Full text: Remote sensing for resources development and environmental management (Volume 2)

632 
2.3. Climatology 
Rainfall in the region is concentrated in the autumn, 
winter and spring seasons. Both temperatures and 
evaporation increase markedly in May and remain 
high until late September. Consequently runoff 
events are restricted to the wet season and streams 
are dry throughout the summer as there is no base 
flow contribution and soil moisture deficits are 
high. The climatological pattern is slightly 
different in the south of the study area, where 
rainfall totals are lower and consequently runoff 
events are rarer, and in the north where it is 
slightly wetter. 
3. SATELLITE DATA AND DESERT GEOMORPHOLOGY 
3.1. Previous work 
Satellite data have been used for geomorphological 
investigations in arid and semi-arid areas by 
several workers and most applications have involved 
landform mapping (eg. Mitchell et al., 1982; 
Sunha & Venkatachalam, 1982 and Van Steen, 1982) and 
surficial material survey (eg. Asem et al., 1982; 
Bird et al., 1982; Davis et al., 1982; Gladwell, 
1982; Hamza et al., 1982; McCord et al., 1982; Sunha 
& Venkatachalam, 1982 and Townshend & Hancock, 1981). 
Few workers have attempted to monitor geomorpho 
logical change although Graetz & Pech (1982) and 
Klemas & Abdel-Kader (1982) have measured river 
channel changes and flooding in arid and semi-arid 
environments. 
Three problems are apparent in these previous 
studies: 
1. They have been limited by the relatively coarse 
spatial resolution of MSS sensors. 
2. They have been limited by the restricted 
spectral resolution of MSS data. The inclusion of 
middle IR bands (1550-1750 and 2080-2850nm) on the 
TM has greatly enhanced the possibility of dis 
criminating between surficial materials, (Bodechetel, 
1983; Gladwell, 1982; Hunt, 1980; Kahle, 1984). 
This enhanced power of surficial material 
discrimination is crucial to ary interpretation of 
sediment dynamics. 
3. The few geomorphological monitoring studies 
that have been undertaken using Landsat data have 
been severely restricted by image availability. 
Archival material has been compared with current 
imagery (Klemas & Abdel-Kadar, 1982) but change 
detection utilising concurrent image interpretation 
and ground verification has been far less 
satisfactory (Graetz & Pech, 1982) because of the 
costs involved in data availability, acquisition, 
ground station receiving policies and atmospheric 
conditions. 
3.2. Change detection 
Jones (1986a, b) has thoroughly evaluated the 
potential of digitally processed TM imagery for 
geomorphological mapping in this area of Tunisia. 
Whilst this research shows the applications that 
can be made using single date imagery, monitoring 
geomorphological change using digital imagery 
requires the use of a multidate imagery and different 
change-detection algorithms. 
In this project image data was supplied as CCT's 
and analysed digitally using a I2S Model 75 image 
processor. Change detection proceedures involve 
either a multidate or a post-classification 
comparison approach. A multidate approach combines 
the two unprocessed images to produce one output 
data set. A post-classification comparison approach 
involves an initial supervised or unsupervised 
classification of the two images. In this study the 
multitemporal approach to change detection was 
preferred because quantitative comparisons between 
the two techniques have shown it to be more accurate 
(Singh, 1984). 
Any change detection study involves scene-to-scene 
image registration to ensure that pixels correspond 
to the same ground locations in each image. Six 
ground control points were used to co-register the 
two TM quadrant images used in this study with an 
average RMS erros of +0.37 pixels, compared to twenty 
seven ground control points used to co-register the 
anniversary 512 x 512 pixel MSS images. This confirms 
the expected improved geometric fidelity of the TM 
compared to the MSS. The ground control points 
chosen in the co-registration procedure were 
permanent features, in the landscape such as road 
junctions and road/rail crossings. 
