Zoning Analogy of Favorable Areas for Rice Cultivation Using Two Methods of Hierarchical Analysis (AHP) and Artificial Neural Network (Case Study of Kermanshah Province, Gilanegharb City)

Subject Areas : Journal of Radar and Optical Remote Sensing and GIS

Milad Bagheri ¹ (Ph.D. Student, GIS and RS Department, Behshti University, Tehran, Iran)
Keyvan Bagheri ² (Ph.D. Student, GIS and RS Department, Tehran University, Tehran, Iran)
Bahram Soleymanpoor ³ (cMaster of Environmental Engineering, Islamic Azad University, Hamadan, Iran)

Received: 2020-07-21 Accepted : 2020-08-28 Published : 2021-02-19

Keywords: Rice, zoning, Neural Network (MLP), Analytical Hierarchy Model,

Abstract :

One of the main pillars of sustainable development in each country is the provision of adequate food at reasonable prices for the people of that community and, given the increasing population and the need for food, identifying and introducing favorable rice cultivation areas in each region is essential. For this purpose, two methods of hierarchical analysis (AHP) and a multilayer perceptron neural network (MLP) using Levenberg-Markov teaching algorithm were used in this study. The effective layers of rice cultivation were compiled and the required maps were compiled including twelve layers including land use map, average annual rainfall, average rainfall Spring season, average autumn rainfall, average temperature Spring season, average autumn temperature, slope, altitude, relative humidity, degree-day distance from the river. Analytic hierarchy model structure is used to determine the weight of layers by analyzing AHP questionnaires. Digital layers the environmental factors in the GIS environment were combined and integrated after assigning AHP weight to each layer. The grid structure is composed of twelve input layers above and eight intermediate layers and an output layer. Land zoning map of rice cultivars was obtained for both models. Thus, in the final map, the results of each of the two models, including five classes, very unfavorable, unfavorable, relatively favorable and favorable, are respectively 22, 43, 25, 7 and 3 percent for The network and results from the hierarchical model are 15, 20, 25, 22, and 18 the total area of the city. The results show that the neural network model is more accurate than the hierarchical model. The total regression coefficient of ninety-four percent of the network, which is the result of all data in the network, indicates the high efficiency of the multilayer perceptron neural network in this zoning.