Simplifying Universal Thermal Climate Index Calculations with Symbolic Regression: Analyzing Extreme Heatwaves in China

Abstract This study develops and validates an interpretable symbolic regression surrogate for the Universal Thermal Climate Index (UTCI) in China’s temperate (Köppen-Geiger Cfa) region. From ERA5 June–August (JJA) data (2013 and 2021; tested on 2022), the search yielded a compact four-predictor linear expression in which air temperature, dewpoint temperature and clear-sky shortwave radiation (J m −2 , preceding-hour accumulation) raise UTCI, whereas wind speed reduces it. Dewpoint temperature captures the humidity that limits evaporative cooling under extreme heat. The surrogate achieved R 2 = 0.793 (RMSE ≈ 1.96°C), keeping most predictions in the correct heat-stress category. The linear form emerged from the symbolic-regression search rather than being predefined, whereas the same procedure recovered nonlinear expressions in other Köppen-Geiger climate zones, including arid (BWh) and subtropical-highland (Cwb) climates. Applied to daily HadGEM3-GC31-LL outputs, it projects a marked rise in mean JJA UTCI by late century, largest in the north, with variance also rising in the south. Under SSP5-8.5, domain-mean UTCI rises by 9.5°C above 1980–2000, exceeding 10°C in the north. Very-strong (UTCI > 38°C) days increase from near zero to approximately 39 per year and 99th-percentile exceedances from about 0.9 to around 76.5, significant under absolute and percentile thresholds. Being linear, the change (+1.19°C per °C warming) decomposes into temperature (86%), dewpoint (15%), radiation (2%) and wind (−3%), and diagnoses a small model bias (roughly +0.76°C) mainly from underestimated wind speed. These single-model results are a high-end estimate. Combining efficient computation with driver-level interpretability, this intuitive surrogate offers a new tool for climate projection and heat-risk assessment and adaptation planning.

Authors

Institutions

Publication Details

Journal
Journal of Applied Meteorology and Climatology
Published
2026-11-13
DOI
https://doi.org/10.1175/jamc-d-26-0061.1
Primary Topic
Urban Heat Island Mitigation
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Simplifying Universal Thermal Climate Index Calculations with Symbolic Regression: Analyzing Extreme Heatwaves in China

Kieran M. R. Hunt, Weikang Kong
5 links
Journal of Applied Meteorology and Climatology
Urban Heat Island Mitigation
article

Simplifying Universal Thermal Climate Index Calculations with Symbolic Regression: Analyzing Extreme Heatwaves in China

Kieran M. R. Hunt, Weikang Kong
article en

Abstract

Abstract This study develops and validates an interpretable symbolic regression surrogate for the Universal Thermal Climate Index (UTCI) in China’s temperate (Köppen-Geiger Cfa) region. From ERA5 June–August (JJA) data (2013 and 2021; tested on 2022), the search yielded a compact four-predictor linear expression in which air temperature, dewpoint temperature and clear-sky shortwave radiation (J m −2 , preceding-hour accumulation) raise UTCI, whereas wind speed reduces it. Dewpoint temperature captures the humidity that limits evaporative cooling under extreme heat. The surrogate achieved R 2 = 0.793 (RMSE ≈ 1.96°C), keeping most predictions in the correct heat-stress category. The linear form emerged from the symbolic-regression search rather than being predefined, whereas the same procedure recovered nonlinear expressions in other Köppen-Geiger climate zones, including arid (BWh) and subtropical-highland (Cwb) climates. Applied to daily HadGEM3-GC31-LL outputs, it projects a marked rise in mean JJA UTCI by late century, largest in the north, with variance also rising in the south. Under SSP5-8.5, domain-mean UTCI rises by 9.5°C above 1980–2000, exceeding 10°C in the north. Very-strong (UTCI > 38°C) days increase from near zero to approximately 39 per year and 99th-percentile exceedances from about 0.9 to around 76.5, significant under absolute and percentile thresholds. Being linear, the change (+1.19°C per °C warming) decomposes into temperature (86%), dewpoint (15%), radiation (2%) and wind (−3%), and diagnoses a small model bias (roughly +0.76°C) mainly from underestimated wind speed. These single-model results are a high-end estimate. Combining efficient computation with driver-level interpretability, this intuitive surrogate offers a new tool for climate projection and heat-risk assessment and adaptation planning.

Journal of Applied Meteorology and Climatology
University of Reading (GB), King Abdullah University of Science and Technology (SA)
Climate action
Urban Heat Island Mitigation
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon