Results, Analysis and Discussion - Historical Land Use/Land Cover Classification Using Remote Sensing - page 159

Chapter 6
Results, Analysis and Discussion
This chapter deals essentially with the results of this thesis, that are then followed
with analysis and discussions. This chapter presents the various LULC-features
classification results (the wide major classes, the irrigated areas development
mapping, and the small detailed agricultural classes), and their accuracies. Factors
which influence the classification results are also discussed. This chapter illustrates
the various LULC-change detection mapping results (pre-classification approach
results and post-classification approach results), and discusses the successes and
the limitations of applying the various remotely sensed data used in this study, to
satisfy investigation into the objectives of the thesis. Statistical records do not
contain all elements of the irrigation projects. The second step in this research
involves employing remotely sensed data to obtain statistical numbers which
represent the areas in over past periods. Here, again, emerges the integration
between statistical data and remote sensing data in study land uses, distribution of
natural coverage and its change across time. In the first step (see Chap. 5.10 ),
statistical numbers have been useful in the spatial determination of the spread of
the targeted needed classes, and are represented in the automated classification
process in order to represent the spectral characteristics of all classes. Plus the use
of the total statistical records in evaluation, the accuracy of the classification needs
to be determined through comparison of the final results of the automated clas-
sification with the results of the traditional human-based survey. The second step,
after obtaining the training-samples from the irrigation projects which have sta-
tistical records or by using the available GPS-points as training-samples, is to
determine the statistics of the regions which have no governmental statistical data.
6.1 LULC-Classification
The spatial resolution of LANDSAT 30 m makes LULC-mapping in some situ-
ations difficult as compared to other platforms such as IKONOS (4 m), SPOT-5
(2.4 m), and QUICKBIRD (less than 2 m) (Jensen 2005 ). Some parts of the ERB