Abstarct: Remote sensing means receiving information without direct contact with the surface, which requires magnetic field. The data of this technology is spatio- temporal. Understanding and analyzing the changes in vegetation cover is very important in several aspects including understanding climate changes, ecological balance and planning necessary conservation measures. The concept of neural networks has gained much significance in the analysis of vegetation dynamics using remote sensing satellite data. In the current study, an attempt is made to predict the vegetation dynamics using Landsat-5, 7, 8 data for the Shahrood city using the Long Short Term Memory (LSTM) network. The dataset consists of Landsat images from 2000 to 2018 with 4 Images from each season. The extracted indices are NDVI, SAVI , RVI from vegetation indexes. In this study, after the atmospheric correction, the desired indices are extracted and time series are constructed. Then prediction is performed using deep learning algorithm. To check the reliability of the experiment we have finalized two diferent regions for investigation. To measure the accuracy of the LSTM network, root mean square error is calculated. To measure the accuracy of the LSTM network, root mean square error is calculated. The trends of the NDVI, SAVI and RVI series are well adapted by the network and it is able to predict the future Index values with good accuracy maintaining RMSE= 0.02 for NDVI, RMSE=0.03 for SAVI and RMSE=0.05 for RVI without providing any supplementary data. By adopting the proposed method, prediction of vegetation changes can be done with more than 99 percent accuracy. In previous studies, using Markov model and logistic regression for prediction of vegetation changes using Landsat satellite data, an accuracy of up 80 percent.