Yang Et Al. - 2013 - The Role of Satellite Remote Sensing in Climate CH
Yang Et Al. - 2013 - The Role of Satellite Remote Sensing in Climate CH
Yang Et Al. - 2013 - The Role of Satellite Remote Sensing in Climate CH
Satellite remote sensing has provided major advances in understanding the climate system and its changes, by quantifying
processes and spatio-temporal states of the atmosphere, land and oceans. In this Review, we highlight some important discov-
eries about the climate system that have not been detected by climate models and conventional observations; for example, the
spatial pattern of sea-level rise and the cooling effects of increased stratospheric aerosols. New insights are made feasible by
the unparalleled global- and fine-scale spatial coverage of satellite observations. Nevertheless, the short duration of observa-
tion series and their uncertainties still pose challenges for capturing the robust long-term trends of many climate variables. We
point out the need for future work and future systems to make better use of remote sensing in climate change studies.
O
bservational data and model simulations are the foundations In this Review we discuss the contribution of SRS to our under-
of our understanding of the climate system1. Satellite remote standing of climate change and the processes involved. We focus
sensing (SRS) — which acquires information about the Earth’s on SRS-enabled discoveries that have substantiated or challenged
surface, subsurface and atmosphere remotely from sensors on board our fundamental knowledge of climate change. Our main goal is
satellites (including geodetic satellites) — is an important component to reveal the unique contributions and major limitations of SRS
of climate system observations. Since the first space observation of as used in these studies. Technical details on instrumentation and
solar irradiance and cloud reflection was made with radiometers retrieval methods can be found in a recent review 12.
onboard the Vanguard-2 satellite in 19592, SRS has gradually become
a leading research method in climate change studies3. Observations of the climate system
The use of satellites allows the observation of states and processes of Conventional land-based observations are typically collected at
the atmosphere, land and ocean at several spatio-temporal scales. For fixed intervals with limited spatial coverage, whereas SRS allows for
instance, it is one of the most efficient approaches for monitoring land continual monitoring on the global scale. This has greatly enhanced
cover and its changes through time over a variety of spatial scales4,5. our understanding of the climate system and its variations (Fig. 1).
Satellite data are frequently used with climate models to simulate
the dynamics of the climate system and to improve climate projec- Global warming. The warming trend of the Earth’s mean surface
tions6. Satellite data also contribute significantly to the improvement temperature since the late nineteenth century has provided evi-
of meteorological reanalysis products that are widely used for climate dence for anthropogenic influences on global climate13. This trend
change research, for example, the National Center for Environmental was first identified by analysing anomalies in time series of near-
Prediction (NCEP) reanalysis7. The Global Climate Observing System surface air temperature over the land that were recorded by weather
(GCOS) has listed 26 out of 50 essential climate variables (ECVs) as stations14. However, the existence of the trend was consistently chal-
significantly dependent on satellite observations8. Data from SRS is lenged due to the biases in weather records15 caused by such things
also widely used for developing prevention, mitigation and adapta- as poor siting of the instrumentation and the influence of land-use/
tion measures to cope with the impact of climate change9. land-cover changes. Satellite data provides an independent way to
Despite the aforementioned contributions of SRS, there are investigate global temperature trends, particularly at the ocean sur-
concerns about the suitability of satellite data for monitoring and face and in the atmosphere.
