Evaluation of a multiple linear regression model and SARIMA model in forecasting 7Be air concentrations

Abstract

[EN] Forecasting the 7Be air concentration is a target value in analyzing fluctuations that could reveal important information on the motions of atmospheric air masses. In this study we first propose a Seasonal Autoregressive Integrated Moving Average (SARIMA) model with a historical data time window of eight years (2007-2014) to forecast 7Be activity. The other proposal is a Multiple Linear Regression (MLR) model for the same time period, in which the atmospheric and meteorological variables are used to forecast 7Be air concentrations. The forecasting performance of both models is evaluated by comparison with real 7Be air concentrations by out-of-sample tests for the 12 months of the year 2015. Considering the high explicative power and the consistently low accuracy of the measurements in the out-of-sample year, the proposed SARIMA model provides good forecasts of 7Be air concentrations. In contrast, the MLR model provides information on the significant meteorological variables that affect 7Be concentrations and could be useful to identify meteorological or atmospheric changes that could cause deviations in these concentrations.