Analysis of Autoregressive Models Used to Predict Climate Change over the Indian Subcontinent

Quantifying the effects of climate change on economies, and the way of life in general is of key importance. This paper marks the first instance of a novel wavelet denoised hybrid autoregressive model being used in the field of remote sensing. The key aspects of noise and non-stationarity of the input signals have been considered in the prediction unlike the previous studies. The MODIS, Sentinel-5P and CHIRPS satellite observations contain noise, part of which is intrinsic to the sensor. Additional noise is also added due to the implementation of image processing, and the wavelength of light used. Non-stationarity of the input data, makes it a challenge for auto-regressive prediction models. Towards these ends, solutions are proposed both for the denoising problem and non-stationary time series analysis. The performance of the novel wavelet denoised EMD-SARIMA model has been thoroughly contrasted with existing autoregressive and hybrid autoregressive models, such as the Box-Cox transformation and MRA based hybrid SARIMA models. Moreover, the existing literature concentrates on a single region of a nation, whereas this study concentrates on the Indian subcontinent as a whole.