understanding climate change10. Climate change studies require The sea surface temperatures (SSTs) of the oceans — which
observations to be calibrated/validated and consistent, and to pro- are directly related to heat transfer between the atmosphere and
vide adequate temporal and spatial sampling over a long period of oceans — serve as important indicators of the state of the climate
time11. However, satellite data often contain uncertainties caused by system16. The Advanced Very High Resolution Radiometer on board
biases in sensors and retrieval algorithms, as well as inconsistencies the National Oceanic and Atmospheric Administration (NOAA)
between continuing satellite missions with the same sensors. The satellites allows us to monitor the SST worldwide. An increase in
use of satellite observations in climate change studies requires a SST has been observed in all ocean basins since the 1970s, with
clear identification of such limitations. an average estimated increase of 0.28 °C from 1984 to 200617. This
Ministry of Education Key Laboratory for Earth System Modeling, Center for Earth System Science, Tsinghua University, Beijing 100084, China, 2State Key
1
Lab of Remote Sensing Science, jointly sponsored by the Institute of Remote Sensing and Digital Earth, Chinese Academy of Science, and Beijing Normal
University, Beijing 100101, China, 3Department of Environmental Science, Policy and Management, University of California, Berkeley, California 94720,
USA, 4Jackson School of Geosciences, University of Texas, Austin, Texas 78712, USA, 5School of Marine and Atmospheric Sciences, State University of
New York, Stony Brook, New York 11794, USA, 6International Institute for Earth System Science, Nanjing University, Nanjing 210093, China, 7Department
of Geography and City Planning, University of Toronto, Toronto, Ontario M5S 3G3, Canada, 8College of Global Change and Earth System Science,
Beijing Normal University, Beijing 100875, China, 9Department of Geography, University of Maryland, College Park, Maryland 20742, USA, 10College of
Environmental Science and Engineering, Tsinghua University, Beijing 100084, China. *email: [email protected]
Space
Atmosphere
Aerosols
Cryosphere
Biosphere
Sea ice
Hydrosphere
Lithosphere
Figure 1 | Remote sensing of the climate system. Remote sensing is carried out by sensors aboard different platforms, including plane, boat and Argo
floats. Ground-based instruments are also used, for example, sun spectral radiometers measure solar radiation. However, satellite remote sensing is
capable of providing more frequent and repetitive coverage over a large area than other observation means . Figure courtesy of R. He, Hainan University.
global trend is similar to that of the near-surface air temperature observed temperatures in the tropical upper- and lower-middle
over land, thus providing a decisive corroboration of the warm- troposphere are also significantly smaller than those simulated by
ing identified from weather records, although the magnitude still climate models28. These remaining differences indicate that either
has considerable uncertainty 18. Satellite observations also reveal model deficiencies or observational errors that were identified in
an uneven warming pattern. The ocean warming trend is highest previous studies have not been fully resolved.
in the middle latitudes of both hemispheres19, stronger east–west
zonal SST gradients have been detected across the equatorial Pacific Snow and ice. The retreat of snow- and ice-cover is an important
Ocean20. The regional variability of SST is closely linked with that of indicator of global warming. Melting of seasonal snow- and ice-
other climate parameters, such as precipitation21. All of these dis- cover can cause a positive feedback by lowering the albedo of the
coveries have led to a better understanding of the role of the oceans Earth’s surface, and the latter contributes to sea-level rise (SLR)13.
in influencing global climate. Data from SRS has played a crucial role in monitoring the dynamics
Surface and tropospheric warming has been predicted as a of snow extent and ice covers.
major response of the climate system to escalating concentrations Snow-cover extent (SCE) over the Northern Hemisphere has
of greenhouse gases22, but initial analyses found no obvious trend been routinely monitored since 1967 using visible-band sen-
in the tropospheric temperature records obtained using microwave sors and passive microwave-band sensors carried by satellites29.
sounding units (MSUs) on polar-orbiting satellites23 — a find- Reconstructed time series based on the NOAA SCE data set and
ing that has challenged both the reliability of surface temperature in situ measurements have shown that, overall, the SCE over the
records and our understanding of the reaction of the climate sys- Northern Hemisphere has been reduced by 0.8 million km2 per
tem to increased greenhouse gas concentrations24. However, after decade in March and April from 1970–2010. A comparison of
removing known problems with the sensors, accounting for the recent SCE with pre-1970 values indicates a 7% and 11% decrease
influence of stratospheric cooling and reducing biases in retrieval in March and April, respectively 30. This trend is consistent with
methods25, new versions of satellite time series all show a warming observed global surface warming. The satellite observations also
trend in the troposphere, except in the Antarctic region26 (Fig. 2). display strong regional patterns of SCE that are affected by regional
This progress, together with improved climate models, has led to climate variability. No significant decrease was observed over North
the conclusion that there is no fundamental disagreement between America in March from 1970–201030, and the decrease of SCE
observed and modelled tropospheric temperature trends of warm- inside Russia ceased after 199031.
ing (provided that uncertainties in both are accounted for)24. The extent of sea ice is primarily monitored by passive micro-
Nevertheless, some questions remain. The observed scaling ratio wave sensors such as the special sensor microwave/imagers (SSM/I).
