Statistical evaluation of snow accumulation and depletion from remotely sensed MODIS snow time series data using the SARIMA model

Abstract

In the remote and challenging terrain of the Himalayan region, accurate measurement of cyclic snow accumulation and depletion is a significant challenge. To overcome this, an attempt has been made in the present study by applying a statistical analysis of MODIS snow time series data with the Seasonal Autoregressive Integrated Moving Average (SARIMA) model from 2003 to 2018 over the Beas river basin. The Box–Jenkins methodology of forecasting is based on the identification using seasonality, stationarity, ACF, and PACF plots; and estimation based on maximum likelihood techniques; and the last diagnostic checking based on the residual and error values have been used. Later, forecasting models have been proposed separately for the snow accumulation period (October–February) as (1,1,1) (0,1,3)19 and for the snow depletion period (March–September) as (1,1,1) (1,1,2)27 after calibration of the data (2003–2015) and the same were then validated using data (2016–2018). The accuracy assessment of the models has been checked using performance criteria like AIC, MSE, and RSS. The comparison of the forecasting models with the observed data showed a good agreement with R2 of 0.83 and 0.89 for snow accumulation and snow depletion, respectively. This research highlights the potential of utilizing satellite data and statistical modeling to address the challenges of monitoring snow cover in remote and inaccessible regions.