Document Type : Research Article
Authors
1 MSc. Student, Environmental Engineering, Saba University, Urmia, Iran
2 Urmia Lake Research Institute, Urmia University, Urmia, Iran
3 MSc in Water Resources Engineering, Urmia University
Abstract
Ignoring structural changes in urban water consumption can undermine forecasting accuracy. Since most methods assume only permanent shifts, this study examines both persistent and reversible consumption regimes in Salmas City (2004–2025) using Bai–Perron, Pettitt, SNHT, and Markov-switching models in R. Bai–Perron, run under mean- and population-based specifications, served as the primary framework, with Pettitt and SNHT providing independent validation and Markov-switching testing whether regimes could reverse over time. Water consumption rose 149%, far outpacing the 44.7% growth in population. All Bai–Perron specifications identified 2015 as the primary breakpoint, closely matching Pettitt (2014) and SNHT (2015). The population-based model also suggested a possible second breakpoint in 2021, though this wasn't confirmed by the other tests. Unlike Bai–Perron's permanent breaks, the Markov-switching model's smoothed probabilities suggest the series leaned toward a lower-consumption regime during 2011–2017 before shifting back — though with only 22 observations, this is best read as a statistically inferred pattern rather than a confirmed change, and one that depends on including population as a covariate. Overall, relying on a single method risks an incomplete picture of these dynamics.
Keywords
Main Subjects
)