(that is, the ratio of atmospheric trend to surface trend) is still sig- The latest pattern identified from the satellite records shows that the
nificantly lower than model projections27. The differences between Antarctic sea-ice extent has increased by 1.5±0.4% per decade from
0.5
0.0
–0.5
1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 2012
Year
RSS
0.5 Mid-troposphere
Temperature anomaly (K)
UAH
0.0
1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009 2012
Year
RSS
0.5 Lower troposphere
Temperature anomaly (K)
UAH
0.0
Figure 2 | Upper atmosphere temperature trends between 1979 and 2012 based on MSU data sets. Data from the University of Alabama at Huntsville
(UAH) and the Remote Sensing Systems (RSS) are shown here. A 6-month moving window smoothing has been applied to the data sets. Both data sets
now show a warming trend, whereas UAH formerly reported a cooling trend in the lower troposphere23. Data from the National Climate Data Center;
http://vlb.ncdc.noaa.gov/temp-and-precip/msu. Figure courtesy of C. Huang, Beijing Forestry University.
1979 to 201032. Researchers hypothesize that this slight increase the mean trend, the acceleration of ice discharge in the Antarctic
was caused by reduced upward heat-transport in the oceans and and Greenland was revealed through the satellite-borne interfero-
increased snowfall33. The situation is reversed in the Arctic, where metric synthetic-aperture radar records39. This finding and subse-
satellite observations show an overall negative trend of −4.1±0.3% quent studies indicate that ice–ocean interaction drives much of
per decade in that period, with the largest negative trend occurring the recent increase in mass loss from the Antarctic and Greenland
in summer 34. This concurs with climate model projections, but its ice-sheets40.
magnitude is larger 35. Whether the Arctic as a whole has reached a The retreat of global mountain glaciers and ice caps (GICs) has
‘tipping point’ (a point in time when the decrease of ice in the Arctic been cited as a clear sign of global warming. However, satellite
cannot be reversed) is still a matter of debate, but some research- observations show that the extent and magnitude of melt for some
ers believe that satellite observations show that a ‘regional tipping glacier systems have been less than predicted. Based on Advanced
point’ has been reached, given that the average age of multi-year ice Spaceborne Thermal Emission and Reflection Radiometer (ASTER)
is decreasing, and the percentage of oldest ice declined from 50% of and Satellite Pour l’ Observation de la Terre (SPOT) data, the
the multi-year ice pack to 10% from 1980 to 201136. dynamics of glaciers in the Himalayan region were highly variable
The mass losses of the Antarctic and Greenland ice-sheets have from 2000–2008. Although 65% of the monsoon-influenced glaciers
been estimated by measuring surface elevation changes from satel- were retreating, 50% of the glaciers in the Karakoram region of the
lite altimetry data (collected by satellites such as ERS-1/-2, Envisat, northwestern Himalayas were advancing or stable41. The mass loss
ICESat and CryoSat-2) or by measuring ice-mass changes using the of global GICs that was assessed using GRACE measurements dur-
Gravity Recovery and Climate Experiment (GRACE) satellite data. ing 2003–2010 (excluding the Greenland and Antarctic peripheral
Both estimates show that the Antarctic and Greenland ice-sheets GICs) was 30% less than the in situ mass-balance result. In the high
are losing mass37,38. The latest studies find that these polar ice-sheets mountains of Asia, the glacier loss-rate in terms of sea-level con-
contributed an average of 0.59±0.2 mm yr–1 to the rate of global SLR tribution was estimated to be only 0.01±0.06 mm yr–1, significantly
since 199238. However, estimates of the mass-loss rates are highly lower than the 0.13–0.15 mm yr–1 calculated by other studies42. In
divergent (Supplementary Table 1). Both GRACE-based and altim- the Karakoram region, the mass balance of glaciers was found to
eter-based methods have limitations, and the efforts to combine the be positive, at 0.11±0.22 m yr–1 of water equivalent between 1999
strengths of the two have achieved only partial success37. Along with and 200843. Although some of these evaluations still contain high
using longer satellite time series did produce smaller rates of increase, sensors carried by some satellites cannot capture climate processes
these results are still regarded as inconclusive due to the brevity of occurring at finer spatial scales. For example, satellite sensors are
the series83,84. The correlation between satellite observed precipita- unable to provide the observations at high enough resolutions to
tion and global surface temperature anomalies at inter-annual scale characterize the small-scale ‘turbulence’ that is associated with the
was also found to be weak during 1988–2008 if large-scale forcing variability of atmospheric temperature and water vapour 91. In addi-
including ENSO and volcanic eruptions were removed76. A survey tion, satellite sensors do not have the required accuracy for detecting
of available satellite-based long-term precipitation products showed cloud-property trends72. This requires new sensors with sufficiently
mostly little to no trend in global precipitation85. These divergent high spatial resolution and accuracy for observing the climate phe-
findings illustrate the problems of detecting a robust global mean nomena of interest. Existing systems provide inadequate temporal
trend of precipitation, a consequence of high variability of precipita- coverage with which to study quickly evolving climate processes.
