Change detection has been one of the basic and crucial needs of management and exploitation of the environment and urban areas. Changes of the Urmia lake have been affected the life of the millions of the Iranian, Turkish and Azerbaijani people and their natural wildlife. Various methods and research studies have been developed to environmental change detection of the Urmia lake. The aim of this research study is to evaluate changes of Urmia Lake in the period of year 2006 until year 2011 by the Landsat 5 satellite images using the supervised fuzzy classifier. Therefore firstly, radiometric correction and image calibration have been applied to the both of the multi-year Landsat data. Then, bands number 4, 5 and 7 are selected as the references processing features according to the results of the previous works. Secondly, the multi-band differential features have been produced by subtracting the corresponding bands of the two Geo-referenced data of the mentioned years. Then two separate and multi-propose classification strategies have been applied to the produced feature space. Also obtained results are compared with the best outcome of the well-known SVM classification method.

 Firstly "two class fuzzy" classifier method on the differential features has been applied. The obtained results provided changed and not-changed classes. Achieved results for overall and average accuracy are 96.25 and 96.50 percent correspondingly. The reached results for "two class fuzzy" classifier are compared with the outcome of SVM classification and are shown the increasing about 17.04 and 10.6 percent for overall and average accuracy correspondingly.   

 Because of the uncertainty, the word "changed" for some area has always been the big challenge. Sometimes the word "changed" can have many levels, which affects the management decisions. Also the changes can be divided as "little change", "mediocre change" and "more changed" classes, according to the different human attitudes. This kind of changes in the case study of the paper can be considered as "wet salt area", "dry salt area" and so on. Therefore secondly, the other fuzzy classifier is used to extraction of the not-changed, "little changed", "mediocre changed" and "more changed" classes. In the "four class fuzzy" classification method the training and test data are remained same as the previously mentioned "two class fuzzy" classification approach, while the defined fuzzy rules are alternated. The achieved classification results for the "four class fuzzy" method are shown the overall and average accuracy about 91.72 and 90.9 percent correspondingly. Moreover the class accuracy for the not-changed, little changed, mediocre changed and more changed classes are 96.14, 85.27, 94.70 and 85.88 percent respectively.

The reached outcomes of the error matrix analysis are shown that the most correlation of the unchanged class is with the little changed class. Likewise the more correlation of the mediocre change class is with the more change class. The obtained result of the change detection for the "four class fuzzy" classification approach according to the human-oriented conceptual of relative changes in a phenomenon has the higher conceptual value.