Any misregistration of the imagery will produce 
errors in the change detection output images since 
boundary pixels corresponding to one surface cover 
type may be compared with boundary pixels of the 
adjacent cover type, resulting in spurious changes 
being detected later. In order to remove these 
possible edge effects a median filter, with a 3 x 3 
pixel square kernel, was passed over all the images 
used in this study before the change detection 
algorithms were applied to the data. 
Image differencing, image ratioing and principle 
components analysis were found to be the most 
meaningful change detection algorithms. Vegetation 
indices (Howarth and Boasson, 1983; Singh, 1984) and 
the ratio differences technique were unsuccessful 
in detecting change. Possibly vegetation changes in 
semi-arid environments are too subtle to be detected 
on the imagery despite the fact that in some 
environments, for instance on playa margins, there 
are geomorphologically significant vegetation changes. 
The most useful spectral bands were MSS Band 7 
(800-1100nm) and TM Bands 3 (630-690nm) and 7 (2080- 
2250nm). 
Differenced images were produced by subtracting the 
median filtered image for the first date from that 
for the second date, and adding a constant to ensure 
that the output values were positive. Ratio images 
were produced by dividing the image for the first 
date by that for the second date. For the principal 
component analysis, the two images being compared 
were treated as one date set. In the analysis, 
Bands 4 (500-600nm) 5 (600-700nm) and 7 (800-1100nm) 
were used for each MSS image, and Bands 4 (760-900nm), 
5 (1550-1750nm) and 7 (2080-2250nm) were used for 
each TM image. Previous research into the use of 
principal component analysis in change detection 
indicates that gross differences due to overall 
radiation and atmospheric changes are contained in 
principal component 1, and that statistically minor 
changes associated with local changes in land cover 
appear in the minor component images (Byrne et al., 
1980; Lodwick et al., 1979; Richardson and Milne, 
1983). Consequently principal components 2 and 3 
were used to detect geomorphological change in this 
study. 
Thresholds, chosen on the basis of previous 
research (NelSon, 1983; Singh, 1984), were applied 
to all change detection output images at +lcr from 
the mean value. FCC images were produced by 
assigning the changes which corresponded to pixels 
with values of <-la to the red gun of a colour 
monitor, those >+la to the blue gun, and either the 
MSS Band 7 or TM Band 7 image to the green gun in 
order to preserve spatial detail. 
4. GEOMORPHOLOGICAL MONITORING AND CHANGE DETECTION 
IN PLAYA ENVIRONMENTS 
4.1. Environmental setting of study playas. 
Three playas have been intensively studied using 
satellite imagery and ground observations in this 
project. 
The Chott 
south of t1 
and they fc 
from scuthe 
Djerid is t 
has an elon 
Fedjadj, whi 
Hamma (770k 
have been t 
(Coque, 196 
Chott el 
enclosed ba 
bounded to 
south by th 
ranges have 
Low ground 
to the east 
4.2. Geomor 
Geomorpholo 
playas from 
and they ca 
the hydrolo 
interaction 
activity, 
seasonal pa 
geomorpholo 
phenomena r 
shorter or 
patterns an 
fluctuation 
of the geom 
they can th' 
This subdiv 
it recognisi 
observed wi 
different i; 
distinct fr> 
techniques. 
Seasonal < 
detection a 
from the di: 
wet and dry 
three playas 
geomorpholo< 
geomorpholoi 
1. - surf; 
moisture) 
2. - surf; 
3. - vegel 
4. - aeol: 
Seasonal j 
vegetation i 
detection ir 
TM imagery c 
greatest chc 
at the end c 
the adjacenl 
well vegetal 
gyonianum ai 
towards the 
Subdued spai 
act as surfc 
into rills, 
depth and 5C 
The individi 
distal end c 
(Fig. 1). 7 
out the yeai 
vegetation t 
The change 
higher middl 
summer. Thi 
higher level 
channels anc 
by higher re 
The high sun 
higher soil
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.