فصلنامه سنجش از دور و GIS ایران
پیاپی 1 (بهار 1388)

In order to solve some problems, as well as to simplify or facilitate some analyses in graphs, some changes can be made in the graphs. Dual graph is one of these changes. Linear dual graph is a type of dual graph that is proposed for presenting graphswith weighted nodes. In this paper, linear dual graph calculus that is based on the linear dual graph is introduced. For this purpose, at first, Linear Dual (LD1) and inverse Linear Dual (LD-1) is introduced, and then, the way of their extraction is explained. After that some applications of this calculus is explained. One of its most important applications is in specification the Hamiltonian cycle in graphs. In other words, by transportation between linear dual graph and the primal graph,we can convert the Hamiltonian cycles whose specification has always been so difficult, to Eulerian ones which can be easily recognized.

Landuse planning is a systematic attempt to model the interaction between human activities and environment, and arranges landuses for sustainable development. One of the main stages in landuse planning is calculation of land ecological capability.This stage includes definition of landuse requirements, assessment of land characteristics and corresponding matching, between these two. Study of definitions and methods related to the calculation of ecological capability shows that it is not possible to calculate an exact value for the suitability of land unit for a specific landuse. In this context, many inaccurate and uncertain processes can be modeled mathematically, using fuzzy logic. Such models can be used for decision making in uncertain conditions.In Iran, Makhdoum’s model is widely applied for ecological capability assessment, in ifferent scales and areas. This is mainly because of its comprehensiveness regarding socio-economic and environmental characteristics of Iran. In this model, phenomena like soil type, climate and slope, which are continuous changes in the nature, are presented in crisp maps. On the other hand, the weight and relative importance of various factors are not covered in Makhdoum model. Methods like fuzzy Weighted Linear Combination (fuzzy WLC) and fuzzy Ordered Weighted Average (fuzzy OWA) are needed to define theweight of ecological characteristics maps. Therefore, fuzzy inference rules are used for the assessment of ecological capability.In this research, Makhdoum model is implemented using crisp and fuzzy approaches. Considering the simplicity of implementing fuzzy logic in raster GIS environments, pixel is selected as a spatial unit, for spatial representation of ecological models. In crisp approach, the ecological models are generated and land suitability index for spatial units is calculated. It should be mentioned that, in Makhdoum’s original model, the suitability of each pixel is defined as either zero or one. Land ecological capability calculation using fuzzy logic includes three stages; normalization and fuzzification of environmental maps, integration of maps using fuzzy inference, and defuzzification of the results. In this research, first, all input maps (slope, elevation, spect, soil texture, soil depth, erosion, climate and geology) and output maps (ecologicalcapability for different classes of agriculture, pasture, urban and rural residential and industrial) are normalized and fuzzified, and the fuzzy membership functions are defined for the inputs and outputs of the model. In the second step, fuzzy data are integrated using fuzzy inference rules. In this process, the fuzzy if-then rules are defined and the related rule-base is created using expert knowledge and the fuzzy membership functions of the input and output of the model. Finally, in the defuzzification stage, the output of the fuzzy inference, which includes several fuzzy numbers, are converted into one crispnumber, using gravity center method and Mamdani decision model. In this research, the fuzzy ecological capability maps are generated for different landuses in Borkhar and Meymeh township.Usually, either the administration units or the ecological ones are considered as spatial units for decision making. According to Makhdoum model, the suitability of each ecological unit and each rural district, for different landuses are calculated by aggregating the suitability of pixels inside these units. Spatial analyses and development capabilities of ArcGIS Ver. 9.3 are used for preparation and analysis of ecological maps. In addition, Matlab Ver. 7.6 is used for defining fuzzy membership functions, creating fuzzy rule-base, and doing other fuzzy calculations.Using fuzzy logic, the ecological characteristics maps can be produced and integrated much closer to the reality, although the amount of calculations grows significantly. Implementation and comparison of Makhdoum model in crisp and fuzzy approaches show significant improvement in the environmental characteristics maps, especially around the boundaries of features and where thematic classes are changing. The method developed and applied here is independent of the number of landuse types and criteria, and can be used for other conditions with minor modifications.