tion, systematic biases associated with instruments, and inadequate This shortfall must be addressed. A possible solution is a constella-
interpretation of the surface and atmospheric properties in the tion of small satellites that observe the same location over a given
retrieval algorithms84,86. Although there is still uncertainty regard- time interval9,72.
ing a general trend, satellite observations have greatly enhanced our The findings of satellite records that are used for trend detection
understanding of the climate processes that control the variability of and retrieval of the absolute levels of climate variables are greatly
precipitation. For example, the effect of SST on the regional precipi- affected by how well the uncertainties associated with the sensors
tation pattern at multi-annual and multidecadal timescales21, and the are resolved. This important step is underscored in the debate on the
prevalence of ‘wet-gets-wetter’ and ‘dry-gets-drier’ trends in tropical trend in solar radiation. Undetected drifts in sensor sensitivity have
regions that are caused by the intensification of summer monsoons, been cited as the main reason for the apparent spectrum of change57.
were detected using satellite precipitation products87. Satellite sensors gradually lose radiometric sensitivity and stability
during their operation, so good calibration is essential. Some satel-
Limitations and solutions lite sensors cannot be recalibrated after launch due to the lack of
Three recurring limitations can be identified from these studies: accurate on-board or on-orbit calibrations. Procedures have been
short data spans of satellite records, biases associated with instru- developed to calibrate these types of sensors but may still contain
ments and uncertainties in retrieval algorithms. uncertainties92. Biases caused by instrument drifts are also common
in satellite data. Satellites go through a slow change of crossing time
Short spans of satellite data sets. Researchers who use short time at the local equator and a decay of orbital height, adding a spurious
series of satellite data might have trouble reliably separating the long- effect to detected trends93. Such biases must be addressed by apply-
term trends from inter-annual and decadal variability. Intuitively, to ing a diurnal correction procedure to the data93 or by determining
most accurately detect trends in climate change, satellite observations the precise orbit position of the satellites94.
should have long-term continuity 88, consistency, and homogeneity Uncertainties can also increase when combining observations
(here this means that the non-climatic factors such as instruments from different satellite systems to form long-term records. If the
and observing methods have homogeneous influences on all observed procedures for merging data from different systems are not well
data). However, there is no uniform requirement for observation developed and calibrated, the uncertainties can potentially be high
length. Some studies have specified minimum requirements for cli- in combined data sets. The report of an abrupt increase in Antarctic
mate variables, ranging from 40 years for tracking the change of ocean sea-ice area was found to be a false detection caused by the shift
colour 89 to 60 years for determining SLR90. The GCOS suggested from the SSM/I system to the Advanced Microwave Scanning
30 years for satellite observation of the climate system8. More studies Radiometer for the Earth Observing System (AMSRE)95. This type
are needed to determine the lengths of satellite time series required of problem can be reduced by allowing an overlap of operating time
for detecting reliable trends in other climate variables. An examina- between instruments so inter-instrument calibrations can be con-
tion of the lengths of ECVs constructed from satellite observations ducted to find out the relative bias96. For instance, launch times of
shows that some time series are already longer than 30 years (Table 1). the Topex/Poseidon satellite and its successors allow for tandem
The availability of more time series with adequate length will depend measurements and intersatellite calibrations. Many new satellite
on our ability to maintain the continuity of existing satellite missions. missions are adopting this approach12.
Biases associated with instruments. Observations of many cli- Uncertainties in retrieval algorithms. Retrieval algorithms are
mate variables are made by satellite sensors that were originally key for converting the electromagnetic signals from satellite sensors
designed for meteorological observations. The coarse-resolution to measurements of climate variables. Uncertainties can therefore
In the version of this Review Article originally published, on page 877, the mass balance of glaciers in the Karakarom region should have
been 0.11±0.2 m yr−1. On page 879, in the second paragraph of the ‘Aerosols’ section, the estimated value for the direct radiative forcing
should have been −1.0±0.34 W m−2. These errors have now been corrected in the HTML and PDF versions of the Review Article.
Nature Clim. Change 3, 875–883 (2013); published online 15 September 2013; corrected after print 27 September 2013; 2 December 2013.
In the version of this Review Article originally published, the temperature anomaly trends for RSS and UAH in Fig. 2b,c should have
been positive values. These errors have now been corrected in the online versions of the Review Article. In the previous corrigendum,
the mass balance of glaciers in the Karakarom region should have read 0.11±0.22 m yr−1.