• Allen, M. R., Mutlow C. T. , Blumberg G. M. C. , Christy J. R. , McNider R. T. , and Llewellyn-Jones D. T. , 1994: Global change detection. Nature, 370 , 182.

    • Search Google Scholar
    • Export Citation
  • Edwards, T., and Coauthors, 1990: The Along Track Scanning Radiometer measurements of sea surface temperature from ERS-1. J. Br. Interplanet. Soc., 43 , 160180.

    • Search Google Scholar
    • Export Citation
  • Fairall, C. W., Bradley E. F. , Godfrey J. S. , Wick G. A. , Edson J. B. , and Young G. S. , 1996a: Cool-skin and warm layer effects on sea surface temperature. J. Geophys. Res., 101 , C1. 12951308.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Haywood, J. M., Allan R. P. , Culverwell I. , Slingo T. , Milton S. , Edwards J. , and Clerbaux N. , 2005: Can desert dust explain the outgoing longwave radiation anomaly over the Sahara during July 2003? J. Geophys. Res., 110 .D05105, doi:10.1029/2004JD005232.

    • Search Google Scholar
    • Export Citation
  • Horrocks, L. A., Candy B. , Nightingale T. , Saunders R. W. , O’Carroll A. G. , and Harris A. R. , 2003: Parameterisations of the skin effect and implications for satellite-based measurement of sea-surface temperature. J. Geophys. Res., 108 .3096, doi:10.1029/2002JC001503.

    • Search Google Scholar
    • Export Citation
  • Kaplan, A., Kushnir Y. , Cane M. A. , and Blumenthal M. B. , 1997: Reduced space optimal analysis for historical data sets: 136 years of Atlantic sea surface temperatures. J. Geophys. Res., 102 , C13. 2783527860.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Lawrence, S. P., Llewellyn-Jones D. T. , and Smith S. J. , 2004: The measurement of climate change using data from the Advanced Very High Resolution and Along Track Scanning Radiometers. J. Geophys. Res., 109 .C08017, doi:10.1029/2003JC002104.

    • Search Google Scholar
    • Export Citation
  • Merchant, C. J., and Harris A. R. , 1999: Toward the elimination of bias in satellite retrievals of sea surface temperature 1. Theory, modelling and interalgorithm comparison. J. Geophys. Res., 104 , C10. 2356523578.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mutlow, C. T., Zavody A. M. , Barton I. J. , and Llewellyn-Jones D. T. , 1994: Sea surface temperature measurements by the along-track scanning radiometer on the ERS-1 satellite: Early results. J. Geophys. Res., 99 , C11. 2257522588.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • O’Carroll, A. G., Saunders R. W. , Watts J. G. , and Horrocks L. A. , 2004: Reprocessing of the ATSR-2 clear-sea consolidated averaged brightness temperature dataset to skin and bulk sea surface temperatures; including a cross-comparison with ATSR-1 SSTs from 1995. Met Office NWP Forecasting Research Tech. Rep. 430, 26 pp. [Available online at http://www.metoffice.gov.uk/research/nwp/publications/papers/technical_reports/2004/FRTR430/FRTR430.pdf.].

  • O’Carroll, A. G., Saunders R. W. , and Watts J. G. , 2005: Comparison of (A)ATSR SST anomalies with other SST datasets. Met Office NWP Tech. Rep. 462, 30 pp.

  • O’Carroll, A. G., Watts J. G. , Horrocks L. A. , Saunders R. W. , and Rayner N. A. , 2006: Validation of the AATSR Meteo product sea surface temperature. J. Atmos. Oceanic Technol., 23 , 711726.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Parker, D. E., Jackson M. , and Horton E. B. , 1995: The GISST2.2 sea surface temperature and sea-ice climatology. Hadley Centre Climate Research Tech. Note CRTN63, 16 pp.

  • Pichel, W. G., 1991: Operational production of multichannel sea surface temperature from NOAA polar satellite AVHRR data. Global Planet. Change, 90 , 173177.

    • Search Google Scholar
    • Export Citation
  • Rayner, N. A., Parker D. E. , Horton E. B. , Folland C. K. , Alexander L. V. , Rowell D. P. , Kent E. C. , and Kaplan A. , 2003: Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108 .4407, doi:10.1029/2002JD002670.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., 1992: Impact of Mount Pinatubo aerosols on satellite-derived sea surface temperatures. J. Climate, 6 , 768774.

  • Sabin, A. L., and Pisias N. G. , 1996: Sea surface temperature changes in the north eastern Pacific Ocean during the past 20, 000 years and their relationship to climate change in north-western North America. Quat. Res., 46 , 4861.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Saunders, R. W., Roberts T. , O’Carroll A. G. , and Horrocks L. A. , 2004: Comparison of ATSR-2/AATSR and TMI sea surface temperatures with the Hadley Centre analyses. Met Office NWP Tech. Rep. 426, 22 pp. [Available online at http://www.metoffice.com/research/nwp/publications/papers/technical_reports/2004/FRTR426/FRTR426.pdf.].

