Urban Heat Island Contamination in USHCN Temperature Records¶

A comprehensive analysis revealing that NOAA temperature adjustments enhance rather than remove urban heat island signals

Key Finding

NOAA adjustments enhance urban heat island signals by 9.4% rather than removing them, with 22.7% of USHCN stations experiencing urban heat island contamination averaging 0.725°C.

Executive Summary¶

This study investigates whether NOAA temperature adjustments successfully remove urban heat island (UHI) contamination from the US temperature record. Analysis of 126 years of data from 1,218 weather stations reveals that adjustments actually enhance urban warming signals rather than removing them.

Primary Results¶

Dataset UHII (°C) Change from Raw p-value Effect Size
Raw 0.662 — 0.004 d = 0.58
TOBs Adjusted 0.522 -21.1% 0.022 d = 0.46
Fully Adjusted 0.725 +9.4% <0.001 d = 0.97

Dual Contamination Pattern¶

The analysis reveals two distinct forms of urban heat island contamination:

  1. Baseline Temperature Elevation: 2.98°C persistent difference between urban and rural stations
  2. Differential Warming Trend: 0.725°C additional warming in urban areas over 126 years
  3. Total Effect: ~3.7°C combined contamination affecting 22.7% of stations

Research Context¶

The Challenge¶

The US Historical Climatology Network (USHCN) provides critical temperature data for climate assessments. NOAA applies adjustments to remove non-climatic biases including:

  • Time-of-observation changes
  • Station relocations
  • Equipment changes
  • Urban heat island effects

The effectiveness of these adjustments in removing urban warming signals has significant implications for climate science.

Urban Heat Islands¶

Urban areas create localized warming through:

  • Reduced evapotranspiration from vegetation loss
  • Increased heat absorption by built surfaces
  • Anthropogenic heat release
  • Modified atmospheric circulation

Given that many weather stations are in or near growing population centers, UHI contamination poses a significant challenge to temperature record integrity.

Methodology¶

Analysis Framework¶

We employed a "steel-man" approach, using parameters that maximize urban heat island detection:

  • Time Period: 1895-2020 (126 years)
  • Baseline: 1895-1924
  • Current: 1991-2020
  • Temperature Metric: Minimum temperatures (strongest UHI signal)
  • Network: 1,218 USHCN stations

Urban Classification¶

Stations classified by distance to population centers:

  1. Urban Core: <25km from cities ≥250,000 (26 stations, 2.1%)
  2. Urban Fringe: 25-50km from cities ≥100,000 (120 stations, 9.9%)
  3. Suburban: 50-100km from cities ≥50,000 (405 stations, 33.3%)
  4. Rural: >100km from cities ≥50,000 (667 stations, 54.8%)

Calculations¶

Temperature Anomaly = Mean(Current Period) - Mean(Baseline Period)
UHII = Mean(Urban Anomalies) - Mean(Rural Anomalies)

Statistical Methods¶

  • Independent samples t-test
  • Mann-Whitney U test
  • Cohen's d effect size
  • 95% confidence intervals via bootstrap

Key Findings¶

The Adjustment Paradox¶

NOAA's adjustment process creates an unexpected U-shaped pattern:

graph LR A[Raw Data<br/>0.662°C] -->|TOBs Adjustment<br/>-21.1%| B[TOBs Adjusted<br/>0.522°C] B -->|Homogenization<br/>+38.8%| C[Fully Adjusted<br/>0.725°C] style A fill:#f9f,stroke:#333,stroke-width:2px style B fill:#bbf,stroke:#333,stroke-width:2px style C fill:#f96,stroke:#333,stroke-width:4px

This suggests that while time-of-observation corrections reduce urban signals, subsequent homogenization procedures more than reverse this reduction.

Absolute Temperature Differences¶

Beyond differential warming trends, urban stations show persistent baseline elevation:

  • Minimum Temperature UHI: 2.98°C (year-round)
  • Maximum Temperature UHI: 0.59°C (summer)

The 5× difference between nighttime and daytime UHI aligns with established urban thermal dynamics.

