Preparing the distribution of Seidlitzia rosmarinus in Semnan East rangeland using ANN model

Artificial Neural Network (ANN) is new information processing structures that uses special methods for biological neural networks. The main purpose of this study was to modeling of Seidlitzia rosmarinus distribution in northeast rangelands of Semnan by ANN model. For this purpose, vegetation sampling was carried out in each vegetation type along three transects of 750 m, on which 15 plots were established with an interval of 50 m. Soil samples were taken from two depths of 0-20 cm and 20-80 cm in starting and ending points of each transect. To provide the prediction map of plant species distribution, different layers of environmental factors used in the model are required. The geostatistics method was applied for mapping soil properties based on the prediction model obtained from ANN method for S. rosmarinus. The back-propagation neural network with three-layer- perceptron network was designed to generate the ANN model and seven neurons in the input layer, ten neurons in the hidden layer, and one neuron in the output layer were used. The accuracy of the prediction map was tested with actual vegetation maps and the Kappa coefficient was calculated to be 72%, indicating a very good agreement. Results showed that this species is distributed in rangelands with a pH of 8.1-8.3, an EC 0.22-0.26 dS/m, in a silty-sandy textured soil, and an altitude of 1600-1750 meters. it is highly correlated with lime and pH in two depths.