Machine learning-based temperature mapping in Sydney: Accuracy, transferability, and scalability
6 affiliations
- 1Climate Change Research Centre, University of New South Wales, Sydney, Australia
- 2Australian Research Council Centre of Excellence for Climate Extremes, University of New South Wales, Sydney, Australia
- 3Australian Research Council Centre of Excellence for the 21st Century Weather, University of Tasmania, Hobart, Australia
- 4Australian Research Council Centre of Excellence for the 21st Century Weather, University of New South Wales, Sydney, Australia
- 5Department of Geography, Ruhr-University Bochum, Bochum, Germany
- 6School of Built Environment, University of New South Wales, Sydney, Australia
Abstract
Urban heat is a growing concern for public health, energy demand, and urban liveability. High-resolution air temperature (Ta) data are needed for developing effective, localised adaptation strategies, but obtaining such data at the city scale remains a challenge as weather stations are often sparse and unevenly distributed. To address this data gap, we developed a machine learning (ML) framework that creates high-resolution, gridded Ta maps using crowdsourced observations and diverse geospatial datasets describing the city, and evaluated its accuracy, transferability, and performance under varying data availability conditions. This framework uses satellite-derived land surface temperature (LST), urban form and fabric datasets, and meteorological variables to train a convolutional neural network (CNN) model. The approach was implemented in Sydney, Australia, producing 30 m gridded Ta estimates (at ~10:00 a.m. local time) with high accuracy on an independent held-out test set (R² = 0.97; RMSE = 0.91 °C), surpassing previously reported ML performance. We further assessed the generalisation and spatial transferability of this method, showing that the CNN model maintained strong predictive accuracy for unseen locations across the city (R² = 0.91–0.93; RMSE = 1.27–1.44 °C). Model performance also remained stable when the number of Ta stations was significantly reduced by ~80%. Performance was more variable for unseen days (R² = 0.66–0.93; RMSE = 1.52–2.60 °C), indicating that ML-based Ta mapping requires training data representative of diverse weather conditions. We find that high-resolution city-descriptive datasets are beneficial but not essential, as comparable accuracy was achieved using only globally available predictors. These results indicate that the proposed framework is transferable to other cities, including those in data-sparse regions. The study provides an effective and scalable approach for city-scale air temperature mapping, which is urgently needed for urban heat assessment, climate adaptation, and public health planning.
Related
- A machine learning model for early season wildfire risk in Ghana’s Guinea savannah assessing discrimination, operational prevalence, spatial transferability, and CMIP6 climate scenarios
- Machine learning surrogates for flash flood hazard mapping: CSI-optimized feature selection and spatial transferability
- Machine-learning prediction of freezing temperature in Pb-contaminated soils remediated with biochar
- The Equilibrium Response of Atmospheric Machine‐Learning Models to Uniform Sea Surface Temperature Warming
- Data Quality Audit and Bivariate Temperature-Humidity Analysis of a Long-Term LoRaWAN Network: Anomaly Detection, PCA, Clustering and Machine Learning
BibTeX
@article{naserikia2026machine,
title = {{Machine learning-based temperature mapping in Sydney: Accuracy, transferability, and scalability}},
author = {Marzie Naserikia and Melissa Anne Hart and Elyas Asadi Shamsabadi and Katrin Meissner and Sanaa Hobeichi and Benjamin Bechtel and Negin Nazarian},
journal = {PLOS Climate},
year = {2026},
month = sep,
doi = {10.1371/journal.pclm.0001050},
url = {https://journals.plos.org/climate/article?id=10.1371/journal.pclm.0001050},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Machine learning-based temperature mapping in Sydney: Accuracy, transferability, and scalability
AU - Marzie Naserikia
AU - Melissa Anne Hart
AU - Elyas Asadi Shamsabadi
AU - Katrin Meissner
AU - Sanaa Hobeichi
AU - Benjamin Bechtel
AU - Negin Nazarian
PY - 2026
DA - 2026/09/09
JO - PLOS Climate
DO - 10.1371/journal.pclm.0001050
UR - https://journals.plos.org/climate/article?id=10.1371/journal.pclm.0001050
AB - Urban heat is a growing concern for public health, energy demand, and urban liveability. High-resolution air temperature (Ta) data are needed for developing effective, localised adaptation strategies, but obtaining such data at the city scale remains a challenge as weather stations are often sparse and unevenly distributed. To address this data gap, we developed a machine learning (ML) framework that creates high-resolution, gridded Ta maps using crowdsourced observations and diverse geospatial datasets describing the city, and evaluated its accuracy, transferability, and performance under varying data availability conditions. This framework uses satellite-derived land surface temperature (LST), urban form and fabric datasets, and meteorological variables to train a convolutional neural network (CNN) model. The approach was implemented in Sydney, Australia, producing 30 m gridded Ta estimates (at ~10:00 a.m. local time) with high accuracy on an independent held-out test set (R² = 0.97; RMSE = 0.91 °C), surpassing previously reported ML performance. We further assessed the generalisation and spatial transferability of this method, showing that the CNN model maintained strong predictive accuracy for unseen locations across the city (R² = 0.91–0.93; RMSE = 1.27–1.44 °C). Model performance also remained stable when the number of Ta stations was significantly reduced by ~80%. Performance was more variable for unseen days (R² = 0.66–0.93; RMSE = 1.52–2.60 °C), indicating that ML-based Ta mapping requires training data representative of diverse weather conditions. We find that high-resolution city-descriptive datasets are beneficial but not essential, as comparable accuracy was achieved using only globally available predictors. These results indicate that the proposed framework is transferable to other cities, including those in data-sparse regions. The study provides an effective and scalable approach for city-scale air temperature mapping, which is urgently needed for urban heat assessment, climate adaptation, and public health planning.
SN - 2767-3200
ER -