An urban heat island indicates an area with relatively warm surface, and most commonly the areas of human disturbance such as towns and cities. The cause of such a phenomenon is distributed in two major components. One factor is the changeof land cover. Vegetation and water area that play important roles in alleviating the rise of air temperature are decreasing, and artificial land covers that have high heat capacity like asphalt and concrete are increasing. Another factor is the increase of artificial heat mission. Especially the emission from automobile and artificial cooling has been increasing recently.Now, there are lots of studies dealing with these topics, but the studies from the viewpoint of quantity are not enough. In this research the urban heat islands of Tehran, as the most important population center and one of the most important industrial center of Iran, is investigated. The aim of this work is investigating the relationship between land cover proportions and UHI with ETM+ images of Tehran. To achieve this, for mapping land surface materials with distinct physical properties from Landsat ETM+, the linear spectral unmixing method was utilized for endmember fraction estimation. The transformed ETM+ image was unmixed into four fraction images (vegetation, soil, high albedo and low albedo). Impervious surfaces were then computed from the high and low albedo images. Multiple regression models were further developed to reveal how land surface temperatures were related to urban biophysical descriptors (i.e., impervious surfaces, green vegetation, and soil). Results indicate that impervious surfaces, because of high heat capacity and anthropogenic heat emissions into the air and dry soil through high heat capacity, were positively correlated; while vegetation was negatively correlated with land surface temperater. Also industrial area has the most positively correlated with land surface temperature, because the anthropogenic heat flux in industrial areas is high.

The growth of technology and industry has caused air pollution in many countries all over the world. Since air pollution has a direct effect on the health of human beings, animals and plants, much attention have been paid by decision makers, experts and researchers to this problem. Iran, especially Tehran city, is not immune to this phenomenon as well; and huge financial, physical and social losses are occured each year.There are many different methods for estimating air pollution. Some of these include: Proximity, Interpolation, LUR, Diffusion and so on, which each of these methods includes some advantages and disadvantages. This study tried to determine spatial and temporal changes of the air pollution. To achievethis, interpolation method was applied. Moreover, to specify the factors that had effects on air pollution, another method- namely, land use regression (LUR)- was used. Since the most important pollutants of the city are CO and PM10, using the Statistic MSE (Mean Square Error), different interpolation methods were first compared to produce air quality maps. Then the air quality maps for these two pollutants were produced for all days during 2004-2005 by optimum interpolation method. Also all of the produced maps were classified based on Air Quality Index(AQI).Air pollution is associated with different factors such as topography, climate, population, transportation system and industry. To determine the most significant factors LUR method was used. Then, using the same method, modeling of the mentioned pollutants in spring was conducted. LUR method consists geospatial information system and multivariate regression methods. In this study, the chosen locations for measuring the pollutants were the same locations where the station were situated. Then traffic volume, land use, population, elevation and distance of the stations from neighborhood roads were measured in different buffers. Finally, by using Multivariate, Backward, Forward and Stepwise regression methods, the relationships among the selected pollutants (as dependent variables) and the other variables (as independent ones) were obtained. The results showed that the best method for estimating CO pollutants is Cokriging method with three secondary parameters- namely temperature, speed and wind direction. However, Spline method for PM10 pollutants yielded the better results.Also, results demonstrated that in December and April CO volume and in March and July PM10 volume were at their highest levels. Also it was found that the most CO pollutant volume were seen in 6, 7, 11, 12 district; while the most PM10 pollutant volume belonged to the 9, 10, 16, 17, 18 district. Results of LUR model demonstrated that the most important factor affecting CO pollutant was the volume of traffic. The most important factor affecting the PM10 was the existence of industrial locations and the final R2 of CO and PM10 Pollutants models were 0.47 and 0.62 respectively.

