Satellite data records as a tool to monitor changes in air temperature

Technical Summary: Satellite Data Records as a Tool to Monitor Changes in Air Temperature

Problem Statement
The years 2023–2024 recorded the highest global temperatures in the instrumental record, driven by the combination of anthropogenic global warming and a strong El Niño event. While surface temperature records (GMST) show unprecedented warming, it remains unclear whether this represents a temporary spike or a fundamental acceleration in the long-term warming rate. Previous studies have yielded conflicting results regarding the statistical significance of warming rate changes, often due to the high variability inherent in climate data and uncertainties in data homogenization. Furthermore, a persistent discrepancy exists between climate model projections (which predict stronger warming in the troposphere than at the surface) and observational datasets (which have historically shown weaker tropospheric warming). This study addresses the need to determine if lower-tropospheric temperatures (TLT) exhibit a trend increase comparable to GMST and to quantify any acceleration in warming rates using long-term satellite records.

Methodology
The authors analyzed global lower-tropospheric temperature (TLT) data spanning 1981–2025 from four distinct sources:

1. NOAA V5.0: Derived from microwave sounder observations (MSU/AMSU-A/ATMS) with specific bias corrections.
2. UAH V6.1: A satellite-derived dataset using different combination coefficients and bias treatments.
3. RSS V4.0: A dataset utilizing off-nadir observations with distinct homogenization algorithms.
4. ERA5 Reanalysis: An equivalent TLT constructed by applying the NOAA TLT weighting function to ERA5 temperature profiles, providing a physically consistent forcing perspective.

To isolate long-term trends from natural variability, the authors constructed an ENSO-aerosol-adjusted TLT record. This involved removing variability associated with:

• El Niño–Southern Oscillation (ENSO): Using the Niño 3.4 index with a 5-month lag.
• Atmospheric Aerosols: Using MERRA-2 reanalysis aerosol optical depth (AOD) data with an 8-month lag to account for volcanic and anthropogenic sulfate effects.

This adjustment reduced annual TLT variability by approximately 46–56%, significantly enhancing the robustness of trend detection.

Two statistical modeling approaches were applied to the adjusted data (1981–2024):

1. Piecewise Linear Regression: Identifying a "changepoint" where the warming rate shifts. The optimal changepoint was determined by minimizing the Root Mean Square Error (RMSE) and ensuring statistical significance (p < 0.05).
2. Polynomial Fitting: Testing linear through fifth-order polynomials. Model selection was based on the statistical significance of the leading coefficient and RMSE behavior. The cubic polynomial was identified as the optimal model for most datasets, as it captured the data structure without overfitting and closely resembled Locally Weighted Regression (LOESS) smoothing.

Key Results

• Identification of Acceleration: After removing ENSO and aerosol effects, all datasets exhibited a statistically significant increase in the warming rate starting around 2015.
• Magnitude of Trends:
  • Pre-2015 (1981–2015): Average warming rate was 0.121 ± 0.025 °C decade⁻¹.
  • Post-2015 (2015–2024): Average warming rate increased to 0.482 ± 0.113 °C decade⁻¹.
  • This represents a four- to five-fold increase in the warming rate across the NOAA, UAH, and ERA5 datasets.
• Acceleration Rate: The cubic fit revealed a continuous increase in the acceleration rate (second derivative), crossing zero around 2000 (NOAA/UAH) or 1997 (RSS/ERA5). By 2024, the acceleration rate reached approximately 0.48 ± 0.12 °C decade⁻² for the adjusted datasets.
• Role of 2023–2024: The temperature jumps observed in 2023–2024 are largely attributed to this ongoing acceleration, further amplified by the El Niño event. The study estimates that the trend change alone contributed an additional 0.22–0.33 °C of warming over the 2015–2024 period compared to a linear trend.
• Projections: Based on the detected acceleration, the authors project an additional 0.5–1.0 °C of lower-tropospheric warming over the next decade (2025–2034). The piecewise model offers a conservative estimate (0.5 °C), while the cubic model suggests an upper-end scenario (1.0 °C).

Significance and Claims
The paper claims that the observed warming in 2023–2024 is not an isolated anomaly but part of a statistically significant, ongoing acceleration in the global warming rate that began around 2015. By utilizing satellite TLT data and rigorously removing natural variability, the study provides evidence that the tropospheric warming rate has increased substantially, potentially exceeding previous surface-based estimates.

The authors emphasize that while the physical mechanisms driving this specific acceleration remain unclear—ruling out greenhouse gas forcing alone (which has increased linearly) and aerosol effects (which were removed)—the data-driven detection of this trend is robust. The study highlights a critical gap in understanding the drivers of this rapid warming, noting that current climate models have not consistently reproduced these observed atmospheric trends or the magnitude of recent temperature increases. Consequently, the authors argue that resolving these mechanisms is essential for improving future climate projections and informing mitigation strategies. The findings suggest that the characteristic timescale of major climate change may be on the order of a decade, implying fundamental limits to the predictability of climate change based solely on data-driven models without a full understanding of the underlying forcing.