The Spectral Signature of Recent Climate Change
- First Online:
- Cite this article as:
- Brindley, H.E. & Bantges, R.J. Curr Clim Change Rep (2016) 2: 112. doi:10.1007/s40641-016-0039-5
Spectrally resolved measurements of the Earth’s reflected shortwave (RSW) and outgoing longwave radiation (OLR) at the top of the atmosphere intrinsically contain the imprints of a multitude of climate relevant parameters. Here, we review the progress made in directly using such observations to diagnose and attribute change within the Earth system over the past four decades. We show how changes associated with perturbations such as increasing greenhouse gases are expected to be manifested across the spectrum and illustrate the enhanced discriminatory power that spectral resolution provides over broadband radiation measurements. Advances in formal detection and attribution techniques and in the design of climate model evaluation exercises employing spectrally resolved data are highlighted. We illustrate how spectral observations have been used to provide insight into key climate feedback processes and quantify multi-year variability but also indicate potential barriers to further progress. Suggestions for future research priorities in this area are provided.
KeywordsReflected shortwave and outgoing longwave radiation Radiative forcing and feedback signatures Natural variability Detection and attribution of climate change
The balance between net incoming solar radiation and outgoing terrestrial radiation at the top of the Earth’s atmosphere (TOA) fundamentally drives our climate system. Perturbations to this balance, either as a result of natural variability or anthropogenic activity, ultimately result in either a heating or cooling of the Earth system such that measurements of the Earth’s radiation budget (ERB) at the TOA are essential in order to understand how our climate is evolving with time.
Characteristics of broadband instruments highlighted in Fig. 1a
Spatial resolution (at nadir) (km)
Temporal coverage (launch date to end of instrument operation)
Broadband OLR and RSW
SW and IR channels
June 1975–October 1978 (Nimbus-6)
October 1978–October 1987 (Nimbus-7)
Earth Radiation Budget Experiment (ERBE) 
OLR and RSW
October 1984–February 1990 (Scanner)
October 1984–August 1999 (Non-scanner)b
Clouds and the Earth’s Radiant Energy System (CERES) 
OLR and RSW
IR window channel
November 1997–September 1998 (TRMM)c
December 1999–present (Terra)
April 2002–present (Aqua)
October 2011–present (NPP)
Geostationary Earth Radiation Budget (GERB) 
OLR and RSW
August 2002–April 2007 (Meteosat-8)
December 2005–January 2013 (Meteosat-9)
July 2012–present (Meteosat-10)d
OLR and RSW
Visible and IR window channels
January 1994–March 1995 (Meteor)
July 1998–March 1999 (Resurs 01–4)
October 2011–present (Megha-Tropiques)e
Characteristics of spectrally resolved instruments highlighted in Fig. 1b
Spatial resolution (at nadir) ( km)
Temporal coverage (launch date to end of instrument operation)
InfraRed Interferometer Spectrometer (IRIS-D) 
April 1970–January 1971
Spectrometer/Interferometer of the German Democratic Republic (SI-GDR) 
June 1977–September 1977
January 1979–June 1979
Interferometric Monitor for Greenhouse Gases (IMG) 
650–3000 cm−1 (3 contiguous bands)
0.1 to 0.25 cm−1, varying with band
October 1996–June 1997
Scanning Imaging Absorption Spectrometer for Atmospheric Cartography (SCIAMACHY) 
240–2380 nm (3 non-contiguous bands)
30 × 240
March 2002–May 2012
Atmospheric Infrared Sounder (AIRS) 
650–2665 cm−1 (3 non-contiguous bands)
0.4–2.1 cm−1, varying with band
Infrared Atmospheric Sounding Instrument (IASI) 
645–2760 cm−1 (3 contiguous bands)
Cross-track Infrared Sounder (CrIS) 
650–2550 cm−1 (3 non-contiguous bands)
0.625–2.5 cm−1, varying with band and operational mode
Early Work: the 1980s
Concern in the late 1970s regarding the potential impact of increases in atmospheric CO2 on the Earth’s climate saw a number of scientific papers devoted to detecting a clear, CO2 induced signal in variables including surface temperature [32, 33], ice cover  and sea level rise . The difficulty in attributing the response of the chosen variable to a specific cause given the natural variability of, and confounding factors within, the climate system was a common theme in these studies. In their work, Madden and Ramanathan mooted the possibility of using spectrally resolved satellite observations of the ERB to provide this attribution, suggesting that whilst increasing CO2 should reduce the outgoing energy within its absorption bands, the expected associated surface heating should enhance emission within more transparent regions of the spectrum. Should this surface heating trigger feedbacks such as changes in cloud amount, the RSW energy would also change.
