Trends in spectrally resolved outgoing longwave radiation from 10 years of satellite measurements

The study addresses how spectrally resolved outgoing longwave radiation (SOLR) has changed over a decade and whether satellite hyperspectral measurements can detect statistically significant trends tied to specific climate drivers and feedbacks. At the Earth–atmosphere system’s top of atmosphere, energy balance is governed by incoming solar radiation, reflected shortwave radiation, and outgoing longwave radiation (OLR). Changes in greenhouse gases and other drivers introduce radiative forcings and feedbacks that alter OLR. Broadband OLR products and techniques (e.g., radiative kernels, partial radiative perturbations) provide valuable insights but depend on accurate knowledge of numerous variables and may mask compensating model biases. Hyperspectral SOLR, the integrand of broadband OLR, offers spectral fingerprints of individual processes, enabling separation of drivers such as CO₂, CH₄, H₂O, and O₃. Building on advances and multi-year stability of hyperspectral sensors (IASI), the authors evaluate 10 years (2008–2017) of IASI clear-sky SOLR to answer: (1) Can statistically significant SOLR trends be detected for channels sensitive to different tropospheric and stratospheric altitudes, and what is their magnitude? (2) Can significant trends be related to known climate processes (e.g., ENSO, PDO, BDC)? (3) Can effects of increasing greenhouse gases be detected in 10 years of IASI measurements?

Broadband OLR studies over recent decades improved understanding of radiative budget changes and feedbacks, leveraging instruments and methods such as radiative kernels and partial radiative perturbations. However, they can obscure compensating model errors. Hyperspectral observations (AIRS, IASI) enable spectrally resolved analyses revealing model biases that do not appear in broadband comparisons. AIRS-based studies identified model biases and, using 10 years of data, reported significant cooling trends in CO₂ v₂-band channels sensitive to the lower/mid-stratosphere, highlighting challenges in modeling and assimilating stratospheric climate. IASI offers gapless coverage over 645–2760 cm⁻¹ with strong long-term stability and has been used for inter-calibration and detecting multi-decadal spectral changes consistent with increased well-mixed greenhouse gases. Earlier IASI work demonstrated small interannual variability sufficient to detect robust changes linked to feedback processes, and cross-mission comparisons (IRIS, IMG, IASI) revealed clear signatures from rising greenhouse gases.

Data: The study uses a complete reprocessed dataset of IASI/Metop-A clear-sky radiance measurements (latest L1C) converted to spectrally resolved OLR (SOLR) over 645–2300 cm⁻¹ at native 0.25 cm⁻¹ sampling for 2008–2017. The upper bound avoids solar reflectance contamination. Conversion to fluxes follows Whitburn et al. (2020). Precision is generally within ±0.005 W m⁻² (cm⁻¹)⁻¹ depending on wavenumber. Clear-sky identification uses AVHRR cloud flags; only scenes with 0% cloud coverage are retained (~14% of observations). Data consist of daily global SOLR on a 2° × 2° grid, separated by day/night; this analysis uses daytime and is restricted to over-sea observations to reduce land-driven heterogeneity. Zonal analyses exclude latitudes poleward of 75° owing to cloud filter performance and sampling issues. Trends estimation: For each channel (0.25 cm⁻¹ sampling; 6621 channels), daily SOLR are averaged zonally by 2° latitude bands. A low-order Fourier series (n = 3) captures intra-annual variability, plus a linear term for the annual trend (LT), following Gardiner et al. (2008). Confidence limits for trends use bootstrap resampling; significance is defined when the 95% CI excludes zero. Global LT maps are also computed on a 2° × 2° grid for selected channels sensitive to different altitudes, and temperature trends are derived from ERA5 reanalysis (24-hourly data averaged per day) at multiple pressure levels. Synthetic IASI spectra and layer-wise radiance change diagnostics (Atmosphit LBLRTM) for tropical and subarctic standard atmospheres help interpret altitude sensitivity and layer contributions.