Mines, as the main sources of raw materials used in the industry, have an inevitable role in today’s human life. Exploration of new mines is an obligation for more raw materials. New methods, technology and information have had great effects on the mine exploration industry. Due to the power and capability of Geographical Information Systems (GIS) in processing the spatial data, they can be effectively used in mine exploration.Mines are recognized based on the preliminary data which are collected in the field. GIS can be used in the processing, weighting, and extracting information out of the data to facilitate the mine exploration process. What is quite important in this respect is nothing but the determination of relative weight of different data layers to combine them. Assigning different weights to the data and/or using different methods to combine them, have a great impact on the final result. Therefore, studying and comparing different weighting schema are objectives as well as the subject of this paper. Weighting methods can be classified into two main groups of data-driven and knowledge-driven categories. For each group one example is studied in this research. To be more specific, Analytical Hierarchical Process (AHP), as a knowledgedriven method, and Artificial Neural Networks as a data-driven one are studied and implemented in this research.These different methods are used to produce the mineral potential map as one of the last steps in the mine exploration process. The information collected from the boreholes are used to evaluate the results. Numerical experimentations showed that the artificial neural network used in this study is the most successful method. It is shown that knowledge drivenmethods are very much affected by the degree of knowledge and specialization of the experts. Meanwhile, different methods of using the knowledge were resulted in different solutions. In this paper Fig. 1 shows the process of calculation of weights in AHP method, while Fig. 2 represents a general form of an artificial neural network used in this study. Fig. 3 illustrates case study area and Fig.4 shows the stages of data preparation. Fig. 5 illustrates input maps and data such as boreholes. Fig. 6 and Fig. 7 show procedures of implementing AHP and artificial neural networks respectively. Fig. 8 represents the percent of correct predictions for implemented methods and at last, Fig. 9 illustrates two mineral potential maps produced by AHP and artificial neural networks. Finally, Tables 1 and 2 represent the weights obtained from AHP and artificial neural networks respectively.

Interpolated data are usually used in many engineering and planning activities such as geology, climatology, natural resources and hazard management. Root Mean Squared Errors (RMSE) is the most frequently used indicator for error assessment and statement. Unfortunately, this indicator is mainly non-spatial and it does not provide the required spatial details about the local reliability of the interpolated layers. Therefore, development of more efficient and practically useful methods for spatial evaluation and visualization of interpolation errors are of prime importance.In this research, after reviewing some of the important limitations of RMSE and interpolation methods, a practical approach based upon the triangulation, for modeling and mapping of the reliability of interpolation has been proposed. Triangles formed by application of the delauney triangulation to sample points have been used as the main spatial entities. Three characteristics of triangles including the area, shape (modeled by the ratio of perimeter to area), and variance of the values of their 3 corners have been used to model the interpolation errors in triangle levels. The reason for choosing theseparameters is their important role in creation of errors, as well as simplicity of their calculation in GIS environment.Point data of elevation from part of Tehran metropolis has been used as the case study for examination and demonstration the usefulness of the proposed approach. Modeling has been based on the least squares fitting in a multiple regression framework. The fitted model has been used for prediction of interpolation errors. Where its validity has been evaluated by the independent, well distributed test sample points with a known elevation.Results of the proposed approach have been encouraging, and close relationship between the actual and predicted (bymodel) errors have been observed. Production of the map of interpolation errors, as in the proposed approach, can be useful for efficient use of the interpolated data. Also, these maps can be used as useful guides for collection of additional samples where the improvement of the interpolated data quality is required. More close examination of the proposed approach in a wider and diverse environmental condition has been recommended.

One of the most applied supervised classification method is Maximum Likelihood (ML) in which a series of statistical parameters such as variance-covariance matrices are estimated. In Hyperspectral remote sensing images, due to the limited number of training samples and their high spectral dimensions, the probability of having singular matrices and/or reduction the accuracy of classification is plausible. To solve this problem, different approaches such as reduction of number of features or ensembling of classifiers can be used. In theory, the acquisition of large number of training data set is feasible, but it is very time consuming. Then in practice there are always some limitations where we believe methods such as Feature Extraction algorithms or ensemble of classifiers for dimensionality reduction can solve that. In this research, Nonparametric Weighted Feature Extraction (NWFE), as well as classifier ensembles, are used simultaneously. For constructing multiple classifiers, manipulation of input features are done in which input feature space is divided into multiple subspaces by using class based NWFE method. A ML classifier is applied on each of the prepared feature subsets, and finally a combination scheme was used to combine the outputs of each individual classifier. In order to fuse multiple classifiers, a measurement level method is suggested using the mean rule. The results show an overall accuracy of 85.67% for NWFE method and 89.26% for classifier ensemble. For method suggested in this paper the overall classification accuracy of 89.34% was achieved. The results indicate significant improvement in classification accuracy, compared to the two methods on which this method is based upon. Despite the closeness of these two accuracies, because of less complexities and feasibility of parallel calculation, the suggested method is preferred.