imposite of the 
.ng the playa 
>un and spring 
ì plunging 
swing spring 
ant limestone 
a Chott el 
bottom of the 
{ can be 
r halophytic 
arrounded by 
because of the 
jetation and 
3 (0.63- 
also relatively 
sund vegetation 
ration. As 
pplies all the 
ly under stress 
d Bands 3 and 4 
station does 
partially dies 
tic vegetation 
In addition the 
Table 1.Profiles of digital values across spring mounds, aioun and associated sand tails. 
SAND TAILS 
177 
174 
169 
180 
171 
180 
176 
180 
178 
179 
179 
178 
202 
176 
224 
189 
222 
205 
221 
211 
212 
207 
204 
193 
197 
182 
181 
180 
171 
178 
161 
174 
159 
176 
164 
174 
SPRING MOUNDS 
203 
201 
206 
204 
203 
210 
0 
218 
229 
196 
0 
98 
140 
124 
111 
118 
0 
118 
133 
120 
59 
0 
48 
94 
110 
0 
131 
110 
221 
196 
181 
0 
48 
43 
203 
222 
AIOUN 
90 
93 
91 
77 
89 
78 
75 
70 
71 
70 
70 
70 
68 
70 
73 
68 
83 
94 
100 
101 
103 
98 
101 
loo 
82 
86 
72 
71 
72 
75 
77 
74 
73 
71 
68 
83 
halophyte communities have lower plant densities 
than the mound spring communities and consist of 
single storey shrubby or herbaceous plants. As a 
consequence of the differences in community structure 
and density single pixel spectral responses in the 
halophyte communities are a mixture of salt-tolerant 
vegetation and bare playa surfaces. Furthermore, 
the dominant types of vegetation are often grey-green 
or reddish green, compared to the deeper greens of 
date palms and other irrigated crops. 
Aioun only occur in the central playa facies in 
the Chott el Djerid where there is no vegetation. 
However there are concentric circles of salt 
effloresences, indicative of variations in surface 
salt concentrations. These create a very distinctive 
circular pattern of variations in reflectance at all 
visible and infra-red wavelengths and so can be 
detected in all TM bands. 
4. DIGITAL IMAGE PROCESSING AND PLAYA GEOMORPHOLOGY 
Digitally processed MSS and TM imagery has been used 
to map playa facies of the Chott el Djerid (Jones, 
1986a,b; Mitchell, 1982; Munday, 1985) and the Chott 
el Fedjadj (Jones, 1986a,b). From Fig. 2 it can be 
seen that the distribution of spring mounds is 
confined to marginal playa facies and that aioun are 
restricted to a more central facies, (Mitchell, 1982) . 
This study builds upon previous work by examining 
in more detail the distribution of spring mounds and 
aioun within playa facies. All of the image 
processing reported on in this paper was carried out 
on a TM image (Path 192; Row 36) taken on 29 January, 
1983. 
The detailed image processing was carried out on a 
subscene of this larger image. This was located so 
that it encompasses the spring mound field on the 
Chott el Fedjadj. It was apparent from a visual 
inspection of different single band images of the 
area that the data showed significant intercorrela- 
tion between bands, although some bands, particularly 
5 (1.55-1.75ym) and 7 (2.08-2.35ym) contained more 
geomorphological information than others. Table 2 
shows the correlation matrix derived from the six TM 
reflective bands in the spring mound test area. 
High correlations were found between all bands (for 
all correlations r= +0.688) with very high 
correlations between the three visible bands, the 
two middle infra-red bands and Band 4 and all other 
bands. It is well known that the TM was designed 
primarily for vegetation discrimination with bands 
selected to take advantage of the spectral response 
of vegetation (Salmonson et al., 1980). The 
implication for geomorphological investigations of 
bare surfaces, such as playas, is that after studying 
an infra-red or FCC image that additional single band 
images provide little additional information unless a 
narrow pixel value range is utilised. 
After comparing the images it was concluded that 
Band 3 (0.63-00.69yjn) was the most useful image for 
analysis. 
This was due to the clear depiction of vegetation 
(by chlorophyll absorption) on spring mounds and 
playa surfaces, which allowed the easy identification 
of spring mounds. The playa facies with spring mounds 
Table 2 TM Bands correlation matrix for spring mound 
area 
1 
2 
3 
4 
5 
6 
1.000 
.957 
1.000 
.925 
.983 
1.000 
.869 
.932 
.936 
1.000 
.734 
.840 
.878 
.872 
1.000 
.608 
.706 
.738 
.746 
.944 
1.000 
was delimited and a contrast stretch was applied to 
this area to enhance the contrast between features 
(Fig. 2). All further image processing was carried 
out on this contrast stretched image. 
One of the main aims of the study was to see if 
linear patterns could be detected in the distribution 
of spring mounds that could be related to faulting 
and jointing patterns in the underlying rocks. The 
alignment of isolated circular geological features to 
indicate underlying geological phenomena has been 
attempted in volcanic terrain but we believe this is 
the first attempt to use such analytical techniques 
in folded sedimentary strata; although lineament 
analysis of faults, lithological boundaries and fold 
axes is well known in similar terrain. 
To highlight the edges in the image prior to 
lineament analysis a series of directional edge 
enhancements were applied to the single band image. 
Edge enhancement operates by passing a digital 
filter or kernal, in the form of a matrix, over the 
data. It has been successfully used in many studies 
of geological lineaments (Bailey et al., 1982). Four 
directional filters were selected (N, NW, W and SW) 
for this study. 
Ratioing of spectral bands was also examined 
initially as it is known to reduce topographic noise 
and enhance spectral differences between surface 
features. However, to effectively use such methods, 
the relationship between the surface materials and 
the spectral responses must be understood. Whilst a 
partial knowledge of the reflectance properties of 
playa salts it known the actual ratio images produced 
poor results and were not used in the later analysis. 
4.1 Lineament analysis of spring mound distribution 
Lineament analysis of spring mound distribution were 
carried out on five images; a Band 3,4,5 FCC, four 
single Band 3 images with different directional 
filters (N, NW, W and SW), (Fig. 4). In addition 
lineament analysis of the known lithological 
boundaries in the area was carried out on a Band 3 
image. 
Software for the lineament analysis was written 
by D. Greenbaum (British Geological Survey, Keyworth, 
Nottingham, UK) for the IIS image processing system. 
The technique involves the visual interpretation and 
drawing of lineaments on the VDU. Lineaments were 
defined as straight lines joining three or more 
spring mounds. The spring mound area is roughly 
rectangular (Figs. 2 and 3) and this means that there 
is a greater probability of longer lineaments, 
connecting greater numbers of mound springs, in 
609