عنوان مقاله [English]
Lakes and wetlands are one of the richest biological resources in every natural ecosystems that any type of damage or destruction on them, affect on the other biological parameters of every ecological environment. Lakes, as the second water resource which cover 1.8 percent of the earth surface, are formed by means of the internal and external processes of the earth.karstic lake is one type of the lakeswhich forms via karstic process. In studying the lakes, beside of the parameters that are relevant to the water quality, depth and area changes of the lake indicate the situation of the lake, too.In the recent years, most of the studies about the lakes bathymetry were accomplished by means of the radar images or field studies butthese methods are very costly. So, presenting a method based on the multispectral images and with much lower cost, can be instrumental.In this research, one of the most important and largest freshwater lakes in Iran, were examined from the view point of the relative depth variation and also changes in lake area in period of 14 years bymeans of the methods which based on the satelliteimage and without field measurements. For this purpose, non-thermal bands of the TM-5 sensor of the landsat satellite, were used.
In this research, non-thermal images of the landsat TM-5, were used for examination of the relative depth and lake area changes during the period of 1985 to 1999. We tried to using the data which has been received in the same month of the yearin each of the two mentioned years, because, the changes that we calculated, not due to the seasonal changes. In order to achieving the mentioned purposes of this research, the following steps were done, respectively : 1- pre-processing of the data. 2- calculate the area of the lake in two years. 3- performing the Spectral Angle Mapper (SAM) classification method. 4- performing Binary Encoding Classification (BEC) method. 5- performing the Spectral Information Divergence (SID) method. 6- using the Sieve filter. 7- producing the classified image of the relative depth,using the Log Ratio Transform and checking the relative depth variations in the mentioned time interval.
Results and discussion
After applying the three classification methods ( SAM, BEC, SID ) for each two years, the results were evaluated to select the best method as the base for calculate the extent changes of the lake, in time range of 14 years. Then, in order to evaluate the precision of the methods that were used in classification the area of the lake, three methods consist of confusion matrix, kappa coefficient and the JM indexwere used. The results showed that the SID and SAM methodshave higher ability. Among them, the SID method with overall accuracy equal to 83.75% and kappa coefficientequal to 0.675, had the best performance. For more accurate examination, we can calculate the changes by means of overlaying the layers.Table 1, figure 6 and figure 8 show the significant increasing of the lake extent. Also, the result of the relative depth calculation, indicate the changes in the lake relative depth value, in 14 years. ( 1985 to 1999 )with regard to the figure 9 ( a,b ), the yellowparts indicate the lower depths and the blue parts represent the deeper areas. As can be seen, a major change has been occurred in the lake relative depth that can be arising from the increasing depth in 1999 and also changes in the situation of lake sedimentation in some areas.For more accurate examination, by means of creating a tree structure classifier, a new classified image ( figure 9, b,c ) is obtained which comprises 4 class land, shallow area, deep and very deep area.( figure 9 )In 1985 and in the eastern part of the lake, we see more of the deeper parts in the lake.in fact, we can say that one of the reasons for this increase in the lake area, is, that the depth of the lake become more uniformand also, decrease of the deeper parts of the lake. These changes are discussed in details in the table 2.For a better understanding of the depth changes conditions, we can make a 3D view of the lake-bed by means of combining the produced image for the depth in each year with the RGB image of the same year.
This research showed that the method based on the use of the reflective bands, can be used to study the changes of the depth in the shallow lakes such as the Parishan lake and can be replaced for the radar imagery and field based costly methods.The SID method showed a high ability On the field of the classification of the lake surface from the other features, with 20% increase in the precision than the SAM common method and 37.5% improvement than the BEC method. Also, our studies showed the high ability of the filtration method which was used, inremoving the isolated and incorrect pixels from the classification.Evaluation results of the lake area changes indicate the 46.42% growth in the area of the lake in a 14-years period. Most of this increase, take place in the western area of the lake, While, the changes of the relative depth show that the many parts of the shallow areas in 1985 has been became to the deeper areas in 1999. Therefore, in this period of time, the western parts of the lake are the most variable parts. From the viewpoint of the depth changes and in the lower level( as compared withthe western parts of the lake ) , there are the eastern edges of the lake which most of the deep areas of them havebecame change in to the average depth, in 1999. In fact, the average depth part of the lake, had the minimum changes from 1985 to 1999 which mainly include the central parts of the lake.