A Topic Revisited: 1990s–Early 2000s
Despite their promise, these theoretical studies did not immediately spark huge interest in the use of spectrally resolved OLR to detect and attribute climate change. Indeed, it was not until the mid-1990s that interest in the topic was revived, principally due to the re-evaluation of observations from the InfraRed Interferometer Spectrometer (IRIS-D) by Goody and co-workers at CalTech. IRIS-D, on the Nimbus-4 meteorological satellite, measured the Earth’s outgoing longwave spectrum from April 1970 to January 1971 . Spanning the range 400–1600 cm−1 (6.25–25 μm) with a nominal spectral resolution of 2.8 cm−1, observations from the instrument had previously been used to provide insight into the spatio-temporal distribution of variables such as methane and cirrus cloud [38, 39]. Motivated both by the prospect of long-term hyperspectral measurements from the proposed Atmospheric InfraRed Sounder (AIRS) and the development of mathematical techniques to optimise fingerprints of climate change , Goody used the IRIS-D measurements to estimate the ‘weather noise’ or short-term variability of the Earth’s OLR spectrum. Combining these estimates with simulated ‘climate forcing’ signals due to, for example, a doubling of CO2 or an increase in incoming solar flux, he introduced the concepts of spectral selectivity and sensitivity. The former term relates to the ability to distinguish between forcings, the latter, the time needed for a signal to emerge from background noise .
Whilst this study indicated that in certain regions of the spectrum a distinct change signal could be identified and attributed to a particular forcing with reasonable confidence, the authors acknowledged that the work was designed more to stimulate further discussion than give definitive answers regarding the most promising detection strategy. The CalTech group continued to refine their approach whilst championing the use of well-calibrated, accurate, spectrally resolved OLR measurements to test climate model performance [41, 42, 43]. Working with other experts, they brought these ideas together, to propose a climate observing system capable of providing benchmark observations for rigorously testing climate model output [44, 45]. Their vision was to establish a system that gave global coverage, measuring quantities that had the information content required to enable scientists to reduce the uncertainty which existed (and still exists) in projections of future climate, with a demonstrable absolute accuracy and precision that would mean that the measurements could be trusted in perpetuity [46, 47]. Their proposed solution involved the space-based measurement of spectrally resolved thermal infrared radiation and atmospheric refractivity, the latter employing satellites within the Global Positioning System (GPS) network.
Inspired by this work, UK researchers were also questioning how observations of the OLR spectrum could be applied to the problem of climatic change. Slingo and Webb used output from the Hadley Centre climate model to simulate the expected clear-sky changes in spectral OLR between 1860 and 2040 in order to examine the signature of climatic response to greenhouse gas forcing . One of their key findings was the very small changes seen in regions sensitive to mid and upper tropospheric water vapour. These arose because of the propensity of the model to conserve tropospheric relative humidity with time, a signature of positive water vapour feedback. In conjunction, a series of papers by Harries and co-workers investigated the potential of developing spatio-spectral detection fingerprints based on the availability of multi-decadal records from narrow-band radiometers, utilising the simulation results to test their approach [49, 50, 51]. The papers highlighted the difficulty of constructing consistent, long-term time series from the operational records and the need to better characterise natural variability in order to assess the significance of any model-observation discrepancy.
Recent Developments: 2000–Present
Since early 2000, significant progress has been made in our ability to monitor changes in spectral OLR and use this information to test climate models. The launch of AIRS aboard the Aqua satellite in 2002, and the subsequent longevity of the instrument, has meant that researchers have finally had the opportunity to test the evolution of the OLR spectrum as simulated by climate models against sustained hyperspectral measurements. In concert, rigorous approaches have been developed to identify the spectral fingerprints of change and evaluate the timescales on which these will emerge given confounding factors such as natural variability and instrument calibration uncertainty . These efforts have led to initiatives to develop a new category of ‘climate change’ satellite missions. Principal amongst these is the Climate Absolute Radiance and Refractory Observatory (CLARREO). Selected as a Tier-1 NASA Decadal Survey mission in 2007 , CLARREO objectives include the provision of in-orbit absolutely calibrated spectral radiances, spanning the longwave and shortwave domains, in concert with GPS radio occultation measurements.
