Water 2025, 17(15), 2204; https://doi.org/10.3390/w17152204 - 23 Jul 2025
Abstract
This study takes the daily temperature series of Taipei City as an example and proposes a data projection method based on Empirical Mode Decomposition (EMD), which effectively resolves the challenge of modeling non-stationary sequences. According to the daily mean temperature records from 1971
[...] Read more.
This study takes the daily temperature series of Taipei City as an example and proposes a data projection method based on Empirical Mode Decomposition (EMD), which effectively resolves the challenge of modeling non-stationary sequences. According to the daily mean temperature records from 1971 to 2023, Taipei has experienced an average warming rate of 0.02 °C per year. After applying EMD, the data were decomposed into 12 intrinsic mode functions (IMFs) and one residual trend. Among them, IMF5, with a period of 352 days (approximately one year), contributes 78% of the total energy, representing the dominant climatic cycle component. In this study, daily temperatures were categorized into five thermal levels: Cold (<12 °C), Cool (12–18 °C), Moderate (18–27 °C), Warm (27–32 °C), and Hot (>32 °C). In addition, using a 5-year moving process based on the annual EMD results, the IMFs and residuals were recombined to generate 390,625 synthetic sequences per year. Results show that the monthly mean temperatures of each year’s simulations closely match the observations, capturing the non-stationary characteristics of temperature variations. The overall classification accuracy of simulated versus observed daily temperature categories ranges from 60% to 71%, with an average of 65.1%. In summary, the EMD combined with the 5-year moving process developed in this study demonstrates a helpful data projection approach with effective reconstruction of periodic structures and stable simulation accuracy. It offers practical value for reconstructing urban climate variability, conducting risk assessments, and analyzing long-term warming trends. Moreover, it serves as a vital tool for modeling non-stationary climate data and supporting future projections.
Full article
(This article belongs to the Section New Sensors, New Technologies and Machine Learning in Water Sciences)