Network Quality Considerations¶

Analysis revealed critical network coverage issues:

  • Pre-1895: Only 64 average stations (inadequate)
  • 1890-1908: 5× expansion potentially creating artifacts
  • Post-1908: Stable 1,218 stations (adequate coverage)

We constrained analysis to post-1895 to ensure reliable results.

Implications¶

For the United States¶

With only 2.1% of stations in urban cores and 36 people/km², the US represents a best-case scenario. Yet we still find:

  • 22.7% of stations affected by UHI
  • 0.725°C warming trend contamination
  • 2.98°C baseline contamination
  • Enhancement rather than removal by adjustments

Global Extrapolation¶

More densely populated regions face greater challenges:

Country Population Density Relative to US Expected UHI Impact
USA 36 people/km² 1.0× 22.7% contamination
UK 275 people/km² 7.6× Much higher
Germany 240 people/km² 6.7× Much higher
Japan 347 people/km² 9.6× Much higher
Netherlands 508 people/km² 14.1× Severe

Climate Science Impact¶

  1. Temperature Records: Both baseline and trend contamination persist
  2. Model Validation: Observational data contains systematic warm biases
  3. Policy Targets: Temperature goals may need recalibration
  4. Attribution Studies: Urban warming conflated with climate signals

Reproducing Results¶

Main Finding¶

# Reproduce Table 2 from paper
ushcn-heatisland analyze adjustment_impact \
  --temp-metric min \
  --baseline-start-year 1895 \
  --current-start-year 1991 \
  --period-length 30

Expected output:

  • Raw UHII: 0.662°C
  • TOBs UHII: 0.522°C (-21.1%)
  • Fully Adjusted UHII: 0.725°C (+9.4%)

Enhanced Analysis¶

# Network quality-informed analysis
cd analysis/ushcn_uhii_analysis_1895_plus
python create_min_temp_uhii_plot_1895.py

Results: 2.975°C minimum temperature UHI

Visualizations¶

# Generate Figure 1
cd analysis/network_visualisation
python create_network_visualisation.py

Conclusions¶

  1. NOAA adjustments enhance rather than remove urban heat island signals
  2. The enhancement (+9.4%) results from homogenization procedures
  3. Total UHI contamination approaches 3.7°C for affected stations
  4. 22.7% of the USHCN network experiences this contamination
  5. Global implications likely more severe in densely populated regions

Future Research¶

Immediate Priorities¶

  • Analyze temperature networks in Europe, East Asia
  • Investigate why homogenization enhances urban signals
  • Compare with satellite temperature records
  • Station-by-station adjustment analysis

Long-term Goals¶

  • Global GHCN analysis
  • Alternative homogenization methods
  • Machine learning approaches
  • Policy recommendation development

References¶

Lyon, R. (2025). Urban Heat Island Contamination Persists in Homogenized USHCN Temperature Records: A 126-Year Analysis. [Journal Name], [Volume], [pages].

Data & Code¶

Scientific Disclaimer¶

This research represents an independent investigation into temperature data processing methods. The findings:

  • Are based on publicly available data and standard statistical methods
  • Do not constitute evidence of deliberate manipulation or conspiracy
  • Highlight unintended consequences of statistical algorithms
  • Contribute to scientific understanding of measurement uncertainties
  • Support improved accuracy in climate assessments

Responsible Use¶

Results should be interpreted within the broader context of climate science research. The identification of urban heat island contamination:

  • Suggests refinement of warming magnitude estimates
  • Supports improved measurement accuracy
  • Benefits all stakeholders in climate science and policy

This research contributes to the continuous improvement of climate science through the identification and correction of systematic measurement biases. Scientific progress depends on the willingness to follow evidence wherever it leads, acknowledge uncertainties when discovered, and refine methods accordingly.