Land cover change detection in Minab watershed using RS and GISJalal Barkhordari Scientific member of the Agriculture & natural resources research center P.o.box 79145-1577 Bandar-Abbas, Iran jbarkhordary@yahoo.com Mohammad Khosroshahi Scientific member of Forests & rangeland research Institute khosromk@yahoo.com 1.Introduction Many types of remote sensing images are routinely recorded in digital form and are processed by computers to produce images for interpreters to study. The characteristics of the images used in this research are three land sat images. The land use term refers to man's use of the land and its cover (FAO; 1990). Although land use and land cover are conceptually different, they are used synonymously in many cases (as in this thesis). One of the main effective factors in hydrological process and soil erosion is land cover changes. In this study the land cover changes of Minab watershed in the recent decades were studied using the satellite images of study area. 2.Study area The Minab River consists of two main branches, the Rudan River in the north and the Jaghin River in the east. The combined runoff of these two rivers reaches to the Minab dam reservoir. The location of the study area is about 100 km north east of Bandar Abbas, the provincial capital of Hormozgan province in Iran  3. Materials & Methodology The following spatial (image) data and historic time series of hydrometeorological data have been used in this project. Satellite and airborne image data and maps: - Landsat MSS (1976); Landsat TM (1988); Landsat ETM (2002). - Black and white aerial photographs, at the scale of 1:55000 - Topographic maps, 1974 &1986, at scale 1:50000. Topographic maps, 2000, at scale 1:250000. Landsat MSS and TM satellite imagery have been used to:
4.Results Therefore, before the supervised land classification of the study area, agriculture and orchard land have been separated from the native vegetation (rangeland and forest) by on-screen digitizing from a colour composite of the area. Then both individual parts were separately classified using supervises classification. Then it appeared that the prepared land classification was not suitable for rangeland and forest area. In addition, a slicing procedure was applied to the NDVI in order to discriminate between rangeland and forest area. Finally, the classified maps of both the low-lying agricultural area (orchard, irrigated agriculture, rainfed and residential) and the higher elevated native vegetated area (rangeland and forest) have been joined. This was done for three images taken in 1976,1988 and 2002, respectively. Accuracy of classification result To access the accuracy of an image classification result, it is common practice to create a confusion matrix or error matrix. In a confusion matrix, classification results are compared to additional ground truth information. To obtain a confusion matrix, a raster map with ground truth information is required (which has not been used to train the classifier). This map is then crossed with the classification result, and displays the cross table. This can be done by choosing the confusion matrix command from the View menu in the ILWIS table window. (Ilwis help. version 3.11). The accuracy of the 1976, 1988 and 2002 land cover maps are showed in the related error matrices (tables 3.3, 3.4 and 3.5). The overall accuracy refers to the number of correctly classified pixels. The overall accuracies obtained are 83.8%, 84.4% and 83.65% for the 1976, 1988 and 2002 maps respectively. The user accuracy is the probability that a reference pixel has been correctly classified, and it is calculated by dividing the diagonal values of each class by the column total. The producer accuracy is the probability that a pixel classified on the map represents that class on the ground (Anderson, 1976). The values obtained for the producer are 80.95%, 75.07% and 81.13%. The user accuracies are 83.8%, 84.4% and 82.72%. These all apply to the 1976,1988 and 2002 land use maps respectively.  Average accuracy= 80.95% Average reliability= 83.79% Overall accuracy= 81.59%  Average accuracy= 75.07% Average reliability= 84.38% Overall accuracy= 86.17%  Average accuracy= 81.13% Average reliability= 82.72% Overall accuracy= 83.65% Re: Residential I.A: Irrigated agriculture Or: Orchard P.R: Poor rangeland M.R: Medium rangeland G.R: Good rangeland Ra: Rainfed Fo: Forest 3.4 Land use maps Three land use maps have been prepared using the available satellite images, aerial photos and other field data for the years 1976,1988, 2002 . The eight different types of land use that have been distinguished in these maps are described in the following.
In 1976, about 90 % of the study area of the Minab catchment was still covered with forest and rangeland. These classes consisted of 50% poor rangeland, 38% medium rangeland and 3% forest . Only in this year the agricultural lands are characterized as rainfed. 3.4.2 Land use map, 1988 In this year, the poor rangeland, medium rangeland, good rangeland and forest covered 87%, 8.5%, 1% and 0.5% respectively. The extent of orchard, irrigated agriculture land and settlements covered 1%, 2% and 0.2%. In this year the extent of rainfed agriculture was zero (table 3.6). In comparison with land u se map of 1976,severe reduction of good, medium rangeland and forest has been noticed. Increase of agriculture lands and residential also have been observed. 3.4.3 Land use map, 2002 In this year, the extent of poor rangeland, medium rangeland, good rangeland and forest land covered 87%, 5%, 1.5% and 1.33%. Orchard, irrigated agriculture and settlements covered 3%, 1.4% and 0.3% respectively (table 3.6). In comparison with land use map of 1988, the most important change is due to cultivated land change to orchard.   Figure 2 land use types of the Minab catchment, 1976,1988 and 2002 3.5 Land use changes 3.5.1. 1976-1988 period: In this period, the maximum change is related to a decrease in the extent of medium rangelands and an increase in poor rangelands, 30% and 34% respectively. This change shows the disappearance of the natural vegetation cover in the catchment. There are different factors accounting for this, such as drought and mismanagement of rangelands. This is based on interviews with local villagers in the catchment and experts of the Natural Resources Office. It was observed that after 1978, the protective programs and exploitation system in the region did not have any follow-up, and a negative trend without any care for natural vegetation was initiated. Also, with the carry-out of agricultural development programs and digging of deep wells, the coverage of irrigated lands in this period has increased (1.2%). The change was from rainfed lands to irrigated lands (table 3.8 and fig.3.6). 3.5.2. 1988-2002 period: In this period, despite the slight increase in rainfall, the natural vegetation cover did not recover to the situation in the late seventies. Only the extent of the orchards increased with about 3%, as an effect of long duration of a drought period combined with mismanagement of the exploitation of the rangelands. These are the two main factors that reduced the probability of natural reclamation of rangeland during these relatively wet years (table 3.7).  3.6 Conclusion In 1976, about 45 percent of the watershed area was covered by rangeland and natural forest (referred to as medium rangeland, good rangeland and forest). Due to continued overgrazing, rangeland cover decreased to 10 percent in 1988 and to 8 percent in 2002. During a period of ca. 25 years, three main land use classes at a large scale have replaced these fertile rangelands. These are characterized by: poor natural cover (e.g. poor rangeland, bare soil and rock out crop), agriculture area (e.g. irrigated agriculture, rainfed agriculture and orchard) and residential area Before 1978(The change time of policies of natural resource management in Iran) basin’s vegetation cover especially medium rangeland had a relatively good condition because climatological condition and management were suitable. After 1978 because of changing in control structures on natural resources that was the beginning of drought period the vegetation cover starts reducing and in the years after drought period, the natural important was not possible. These changes have an important effect on hydrologic regime of the basin in this period. Reference
|