Insights from AIRS
Huang and Yung were perhaps the first to use AIRS observations to directly evaluate the variability of the OLR spectrum . They used Empirical Orthogonal Function (EOF) analysis of the AIRS observations to investigate the modes of spectral variability in different climate zones. Although employing only 1 month of data, the authors showed that on this timescale whilst the contrast between cloud top and surface temperature was the dominant factor in driving variability, the patterns and ordering of the modes varied with zone.
Focusing more on factors related to climate sensitivity, Huang and Ramaswamy investigated the spectral variation in the super greenhouse effect (SGE) as manifested in AIRS and GFDL . The SGE phenomenon, essentially a strong anti-correlation between sea surface temperature (SST) and OLR within the tropics, has a strong regional pattern, tending to occur as zones transition between ascent and descent due to the seasonal shift of the Hadley circulation . Analysis of observed and modelled OLR spectra in SGE regions demonstrated the ability to identify compensating errors not only in the mean model spectrum but also in the responses of cloud and the water vapour vertical distribution to a changing SST.
The AIRS studies indicated the power of using spectral observations to directly test climate model performance, but techniques to formally discriminate between different feedbacks and unambiguously detect change using radiance spectra needed development. Recognising this, Leroy et al. derived feedback signals due to temperature and water vapour changes realised by an ensemble of the World Climate Research Programme’s Coupled Model Intercomparison Project phase 3 (CMIP3) climate models under a particular emissions scenario . They showed an optimal detection technique, incorporating uncertainty in the spectral shape of the feedback signals, could be used to distinguish different signals. They also noted the dominant role of inter-annual variability in determining the accuracy with which a particular feedback could be identified given a 20-year record length.
Unlocking Information from the RSW Spectrum
The previous sections highlight the relatively large body of work concerned with discerning climate forcing and feedback processes directly from OLR spectral radiances. Analogous efforts utilising the shortwave spectrum are less well-developed. However, the realisation that the two regions contain complementary information regarding key feedback processes, particularly those involving cloud, coupled with the availability of hyperspectral observations of RSW, has motivated a number of recent studies. Roberts et al. used reflected spectral radiances from the Scanning Imaging Absorption Spectrometer for Atmospheric Cartography (SCIAMACHY) to quantify and attribute the variability in the hyperspectral observations covering the ultraviolet (0.3 μm) through to the near infrared (1.75 μm) . Employing principal component analysis (PCA) techniques similar to those employed in OLR studies, they showed that for all of the cases studied, six components or fewer explain over 95 % of the variance in the SCIAMACHY spectra. Perhaps more interestingly, they were able to relate specific PCs to variations in cloud, water vapour, vegetation and sea ice. An alternative approach to characterise variability was proposed by Jin et al. who derived spectral radiative kernels to explore the sensitivity of the RSW spectrum to perturbations of individual parameters such as water vapour, aerosol optical depth and cloud properties. They found, at low-mid latitudes, interannual variability in the RSW spectrum was generally dominated by variability in cloud amount and optical depth . At higher latitudes, snow and sea ice played a more important role. Whilst these effects dominate at wavelengths below ∼1.3 μm, the impact of water vapour and cloud height variability is manifested within the strong water vapour absorption bands across the RSW spectrum.
Bringing It Together: a Pan-Spectral Approach and the Question of Natural Variability
Clearly, a model needs to accurately capture natural variability to reliably inform on when a climate change signal might be detectable. In the RSW domain, attempts have been made to compare the variability diagnosed from SCIAMACHY with that captured by the CCSM3 OSSE simulations. Roberts et al. used a PCA approach to show that, for selected observation rich months, the two data sets share a large fraction of their spectral variability. This suggests that the OSSE is adequately capturing the spectral variability of the early twenty-first century as observed by SCIAMACHY . Jin et al., focusing on interannual variability over relatively large spatial domains, showed the normalised OSSE and SCIAMACHY reflectance variabilities are both typically less than 1 % across the RSW spectrum for all regions considered. Their work also highlighted a reduction in inter-annual variability as spatial and temporal averaging scale increases, with a reduction of ∼50 % when moving from monthly to annual averages [70•].
