Scholarlink

<p>Understanding the temporal and spatial dynamics of urban thermal patterns is critical for assessing long-term environmental change in rapidly urbanizing regions. This study integrates land use/land cover (LULC) analysis and machine learning-based wavelet modeling to investigate urban heat variability in Delhi from 2018 to 2023. LULC maps were developed using Landsat-8 and 9 imagery with high classification accuracy (Kappa: 0.86 in 2018, 0.89 in 2023). Between 2018 and 2023, built-up areas expanded from 22.43 to 30.53%, while green spaces declined from 42.47 to 35.36%, intensifying the urban heat island (UHI) effect (mean UHI rose from 0.43 to 1.69). The Continuous Wavelet Transform (CWT) revealed dominant periodicities across 52 scales, capturing both short- and long-term climate cycles. IMD temperature data showed high Lag-1 autocorrelation (0.96), indicating persistent thermal patterns. Seasonal decomposition indicated rising trends, with peak temperatures of 35.5&#xa0;°C in 2023 and projected cooling to 14.48&#xa0;°C by 2026. The Prophet model demonstrated high forecast accuracy (MAE = 1.94, R² = 0.88), identifying anomalies in 2025 and 2029. These results underscore the need for real-time environmental monitoring and green infrastructure planning. Future work should integrate high-resolution remote sensing and hydrological modeling to refine climate adaptation strategies in urban regions.</p> Graphical Abstract <p>This study uniquely integrates wavelet-based time-frequency analysis with machine learning forecasting to capture multi-scale urban temperature dynamics, offering enhanced precision in predicting thermal trends and climate anomalies. The assessment of urban thermal climate was done using long-term IMD temperature data. It proceeds with trend analysis to identify persistent thermal patterns and inter-annual variability. The process then applies time-frequency analysis using the Continuous Wavelet Transform to detect dominant periodicities. Machine learning-based forecasting, including Prophet modeling, predicts future temperature trends. This integrated approach highlights the significance of multi-scale analysis and predictive modeling in understanding urban climate dynamics and supporting resilient urban planning strategies.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Mapping Time-Frequency Climate Shifts in Cities: A CWT and ML-Based Predictive Study

  • 期刊论文

  • 发表日期：

    2025年8月25日

查看全文
  • Sagar Tomar,
  • Kishor S. Kulkarni,
  • Aqil Tariq

摘要

Understanding the temporal and spatial dynamics of urban thermal patterns is critical for assessing long-term environmental change in rapidly urbanizing regions. This study integrates land use/land cover (LULC) analysis and machine learning-based wavelet modeling to investigate urban heat variability in Delhi from 2018 to 2023. LULC maps were developed using Landsat-8 and 9 imagery with high classification accuracy (Kappa: 0.86 in 2018, 0.89 in 2023). Between 2018 and 2023, built-up areas expanded from 22.43 to 30.53%, while green spaces declined from 42.47 to 35.36%, intensifying the urban heat island (UHI) effect (mean UHI rose from 0.43 to 1.69). The Continuous Wavelet Transform (CWT) revealed dominant periodicities across 52 scales, capturing both short- and long-term climate cycles. IMD temperature data showed high Lag-1 autocorrelation (0.96), indicating persistent thermal patterns. Seasonal decomposition indicated rising trends, with peak temperatures of 35.5 °C in 2023 and projected cooling to 14.48 °C by 2026. The Prophet model demonstrated high forecast accuracy (MAE = 1.94, R² = 0.88), identifying anomalies in 2025 and 2029. These results underscore the need for real-time environmental monitoring and green infrastructure planning. Future work should integrate high-resolution remote sensing and hydrological modeling to refine climate adaptation strategies in urban regions.

Graphical Abstract

This study uniquely integrates wavelet-based time-frequency analysis with machine learning forecasting to capture multi-scale urban temperature dynamics, offering enhanced precision in predicting thermal trends and climate anomalies. The assessment of urban thermal climate was done using long-term IMD temperature data. It proceeds with trend analysis to identify persistent thermal patterns and inter-annual variability. The process then applies time-frequency analysis using the Continuous Wavelet Transform to detect dominant periodicities. Machine learning-based forecasting, including Prophet modeling, predicts future temperature trends. This integrated approach highlights the significance of multi-scale analysis and predictive modeling in understanding urban climate dynamics and supporting resilient urban planning strategies.