Land Use Classification and Determining the Pattern of Changes for 2014-2017, using OLI Sensor’s Data

Land use is an aggressive process applying to human activities and different uses accomplished over land. It can be argued that human actions can lead to significant changes in current state of earth’s surface. Changes in surface cover (land cover change) may in turn lead to alternations in balance of energy, water, and geochemical fluctuations at local, regional or global levels. Thus, studies on different land uses changes seem necessary in general environmental evaluation. LULC change detection include implementing multi-temporal Remote Sensing (RS) knowledge to analyze the historical LULC data (maps) and therefore helps in determining the trend of changes associated with LULC properties.

Image processing and performing supervised image classification helps to extract information from imageries. In this study, ENVI 5.3 software was used for processing two selected imageries in this project (2014 and 2017). Five LULC classes were established as forest, bare land, vegetation, mountain and water body. For each LULC class, 500 samples were collected at least and used for the supervised classification of images in ENVI. About half of these samples, which were used as “training samples” were collected from the study area through Land Surveying Geographical Positioning System GPS (ground truth data) and Google Earth images. The first step in pre-processing of LANDSAT 8 OLI data in this study referred to the collection of training samples for each class and validating the geometric accuracy of Landsat images, while the next step belonged to the conversion of DNs into At-Satellite radiance using algorithms such as FLAASH. Two dated Landsat images were compared via the supervised classification technique. In this classification technique, two or more images with different dates are independently classified. Maximum Likelihood Classification (MLC) algorithm as a supervised classification method was carried out using training areas and test data for accuracy assessment in ENVI 5.3 and accuracy assessment was done for both images using ENVI v5.3.

In order to recognize the past land use pattern of Tarom, researchers first focused on imagery of Landsat 8 ETM+ for the year 2014. Summary of supervised classification accuracy for the 2 different time frames (2014 and 2017) found from accuracy assessment showed that the highest accuracy was found for 2014 supervised classification (92.16% accuracy). Kappa value is also used to check accuracy in classification and having a Kappa value (0.81–1.00) denotes almost perfect match between the classified and referenced data. Different LULC classes had been recognized and used as the base map. From the identified LULC classes, Mountain area by 3524 km2 (62.75% of total land area) was the highest category, after which, came bare land areas with 1295 km2 (24.0%) coverage and vegetation area with 194.6 Km2 (3.7%). Forest was the next class with (2.7%) coverage whereas, water body (1.4%) and unknown pixels 8 km2 (0.15%) specified the least amount of coverage, respectively. Based on the 2017 image classification results, the highest category belonged to mountain area (3532 Km2, sharing 67.7% of total area). The remaining land uses were bare land (23.21%), forest (2.73%), vegetation (4.3%), and water body (1.75%). The unknown and uncategorized pixels were identifiable in this stage that shared 0.31% of the total area. The relative changes in land use and land cover from 2014 and 2017 images showed some irregular patterns in the study area. Land-use change from this period showed positive changes in most of the categories. About 31.4 Km2 of vegetation area had increased in 2014–2017 period which showed a positive change of (+16.14%). While a negative decrease (83 Km2, -6.4%) in bare land category. The results showed that the extraction of adequate samples from different classes of land cover/land use would increase the possibility of correct distinction of image pixels received from the satellite and accurate extraction of LULC classes. Thus, obtaining accurate results from the classification of images via the maximum likelihood method is depending on adequate and appropriate training samples. The trend of land-use changes found in this study, especially percentage increase in forest land and a decrease in bare lands will be helpful for policymakers to make appropriate decisions.

Land cover is the physical material at the earth’s surface and an essential variable which links the physical environment by human activities, and land use is the description of how the land has been utilized for the socio-economic activities purposes. Population growth increases the demand for food, water, and energy, which causes a prompt change in land cover and pattern of land-use. The mentioned process depends on the social and economic development of the nation. In order to have appropriate and unrestrictive management of natural resources (water and soil), it is necessary to have complete information about the pattern of land use and its alteration pattern over time. Thus, it can be concluded that remote sensing is a proper technique to investigate the land-use changes using satellite imagery. Spatiotemporal analyses of LULC help us to manage the environmental changes, which are an appropriate tool for decision-makers on water resources’ to enhance their decisions. In the presented study, LULC map for Tarom basin, Iran, acquired from OLI sensor data sets (Landsat-8) by applying a pixel-based classification method (MLC) with the aid of remote sensing technology. The results that are presented in this study proved the usefulness, effectiveness and also convenience of the MLC technique for generating land-use maps by using a free archive of Landsat data and processing the digital images through the ENVI software. Accuracy assessment using overall accuracy and kappa coefficient for 2014 and 2017, shows the performance of the used algorithm. What matters most in this regard is the accuracy, speed, and quality of land-use maps. In the present study, it was shown that due to high speed and accuracy in generating land-use maps of Tarom, MLC method, would act as the best classification method in this area. However, it is suggested to classify the data by using other methods and compare the results with image outputs provided by Landsat 8 satellite.