Conclusions and Future Directions
The work highlighted in this review article illustrates the progress that has been made in measuring and interpreting the outgoing energy spectrum of the Earth across both the shortwave and longwave domains over the last five decades. A key advance concerns the direct use of spectral observations to quantitatively evaluate climate models. These comparisons have identified inadequacies in climate model representations of processes relating to the response of cloud, water vapour and temperature that could not be diagnosed from broadband, spectrally integrated measurements because of compensating effects across the spectrum. As such, they provide an illustration of why such measurements could prove invaluable in reducing the uncertainty range that is currently associated with projections of future climate change .
Considering our future climate, this article has highlighted that whilst the potential of directly using spectrally resolved measurements to diagnose and attribute changes in climate forcing and feedbacks has been recognised for over 40 years, it is only relatively recently that hyperspectral measurements have been of sufficient quality and duration to begin to quantify the observed variability in the reflected shortwave and outgoing longwave spectrum over annual timescales and longer. Such assessments are crucial, as they provide insight into when specific signatures associated with, for example, expected changes in greenhouse gases, aerosol or cloud might be expected to emerge from this ‘natural’ variability. Moreover, with careful consideration of instrument sampling characteristics, these data can be used to evaluate whether our current generation of current climate and earth system models are able to capture the observed spectral variability. It should be noted that the effort required to transform climate model output such that it can be directly compared with the observed metrics (radiance, reflectance, etc.) is not trivial [27•]. Nonetheless, similar ‘satellite simulators’ have already been developed for active and narrow-band passive instruments  and have been used to investigate the ability of a variety of different climate models to correctly capture the seasonality associated with different cloud regimes  and to diagnose the reasons for specific common climate model biases . We advocate that developing this capability further for hyperspectral observations should be a high priority. Such an effort would be particularly timely given current and planned future instrumentation (e.g. AIRS, IASI, CrIS, IASI-NG ) and could potentially incorporate developments in our ability to perform fast, accurate, high spectral resolution radiative transfer calculations  whilst retaining consistency with assumptions made internally within the relevant models .
Aligned with the need for the development of suitable tools to directly exploit the information contained in pan-spectral observations, there must also be recognition that, whilst the current generations of hyperspectral instruments are providing observations of unprecedented quality, they were not designed specifically for climate change monitoring. Indeed, we would contend that there is currently no hyperspectral instrument in space that possesses the level of in-orbit, SI traceable calibration or sampling characteristics needed to provide a benchmark record of the true climate state. Efforts are ongoing to rectify this situation via, for example, the CLARREO [81••] and Traceable Radiometry Underpinning Solar and Terrestrial Studies (TRUTHS) initiatives . A CLARREO Pathfinder mission is scheduled for launch in 2020 on-board the International Space Station with the goal to demonstrate the target in-orbit calibration accuracy across the RSW domain, and efforts continue to secure funding for the full mission (currently scheduled for launch no earlier than 2023). Implementation of missions with such benchmarking capability is key to understanding discrepancies between contemporaneous measurements from different instruments and, more critically, to being able to make robust claims concerning real spectral changes that have occurred between observing periods when the instrumental record is not continuous in time .
Finally, we would also note what we consider to be a major missing piece in our understanding of the response of the Earth’s energy budget to climate change. Although in the global mean, approximately half of the Earth’s OLR is located at wavelengths greater than ∼15 μm, within the so-called far-infrared (FIR) , the region has never been measured spectrally, in its entirety, from space. The FIR is highly sensitive to mid-upper tropospheric water vapour and to cirrus cloud, both of which critically influence the Earth’s radiative budget and climate sensitivity . Moreover, very recent work has suggested that the FIR may have a more important role than previously recognised in modulating high-latitude climate response and future change [85, 86]. We suggest that future missions capable of providing spectrally resolved measurements across this region would ensure we can have confidence we are correctly characterising the full OLR spectrum across the complete range of Earth scenes and address a long-standing shortcoming in our observational capabilities.
The authors would like to acknowledge Daniel Feldman, Yi Huang and Xianglei Huang for kindly supplying data and modified figures for this manuscript. We also acknowledge Jacqueline Russell and Amy Seales who assisted in the manuscript preparation.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.