Monitoring and modeling changes in urban development pattern using satellite images and artificial neural network model (Case study: District one of Rafsanjan city)

The excessive and incorrect use of existing natural resources requires increasing the evaluation of the components of the resources and examining the changes that have happened in the past. Therefore, detecting and predicting changes is necessary to take care of an ecosystem, especially in areas with rapid and often unplanned changes in developing countries.

The present research method is applied in terms of purpose and descriptive-analytical in terms of method. And using visual and drawing techniques, the images of TM Landsat 5 sensors in 1998, 1992, 1986, ETM+ Landsat 7 in 2010, 2004 and OLI Landsat 8 sensor in 2016 have been used. In addition, the digital maps of the region were used to check the geometric correction of the images and also as auxiliary data in the interpretation of the images and prediction of changes. After confirming the geometric and radiometric quality of the images according to the characteristics of the region, the existing land uses were classified by the method of integrated interpretation of the images of all six time periods into four classes of urban areas, pistachio orchards, barren lands and salt fields using the maximum probability method, And after validation, the average accuracy of Kappa was 83% and the average overall accuracy was 89% for the six land use maps produced.

The land use maps produced from the classification and predicted from the models in 2016, 2010, 2004, 1998 were compared and their accuracy was evaluated using the Kappa index. The results showed that the average kappa accuracy for the neural network model was 76%.

It shows the coordination between the amount and location of actual and predicted changes, and as a result, the relatively good performance of the LCM program in predicting land use changes.