The analysis of vegetation dynamic based on NDVI time series using ARIMA Model

NDVI is a measure of surface reflectance widely applied for monitoring vegetation dynamics at regional and global scales. Autoregressive integrated moving average (ARIMA) analysis was used to model NDVI time series extracted from ETM+ images from 1999 to 2019. First NDVI for each year was calculated using Red and NIR bands, then the values of 411 randomly selected samples on the NDVI images were extracted, and the average of the NDVI of the samples was considered as the representative of NDVI for each year. Therefore, the time series of NDVI was produced. Using ACF and PACF correlogram, the order of model parameters was estimated, and the ARIMA model was performed and verified. The residual analysis of the selected models showed that all models, including ARIMA (2,0,2), ARIMA (2,0,3), and ARIMA (2,0,5) with low errors and high coefficients of determination could capture the linear structure in the NDVI time series. Furthermore, the results of the validation models indicated that ARIMA (2,0,2) had the smallest error in all three error criteria, equal to 0.006, 0.003, and 4.40 for RMSE, MAE, and MAPE errors, respectively, and therefore is the best model for simulating vegetation dynamics in the study area.

The analysis of vegetation dynamic based on NDVI time series using ARIMA Model Keywords:

NDVI , Time series , ARIMA