Enhancing Building Environmental Simulation through Climatic Variable Forecasting Using ARIMA and VAR Models with Box-Cox Pre-Processing

Abstract

Accurate forecasting of climatic variables is vital for reliable building-environment simulations and energy-performance analysis, so this study compares univariate and multivariate time-series approaches—ARIMA and ninth-order VAR—using Box-Cox transformation to stabilize variance and improve homogeneity. After transformation, series are tested for stationarity and subjected to diagnostic validation; ARIMA models are fitted to individual variables while the VAR model captures dynamic interrelationships, with model selection guided by information criteria and forecasting evaluated by error metrics and graphical inspection. Empirical results from northern Iraq show Box-Cox preprocessing improves data stability and parameter estimation for both approaches, ARIMA delivers strong forecasts for single variables, and VAR captures inter-variable dependencies but does not consistently outperform univariate models. Overall, the findings underscore the value of appropriate preprocessing and the need to weigh inter-variable correlations when choosing forecasting methods for building simulations, offering a practical framework to improve climatic inputs for sustainable building design. © 2026 NSP Natural Sciences Publishing Cor.