Atmospheric aerosols
; Errors
; Forecasting
; Mean square error
; Time series
; Uncertainty analysis
; Cosmogenic radionuclides
; Forecasting accuracy
; Mean absolute percentage error
; Predictive techniques
; Root mean square errors
; Sarima models
; Seasonal autoregressive integrated moving averages
; Variance of forecast errors
; Time series analysis
; beryllium 7
; atmospheric chemistry
; atmospheric modeling
; beryllium
; concentration (composition)
; cosmogenic radionuclide
; error analysis
; forecasting method
; numerical method
; prediction
; time series analysis
; tracer
; accuracy
; Adapted Mean Absolute Percentage Error
; aerosol
; air
; Article
; forecasting
; prediction
; priority journal
; Root Mean Square Error
; Seasonal Autoregressive Integrated Moving Average model
; statistical model
; time series analysis
; uncertainty
Scopus学科分类:
Environmental Science: Water Science and Technology
; Earth and Planetary Sciences: Earth-Surface Processes
; Environmental Science: Environmental Chemistry
英文摘要:
7Be is a cosmogenic radionuclide widely used as an atmospheric tracer, whose evaluation and forecasting can provide valuable information on changes in the atmospheric behavior. In this study, measurements of7Be concentrations were made each month during the period 2007–2015 from samples of atmospheric aerosols filtered from the air. The aim was to propose a Seasonal Autoregressive Integrated Moving Average (SARIMA) model to develop an explanatory and predictive model of7Be air concentrations. The Root Mean Square Error (RMSE) and the Adapted Mean Absolute Percentage Error (AMAPE) were selected to measure forecasting accuracy in identifying the best historical data time window to explain7Be concentrations. A measure based on the variance of forecast errors was calculated to determine the impact of the model uncertainty on forecasts. We concluded that the SARIMA method is a powerful explanatory and predictive technique for explaining7Be air concentrations in a longterm series of at least eight years of historical data to forecast7Be concentration trends up to one year in advance. � 2017 Elsevier Ltd