Trend and Time Series Analysis of Vegetation Dynamics Using Satellite Data: A Case Study of Uttarakhand, India

In this work, aim is to collect time series data for Normalized Difference Vegetation Index (NDVI) band using google earth engine (GEE) and MOD13A1 V6.1 product for the region of Uttarakhand, Uttarakhand districts and Himachal Pradesh. Thereafter investigation and comprehension of the viability of using MODIS NDVI satellite data time series to identify trends and give a forecast model. Time series data was collected using Google Earth Engine for NDVI indices for the period of the year 2010 to 2022. Trend analysis and time series analysis were performed for collected data. NDVI time series data set is collected using GEE for Uttarakhand state, its districts and Himachal Pradesh State. The Mann- Kendell (MK) method is used to find trend analysis of above regions. NDVI time series data is divided into train and test dataset. Five forecasting models are used to forecast NDVI time series dataset i.e., Long short-term memory (LSTM), Bidirectional Long short-term memory (BiLSTM), Support vector regression (SVR), Autoregressive Integrated Moving (ARIMA), Adaptive Neuro fuzzy interference system (ANFIS) models are trained using train data and are used to generate the predicted value. The predicted value is then compared with test data using various metrics for forecasting NDVI times series. Trend analysis of NDVI shows an increasing trend in NDVI values for Uttarakhand and its districts as well as Himachal Pradesh. ANFIS model resulted R2 value of 0.6702, Stacked LSTM model resulted R2 value of 0.7541, Bidirectional LSTM model resulted with highest R2 value of 0.8365, Autoregressive Integrated Moving Average (ARIMA) model resulted with lowest R2 value of 0.153, SVR model resulted with R2 value of 0.6719.