Prediction of Seasonal Snow Accumulation and Depletion by SARIMA Model using MODIS data

Abstract Autoregressive Integrated Moving Average (ARIMA) and seasonal ARIMA (SARIMA) models are statistical techniques generally used in analyzing and forecasting seasonal, periodic cyclic, and non-stationary time series data. This paper presents the use of the Seasonal Autoregressive Integrated Moving Average (SARIMA) method for developing a forecasting model that computes seasonal snow accumulation and depletion in the snow dominant area of the Beas river catchment. A time-series data of 8- days average snow covers acquired by Terra and Aqua sensors of MODIS (Moderate Resolution Imaging Spectro-radiometer) optical satellite has been utilized (2003 – 2018). The Box – Jenkins methodology has been performed separately by splitting yearly data into two main seasons snow accumulation (Oct. – Feb.) and snow depletion (March – Sept.). Two SARIMA models, one for snow accumulation as (1,1,1) (0,1,3)19 and the second for snow depletion as (1,1,1) (1,1,2)27 were identified by visual inspection of ACF and PACF plots using data (2003 – 2015) and then accuracy assessment has been done using performance criterion like Akaike’s Information Criterion (AIC), MSE and RSS, etc. The performance of the resulting models was then validated using data (2016 - 2018) and the comparison of both the models showed a good agreement between the simulated and observed data with a coefficient of determination (R2) of 0.829 in snow accumulation and 0.893 in snow depletion. Finally, the study advised, that the identified models could be adequate to forecast the weekly snow accumulation and depletion at least for the next 3- years to predict hydraulic events such as flood forecasting, runoff estimation, and hydropower assessment.