Results, Analysis and Discussion - Historical Land Use/Land Cover Classification Using Remote Sensing - page 184

with ETM+-data. Using TM-data, the accuracies were close. The greatest prob-
lems were found while classifying the artificial surfaces, especially by using MSS-
and TM-2007-data. The classification accuracy of barley and corn was relatively
slight. The details of these classification problems were discussed for each LULC-
class and for each used remotely sensed data in Sects. 6.1.1 , 6.1.2 , and 6.1.3 .
For evaluation of the change detection products, for instance, Macleod and
Congalton ( 1998 ) proposed an adjusted change detection mapping products
method for evaluating their accuracy. This process is based on Congalton ( 1991 )
and requires the regular confusion matrices to be applied on bi-temporal resulted
change maps. These matrices can represent all combination classes of occurred
change. A simplified no-change change matrix was also proposed.
The essential problem for assessing the accuracy of change detection products
is the gathering of truth reference data, where the conditions of the initial time
period (i.e., 1975) cannot be revisited as the land use has been significantly
changed. Therefore, accuracy assessment depended on the remotely sensed data of
the final time (i.e., 2007), in addition to ancillary data.
6.3 LULC-Change Detection Mapping
This section gives a historical view of the different LULC-features in the study
area. It describes the rate of their changes over the last 32 years, in particular the
irrigated agricultural lands. The results are essentially presented by maps, statis-
tics/tables and graphs.
6.3.1 Pre-Classification Results
Each change detection process analysis/result, whether a pre-classification
approach is used or a post-classification approach (see Sect. 6.3.2 ), consisted of
four major components, which were: Measure of changes/quantity. This provided
the quantity of the occurred change and measured the area/s of change/s to provide
statistical numbers, i.e., how much was/were the change/s?
Pre-classification approach results (Fig. 6.10 ) were generalized but very
effective in relation to estimating the occurred change on the cultivated areas,
especially when these areas were vegetated and not fallow. The total change in the
study area (5,062,082 ha, 100 %) between 1975 and 2007 was about 600,967 ha
(11.93 %), in which 238,646 ha (4.74 %) was changed from natural vegetated
areas to bare areas, and 362,321 ha (7.19 %) changed from bare areas to cultivated
areas (especially to irrigated). Areas recording no change were about
4,461,115 ha, 88.62 %. In comparison, the results of the three previously-men-
tioned approaches generally changed classes with those that resulted from
applying the post-classification approach (see Sect. 6.3.2 ), but for the duration