جستجوی مقالات مرتبط با کلیدواژه « نزدیک ترین همسایه » در نشریات گروه « محیط زیست »

Nowadays, most developing countries are facing rapid land use/cover changes due to excessive population growth and lifestyle changes for more income. Land use/cover change includes a wide range of changes in the earth's surface and is one of the important factors that cause extensive changes in human activities and the natural environment. In addition, it is effective in all environmental functions, and in terms of sustainable development, it has a great impact on global climate changes and the resulting ecosystem responses. Therefore, the accurate and fast change detection of land use/cover is very important for understanding the relationships and interactions between humans and natural phenomena. In this regard, the development of remote sensing technologies and improving the spatial, temporal, spectral, and radiometric resolution of satellite images along with increasing the performance of this technology in terms of data integration, storage, analysis, and transmission, significantly improved the detecting, predicting and monitoring of land use/cover changes. So by increasing accessibility of satellite images, we can have a more comprehensive images of the land use situation.In this regard, different methods and data are used to detect changes in land use/cover. In general, two pixel-based and object-oriented methods are used to classify satellite images, each of which has its own advantages and disadvantages. Pixel-based classification is a traditional classification that uses a combination of spectral responses of all pixels in a training set for a given class and is very suitable for data with low to moderate spatial resolution. To perform this classification, we need educational samples, which are usually obtained from aerial photos, satellite images, or through field surveys and collection. Among the algorithms used in this classification method are neural network, support vector machine, maximum likelihood, and random forest algorithms, which provide more suitable results. But the object-oriented classification method is based on the information of a series of similar pixels in terms of different structures, the sum of these pixels and their information is called an object, and like the pixel-based method, It does not consider pixels separately, but instead of using pixels as the minimum unit, it divides the image into objects and separates spectral, spatial, contextual and textual features between them. Considering that accurate detection and monitoring of land use / cover changes is necessary for sustainable land planning and management, as well as responding to today's increasing demands. The objective of this research is to detect and monitor 28 years of land use/cover changes in Ardabil, Namin, and Astara cities between 1992 and 2019 using Landsat and Sentinel 2 images and comparing the classification of these images with pixel base and object-oriented methods.

To prepare a land use/cover map, first, the studied area was divided into agricultural, fallow, barren land, forest land, rangeland, residential, and water bodies classes. Then from each of the classes according to their area and distribution, 60 to 170 training samples were collected by GPS Garmin models 64 in 2019 and for the year 1992, training samples were taken using Google Earth. Also, a Landsat 5 image with a spatial resolution of 30 meters corresponding to July 1992 and a Sentinel 2 image with a spatial resolution of 10 meters corresponding to July 2019 were used. In the next step, in order to ensure the quality of data and image bands, the images were corrected in terms of radiometric and atmospheric errors using FLAASH and Dark Object Subtraction methods by ENVI 5.3 and QGIS 3.10 softwares. In the end, to land use/cover mapping of the study area, pixel-based methods (neural network, support vector machine, maximum likelihood, and random forest) and object-oriented (nearest neighbor) methods were used in ENVI 5.3 and eCognition softwares. Finally, in order to evaluate the accuracy of the user maps and land coverage of the study area and compare the different algorithms used in this research, the classification error matrix was extracted using one-third of the collected educational samples, which was used by Parameters of overall accuracy, kappa coefficient, producer's accuracy, and user's accuracy, the final evaluation of prepared algorithms and maps were done.

In this research, the land use and land cover map of the study area for the years 1992 and 2019 were extracted using basic pixel methods (neural network, support vector machine, maximum likelihood, and random forest) and object-oriented methods. The results of this research showed that the object-oriented method (nearest neighbor algorithm) is more accurate than the pixel-based method (artificial neural network, support vector machine, maximum likelihood, and random forest algorithms), with overall accuracy and kappa coefficient 90%, 0.80 for 1992 and 93%, 0.91 for 2019 respectively. The results showed that the object-oriented method has more ability to prepare land use/cover maps compared to the pixel-based methods, and the maximum likelihood algorithm has the least ability among the used pixel-based algorithms. The results also showed that in this period of time, the area of agricultural, fallow, barren lands, and residential areas has increased and the area of forest land, pasture, and water bodies has decreased. Rangeland with a decrease in the area of 72272 hectares and fallow land with an increase of 64010 hectares had the largest area change in the studied area. This research, which evaluated a large area using satellite images for a period of 28 years, shows many changes in land use. Most of these changes were related to the conversion of range lands to uncovered lands such as fallow, residential, and barren lands. This problem shows that the land use of the studied area was initially towards the expansion of agriculture, then due to the changes in the conditions such as the occurrence of drought and water scarcity, the tendency is towards the abandonment of these lands. According to the results, it can be concluded that in the studied area, rangeland were converted into croplands, and then due to the low income of agriculture with traditional methods, these lands were abandoned which caused soil erosion, dust phenomenon and etc. Therefore, appropriate management of land use and land cover in the study area is essential for sustainable development.

The first step in proper management of forests is having enough information about growth, development and how the forest stands structure are influenced by environmental factors. Considering the importance of forest stands in the north country and environmental values of the study area, structure of the existing species in Dashtenaz forest protected area in Mazandaran province has been investigated using the nearest neighbor indices. To investigate the structure of the study area, 56 hectare of the forest stands were statistically analyzed, and Clark and Evanes, Uniform angles, Shannon-Wiener, Mingling, DBH dominance and DBH differentiation indices were used to investigate the structure. The results showed that the mean Clark and Evans and uniform angles indices were 0.70 and 0.55 respectively, indicating a distribution between random and cluster patterns. Mean indices of Shannon-Weiner and Mingling were 0.40 and 0.06 respectively. Knowing that the study area is mostly composed of oak, the Mingling index value was calculated to be low​​. The mean DBH dominance index was calculated as 0.50, indicating the dominance of DBH of Quercus Castaneifolia and recessivity ofUlmus Glabra. Zelkova Carpinifolia was medium in terms of DBH dimension. Also average DBH differentiation index was calculated as 0.32, representing a medium difference between the trees next to each other in terms of DBH dimension. The indices used in this study can be used to describe the structure and ecological property of different species of trees.

Considering the importance of forest structure and interaction among different tree species, this study conducted aimed to determine and compare spatial pattern, heterogeneity, and dimensions of Hawthorn and Maple trees with other neighbor trees. For determining Hawthorn and Maple trees structure, 30 individual of each trees were identified and studied. Based on the results, uniform angle index for Hawthorn and Maple trees were 0/744, and 0/733 respectively, which would propose a clumped pattern for these two species. Computed mingling index values for Hawthorn and Maple were 0/877 and 0/955 respectively that show a high tree species mixture. Also, the average of crown canopy differentiation and crown canopy dimension were 0/582 and 0/322 respectively, for Hawthorn tee and it was 0/42 and 0/656 respectively for Maple tree. Based on mean values, these results show that Hawthorn had high level of differentiation but was inferior in comparing to its neighbors; whereas Maple has intermediate differentiation and is dominant to the other neighbors. The average distance to nearest neighbors for Hawthorn and Maple were calculated 7/374 and 6/278 respectively. The results of an absolute discrepancy algorithm (AD) showed a high differentiation of Hawthorn and Maple mixture in comparing to the other indices. It is concluded that maple is superior to hawthorn in dimension and because of low frequency of these two species, they hold high values of mingling indices.