Modeling Land Use Change Using Perceptron Neural Network (Case Study: Lahijan City)

 Population growth and migration of (from or to) cities has led to the construction of unstructured and large changes in the spatial structure and expansion of cities. This causes changes in the surface of the earth and the conversion of natural effects of the earth such as soil and vegetation to the urban texture. So, the first consequence of the expansion of cities is land use change. Today, land use change and land cover have become a major challenge in many countries. Hence, the study of these changes plays a major role in the world's environmental studies. In order to better manage natural and human ecosystems and develop long-term planning, it is necessary to model land use changes and predict future changes.

The research method is applied in terms of purpose and  the nature and method of descriptive-analytic research, and the method of data collection in this study is also a library research. In this study, for land use changes during the 29-year period, images were first provided  from the website of the Geological Survey of the United States. Then, using ENVI software, the pre-processing operation was performed to apply atmospheric and radiometric corrections. Also, the specimens of educational and supervised classification of images for land use in four levels (lands, rice field, forests, gardens and Water zone) were studied. Then, in the IDRISI SELVA software, simulation was used to predict future changes using the perceptron neural network.

Before the main analysis of the data and the extraction of the information, it is necessary to perform the pre-processing operation. Then several time satellite images used in the research after atmospheric and radiometric corrections were used to prepare the land use map and Maximum likelihood algorithm was used to classify the desired classes. The selection of effective variables in predicting urban growth is an important and useful information for the user to understand the desirability of land use change. Therefore, in the present study, distance variables from the road are considered as independent static variables, and distance from the landfill, distance from the land, and the distance from the forest and gardens are considered as independent variables were used. Among the models that are used in the simulation of land use change, neural networks are multilayered perceptron. Therefore, this model was used to simulate land use changes in this study. Finally, according to the Kramer coefficient, the distance from the road has the least effect and the distance variable of the land has the greatest impact on land use change and transmission potential modeling. Then, user-potential mapping maps were generated through multi-layer perceptron neural networks for an 8-year time span. Also, in the maps produced, regions with a warm color spectrum have the greatest potential for change, and are more vulnerable to areas with a cool color spectrum.

Today, land use change and land cover have become a major challenge in many countries. These changes have a direct impact on environmental components such as soil, water and atmosphere. Which This causes changes in the surface of the earth and the conversion of natural effects of the earth such as soil and vegetation to the urban texture. Due to the fact that the city of Lahijan, like many other cities in Iran, has faced expansion of construction in recent years, so, today, the city has undergone significant changes in land use. The purpose of this study is to model and predict land use changes using the Multilayer Perceptron, . In this regard, in order to implement this model, Landsat classified satellite images for the four periods of 1989, 2000, 2010 and 2018, as well as four independent variables including distance from the road, distance from Shalizar, distance from the forest and gardens, And and distance from the land, were built to simulate land use changes. The study resulted in the generation of transmission potential mapping with the 84.58 accuracy index, which shows that the distance from the land constructed  the greatest impact and the distance from the road has the least effect on land use change variations.