• Greenhouse gas fingerprints: Significant negative SOLR trends attributable to increasing CO₂ and CH₄ are observed in the ν₂ and ν₃ CO₂ bands and the ν₄ CH₄ band, with magnitudes −0.05 to −0.3% per year. These reflect increased absorption (reduced outgoing radiance) by rising GHG concentrations.
• Window regions (surface-sensitive): In atmospheric window regions (795–970, 1070–1230, 2090–2170 cm⁻¹), trends mainly follow surface temperature changes. Tropics and mid-latitudes show positive trends about +0.03 to +0.05% per year (≈ +0.5 to +1.3 × 10⁻⁴ W m⁻² per year), consistent with SST evolution tied to two La Niña events (2007–2008, 2010–2011) and a strong El Niño (2015–2016) and associated PDO phases. High northern latitudes (>60°N) show negative zonal trends (−0.1 to −0.4% per year; ≈ −5 × 10⁻⁴ to −2.5 × 10⁻⁴ W m⁻² per year), influenced by the North Atlantic Warming Hole (NAWH) and regional SST patterns; however, uncertainties are large due to sampling and cloud screening.
• Mid-tropospheric H₂O-sensitive channels: Strong regional patterns are linked to dynamical changes (convection/subsidence) modulated by ENSO/PDO. Convergence zones (enhanced humidity/absorption) exhibit negative SOLR trends; subsidence/drying regions (e.g., tropical Western Pacific) show positive trends.
• Upper-tropospheric/strong H₂O channels (~150–250 hPa, e.g., 1507 cm⁻¹): Predominantly negative trends across tropics and mid-latitudes, consistent with lower stratospheric cooling (BDC strengthening and CO₂ increase) and increased upper-tropospheric H₂O via Clausius–Clapeyron, with a few regional exceptions.
• O₃ ν₃ band (~1042–1060 cm⁻¹): Zonal trends are strongly positive in the tropics and subtropics (+0.05 to +0.12% per year; ≈ +0.8 to +1.5 × 10⁻⁴ W m⁻² per year). Spatial heterogeneity indicates ENSO-related modulation of lower-stratospheric O₃ via BDC: enhanced tropical upwelling during El Niño reduces O₃ (yielding positive SOLR trends), whereas mid-latitude downward motions increase O₃ (negative trends).
• High latitudes: In centers of strong absorption bands (CO₂ ν₂, ν₃; CH₄ ν₄; H₂O ν₂ strong lines), trends are positive (0.2–0.4% per year) reflecting stratospheric warming linked to a strengthening BDC; in CO₂ band wings, trends become more negative due to increased transparency to surface emission and the direct CO₂ increase.
• Additional species: Decreasing spectral features consistent with declining CFC-11 (~847 cm⁻¹) and CFC-12 (~923 and ~1161 cm⁻¹) are detected, consistent with Montreal Protocol impacts. Changes in N₂O (ν₁ near ~2224 cm⁻¹) are not clearly detectable, likely masked by overlapping CO₂/H₂O variability.
• Methodological insight: Zonal means can obscure strong regional disparities; global maps reveal teleconnection patterns (ENSO, PDO, Antarctic Dipole) and ocean-basin specific trends.

Findings demonstrate that high-spectral-resolution SOLR from IASI can detect statistically significant trends associated with both thermodynamic (temperature/SST) and compositional (GHG, O₃, H₂O) changes over a decade. The observed negative trends in CO₂ and CH₄ bands directly evidence enhanced greenhouse gas absorption. Temperature-driven patterns, especially in window regions and H₂O-sensitive channels, align with ENSO and PDO variability and associated circulation changes (convection/subsidence). Upper-tropospheric and lower-stratospheric behaviors are consistent with a strengthening Brewer–Dobson circulation and CO₂-driven stratospheric cooling, while O₃ trends reflect ENSO-modulated upwelling and regional circulation impacts. The spectral resolution enables attribution to individual gases and identification of compensating processes that would be muted in broadband OLR, addressing the study’s research questions and underscoring the utility of SOLR for climate process evaluation and model diagnosis.

The study provides a proof of concept that 10 years of IASI-derived spectrally resolved OLR can reveal small but significant trends linked to specific climate drivers. Clear spectral signatures of rising CO₂ and CH₄ are detected (−0.05 to −0.3% per year in corresponding bands), along with ENSO/PDO-driven temperature and humidity variations and O₃ changes tied to stratospheric dynamics. The unprecedented spectral sampling (0.25 cm⁻¹) highlights processes and compensating effects invisible in broadband fluxes, offering a powerful tool for evaluating climate models. With additional IASI platforms (Metop-B/C) and the forthcoming IASI-NG (0.125 cm⁻¹) extending records by at least 25 years, future work can: (1) lengthen time series to strengthen trend detection and attribution; (2) combine with AIRS/CrIS to assess diurnal cycles and cross-validate trends; (3) perform detailed model–observation spectral comparisons to diagnose and reduce model biases; (4) expand to land surfaces with improved cloud/scene screening.

• Time span is relatively short (2008–2017), limiting detection of small trends and separation of forced signals from internal variability.
• Clear-sky restriction and a conservative cloud flag (0% coverage) yield limited sampling (~14%), especially at high latitudes; zonal trends above ~60–75° have high uncertainties due to low observation counts and cloud filter performance.
• Analysis is restricted to ocean scenes to reduce heterogeneity; however, coastal regions may still be influenced by land processes.
• Heterogeneous distribution of clear-sky observations can bias zonal averages; zonal averaging can mask strong regional disparities.
• Disentangling contributions of temperature vs composition is challenging; overlapping spectral features (e.g., N₂O vs CO₂/H₂O) hinder detection in some bands.
• Upper latitude cutoff at 75° and daytime-only focus may omit relevant variability.