Forecasting water consumption in logistics distribution centers using multiple linear regression

Abstract

The expansion of logistics distribution centers (DCs) demands reliable tools for forecasting water consumption to support investment and efficient management. This study develops and validates a monthly forecasting model for DCs using multiple linear regression with panel data. The replicable methodological flow included variable selection, panel construction with data from nine Brazilian DCs, model specification, estimation with robust inference (clustered standard errors), and evaluation of explanatory power, coefficient significance, multicollinearity, and residuals. Predictive performance was compared with the per-capita method. Results indicate positive and significant effects of service population, mean maximum temperature, minimum hydrostatic pressure, and numbers of flush toilets and showers. The number of faucets had a minor positive effect, while distributed cargo volume was not significant. The model showed strong explanatory and predictive performance, producing estimates closer to observed values and outperforming the per-capita approach, offering practical benefits for planning and water management in DCs.

Keywords
Water consumption; Logistics distribution centers; Multiple linear regression; Consumption forecasting; Statistical modeling