimum curvature algorithm. It is visually striking to 
see that, apart from a few exceptions, the multi-temporal dataset 
shows an almost constant shrinking of the lake surfaces and even 
the disappearance of one small water body. Figure 5 illustrates 
the difference in lake surfaces for the whole period. The triangu 
lar area at the bottom of the 1984 image, was part of the eucalyp 
tus plantation and is now regenerating. 
Figure 5: Comparison of the lakes for the same time of the year 
in 1984 (left) and 2009 (right). 
Two lake extraction methods were used: 1) threshold of the pos 
terior probability of the maximum likelihood classification and 
2) threshold of the modified normalized difference water index 
(MNDWI). The former approach yielded far superior results in al 
most all images. We attribute this result to the presence of aquatic 
vegetation and turbidity (for the Formosa lake) mixed with the 
water which tends to increase reflectance in the near and mid- in 
frared. Figure 6 shows the extreme example of the Formosa lake 
which is outside the State Park and suffers from eutrophication 
and aquatic vegetation bloom. Figure 6 should also be compared 
with the false color image at the top right of Figure 5. 
(a) Classification -5m (b) NDWI threshold -5m 
Figure 6: Comparison of the contour extraction methods using 
classification (a) and MNDWI threshold (b). 
By using the posterior probability of a single water class, we 
found that there was always an easily identifiable break between 
the water and non-water classes that made the selection of a thresh 
old very easy. The threshold was applied to all 50 images and 
the area of all six lakes computed for every date. The graph in 
Figure 7 shows how these areas have changes between 1984 and 
2009. Table 2 gives an over view of the shrinking of the six lakes. 
The lake areas of 1990 are also indicated for being the record size 
for all lakes. While lake “Pista” has completely disappeared since 
2000, four other lakes have lost between 59 and 80% of their area. 
The lake “Azul” has somewhat retained much more of its origi 
nal area (loss of 29%) and it is also the only lake surrounded by 
hydromorphic gley soil with a higher clay content. 
Since we did not have reliable elevation data at the time of writ 
ing, the areas water surfaces could not be associated with precise 
Areas 
km 2 
Pista 
Très 
Lakes 
Meio 
Sede 
Azul 
Formosa 
1990 
4962 
28778 
37413 
56402 
105389 
296237 
1984 
375 
14795 
32471 
39030 
92670 
291502 
2009 
0 
2928 
7228 
12243 
65829 
170409 
% loss 
100% 
80,2% 
77,7% 
68,6% 
29,0% 
58,5% 
Table 2: Comparison of the areas of all six lakes between 1984 
and 2009 with the shrinking expressed in percentage (1990 was 
the record year for all lakes). 
Figure 7: Graph showing the evolution of the area of all six lakes 
for the period 1984-2009. 
altimétrie level measurements. These data will be available at the 
third quarter of 2010. Using the digital elevation model (DEM) 
from the ASTER sensor, and overlaying the contours over it we 
were able to estimate the lowering of the water level for the 1984- 
2009 period to about 1 meter for the lake “Azul” and to slightly 
over 2 meters for the lakes “Sede”, “Meio”, “Très” and “For 
mosa”, being outside the State Park. Figure 8 shows the 1984 
and 2009 levels on the ASTER DEM profile. 
Figure 8: Water level differences between 1984 and 2009 on an 
ASTER DEM profile for lake “Formosa”. 
The validation of the data was done using the approach described 
in section 2.4. Table 3 shows the validation obtained with both 
control datasets (GPS and Ikonos image) and with the two meth 
ods of comparison (simple comparison of areas and “intersection 
-r- union” approach). As expected, the accuracies with the latter 
method are slightly lower but since all accuracies but two are well 
above 80%, we conclude that both our extraction method and our 
geometric correction are within very acceptable boundaries. Fig 
ure 9 shows the contours extracted from the Landsat image of 
2010 and the GPS survey coutours for three of the lakes. 
3.3 Statistical Testing 
Spearman’s correlation test was applied to the area series of all 
lakes along with the AW data for the same period. The results