desertification developed as stochastic process that favoured to mathematical modelling of the ecosystem 
dynamics using Markovian approach. 
Study area occupies the lower part of the Amudarya delta between emerged coast of the Aral sea on the 
North and residual highlands Kyzyldzhar and Kushkantau on the South, between the high plateau Ustyurt on 
the West and the desert sands Kyzylkum on the East. This area presents a flat delta plain deposed by 
alternation of lake loams and alluvial sands. The flat delta plain is transected by many river channels, river 
levees, active and dead arms, old river beds where prevailing light alluvial deposits. Flood plains with silt 
depressions, river lakes, backswamps lie between river facies and underlied by heavy alluvial and lacustrine 
deposits. Ground waters are very spotted from 0 to 10 m of depth, from 0.5 to 5 g/1 of mineralization. Soil 
and vegetation are heterogeneous also: from deserts to swamps, from meadows to solontchaks. On space 
photographs enlarged to 1:200 000 we reliable recognize 10 ecosystem classes area of which could be measured 
with relative error less then 2% (tab 1.1). These 10 ecosystem classes were shown on the ecological maps at 
scale 1:200 000. 
Methods of the Dynamics Studies 
The dynamics studies requires optimization and verification of main remote sensing features of study area such 
as probability of correct recognition of ecosystems on aerial and space photographies, representative size of 
study area, and reliable time interval between successive surveys. 
The first requirement is a optimal size of study area for revealing of the ecosystem dynamics as a stochastic 
process. Too small study area couldn't reveal all interactions between ecosystems and couldn't ensure 
Markovian approach. Previous experiences proved that study area less then 100 sq.km is nonsufficient for 
mathematical modelling of the ecosysytem dynamics (for example, 76 sq.km [Debusche et al., 1977], 6 sq.km 
[Vinogradov, Popov, 1982]). Too large area of study area could be undesirable also. For example, study area 
of the Amudarya delta more then 1000 sq.km was found excessive as included too heterogeneous landscapes 
which hadn't transition traces at the studied time interval. This experience proposed that study area more then 
1000 sq.km is surplus for mathematical modelling of the ecosystem dynamics during recent time. Thus, size 
of optimal study area for mathematical modelling lies between 100 and 1000 sq.km. 
Number of the ecosystem classes suitables for mathematical modelling is restricted by probability of correct 
identification depending on scale of mapping and quality of aerospace images. Because of this, relative area 
changes during one time interval between successive surveys should be rather more then relative error of photo 
interpretation of each ecosystem classe. French authors [Debusche et al., 1975] found the minimum threshold 
of 3% in the change coefficient, Russian authors [Vinogradov et al., 1980(1979), 1988] believed that the 
probability of correct recognition should be not below 96-98% also. 
In terms of threshold error 2-4% the sufficient time interval between successive aerospace surveys depend on 
rate of the ecosystem dynamics otherwise the error of photo interpretation can exceed the ecosystem area 
changes. Stable and slight dynamic ecosystems, that is, with area changes of less then 0.5% area per anuum, 
require revision not less than once in 8-10 years. Moderate dynamic ecosystems with area changes near 1% 
area per anuum require revision every 5-6 years; middle dynamic ecosystems (2-3% area change) every 2-4 
years; and high dynamic ecosystems (4% area change 3 and more) require inspection every year [Vinogradov, 
1984]. 
There are several developed technologies that can be used for comparison successive aerial or space 
photographs for detection of the ecosystem changes. This comparison is executed by comprehensive 
combinations of visual and machine processing. Sometimes the systems work by detection the differences 
between compared images before classification of detected changes. That is "Delta Data Method" [Weismuller 
et al., 1977]. At the other time, in contrast, images are firstly automatically classified by a single programme 
to produce ecological classes and are then subtracted before classified displays on the output. The second 
procedure "Post Classified Comparison Method" is more suitable for monitoring of the complex ecosystem 
780