the classifi- 
rough pixel 
ification does 
ith a unicon- 
t possible to 
and environ- 
areas depends 
.a and Raitala 
984a), aquatic 
1985), water 
1980, Arkimaa 
1984b). The 
substantially 
:ffective vari- 
3 of satellite 
ones can be 
Computer- aided 
provide an im- 
vely different 
t quantitative 
'eloped further 
d in routine 
In the most detailed studies it was found possible 
to obtain at least five distinct class categories 
from within quite simple water areas by means of 
supervised Landsat classification procedure. These 
categories consisted of 11 to 17 classes, some of 
which slightly overlapped in groups to form the 
categories. The five categories established may be 
quite reasonable, while the number of individual 
classes and their unambiguous validity somehow also 
depend on the complexity of the aquatic variables 
present and on fortunate circumstances in finding 
ideal representative reference fields for each of 
the classes. Usually there are several variables 
forming a surface complex and small number of re 
ference fields must be chosen to cover all combina 
tions of the surface variables. This may result in 
a small number (two to four) of related classes 
forming an association of intermingled classes or a 
category which includes a certain range of minor 
variations on the ground. 
Landsat multispectral classification system over 
aquatic areas may be most advantageous when used in 
conjunction with ground-truth data. Scientists 
familiar with basic variables within particular 
water areas can use the method to yield rapid up-to- 
date information of the state of and changes in the 
aquatic environment in a valid and cost-effective 
manner. 
1 waters of the 
open deep sea, 
submersed veg" 
ects, G *= di 
scharge from a 
Figure 3. Detailed study of the shallow water areas 
of the Bay of Liminka (cf. lower part of Fig. 2) in 
dicates different vegetation complexes as follows 
(Raitala et al. 1984c): II = Eleocharis acicularis, 
Eleocharis palustris, III = Eleocharis palustris, 
IV = Scirpus, Eleocharis acicularis, V = Eleocharis 
palustris, Potamogetón friesii, VI = Eleocharis pa- 
lustris, Eleocharis acicularis on hard and clayey 
bottom, VII = Eleocharis palustris, Eleocharis aci- 
cularis on soft and stony bottom, VIII = ZannicheIlia, 
Limoselia, IX = Eleocharis acicularis, Chara. 
4. MULTITEMPORAL MONITORING 
Because almost all of the variables affecting the 
reflected spectrum from within water areas are also 
time-related the most appropriate application of 
aquatic remote sensing is multitemporal monitoring 
of environmental changes. In Finland the recent 
littoral processes are most effective in two diffe 
rent places: around reservoirs with annually 
fluctuating water levels and along the seashore of 
the northern Gulf of Bothnia, with a maximum land 
upheaval of 1 cm a year. 
The Porttipahta water reservoir displays an extre 
mely intense vertical water level amplitude of 11 m, 
from very low in early spring to maximum flooding in 
autumn. The erosion of former terrestrial forest 
and bog vegetation and the underlying soil has been 
intense. Because only two images presenting lower 
and higher medium water levels were used two diffe 
rent presentations were prepared to display areas 
with different erosion effect between those two 
water levels and between the high medium-water level 
and the highest water level (Jantunen and Raitala 
1984). The wind-induced fluctuations in the water 
level among the seashore off the city of Oulu are 
much smaller. Similar multitemporal monitoring evi 
dently displays largely seasonal variations in 
vegetation within the shallow hydrolittoral areas, 
differences in water quality, but also some differe 
nces in depth relations. 
Although only very simple arithmetic procedures 
were involved in comparing and manipulating the 
corresponding pixels, the results were encouraging 
and important enough to be evaluated in respect to 
further time-related monitoring of these and other 
related areas. The study areas were especially good 
for testing and evaluating the potential of multi- 
temporal satellite remote sensing because environ 
mental changes were and will continue to be so 
evident. Multitemporal satellite remote sensing 
will clearly offer a useful tool in monitoring 
environmental changes within other aquatic areas, 
too, although the exact outcome of this experimenta 
tion will nevertheless depend on future development 
of the Landsat and other satellite programs them 
selves (Beardsley 1986). 
Figure 4. Multitemporal monitoring ot the sea area 
off Oulu. The darker the symbols are the more dis 
tinct have seasonal and water level changes been.