  • Smith, D. L., Delderfield J. , Drummond D. , Edwards T. , Mutlow C. T. , Read P. D. , and Toplis G. M. , 2001: Calibration of the AATSR instrument. Adv. Space Res., 28 , 3139.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wentz, F., and Meissner T. , 2000: Algorithm Theoretical Basis Document (ATBD) version 2: AMSR ocean algorithm. RSS Tech. Proposal 121599A-1, 66 pp. [Available online at http://www.ssmi.com/papers/AMSR_Ocean_Algorithm_Version_2.pdf.].

  • Wentz, F., Ashcroft P. , and Gentemann C. , 2001: Post launch calibration of the TRMM Microwave Imager. IEEE Trans. Geosci. Remote Sens., 39 , 415422.

  • View in gallery
    Fig. 1.

    Global mean (A)ATSR SSTs and HadISST1 from 1995 to 2005.

  • View in gallery
    Fig. 2.

    Global bias-corrected (A)ATSR, HadISST1, and nighttime AVHRR SST anomalies, all coincident.

  • View in gallery
    Fig. 3.

    Global bias-corrected (A)ATSR, HadISST1, and nighttime AVHRR SST anomalies, all coincident except for (A)ATSR.

  • View in gallery
    Fig. 4.

    Global bias-corrected (A)ATSR, HadISST1, nighttime AVHRR, and AMSR-E SST anomalies, all coincident.

  • View in gallery
    Fig. 5.

    Tropical bias-corrected (A)ATSR, HadISST1, nighttime AVHRR, and TMI, all coincident.

  • View in gallery
    Fig. 6.

    Bias-corrected (A)ATSR, HadISST1, and nighttime AVHRR, all coincident, for the (a) Northern Ocean (20°–60°N, 180°–180°) and (b) Southern Ocean (20°–60°S, 180°–180°).

  • View in gallery
    Fig. 7.

    Tropical Pacific (5°S–5°N, 120°–170°W), bias-corrected (A)ATSR, HadISST1, nighttime AVHRR, and TMI, all coincident.

  • View in gallery
    Fig. 8.

    Global SST anomalies for July 2003 for different SST sources: (top left) non-bias-corrected (three-channel night) AATSR, (top right) bias-corrected (three-channel night) AATSR, (middle left) nighttime non-bias-corrected AVHRR, (middle right) HadISST1, (bottom left) TMI, and (bottom right) AMSR-E.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 256 56 8
PDF Downloads 165 34 2

The Measurement of the Sea Surface Temperature by Satellites from 1991 to 2005

Anne G. O’CarrollMet Office, Exeter, United Kingdom

Search for other papers by Anne G. O’Carroll in
Current site
Google Scholar
PubMed
Close
,
Roger W. SaundersMet Office, Exeter, United Kingdom

Search for other papers by Roger W. Saunders in
Current site
Google Scholar
PubMed
Close
, and
James G. WattsMet Office, Exeter, United Kingdom

Search for other papers by James G. Watts in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

A near-continuous series of global retrievals of sea surface temperature (SST) has been made from the Along-Track Scanning Radiometer (ATSR) series of instruments from 1991 to 2005. To analyze possible long-term trends in the global or regional SST throughout the period daily anomalies are computed using a 1961–90 daily climatology, averaged into global monthly means, and plotted as a global time series. To evaluate any biases in these anomalies they are compared with other satellite SST datasets that have been computed and compared over the same time period. Global infrared satellite SST data have been received from the Advanced Very High Resolution Radiometer (AVHRR) series, microwave SST data from the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI), and global microwave SST data from the Advanced Microwave Sounding Radiometer (AMSR)-E on Aqua. Additionally, the anomalies have also been compared with the Hadley Centre Global Sea Ice Coverage and Sea Surface Temperature (HadISST1) anomalies. HadISST1 is a globally complete 1° SST analysis compiled from in situ and bias-corrected AVHRR SSTs at the Met Office (UK).

The results of the study show the high accuracy of the Advanced Along Track Scanning Radiometer (AATSR) SSTs, but there are concerns with the NOAA-14 AVHRR data (1996–2000) being biased cold, especially in the Northern Hemisphere, and the AMSR-E SSTs (version 4), which show unexplained biases. Since 1999 TMI SSTs appear to have a consistently warm (∼0.2 K) bias relative to the infrared sensors and HadISST1.

The time series in (A)ATSR SSTs indicate the possibility of warming trends between 0.1 and 0.2 K decade−1, but the remaining ATSR-1 data are required to confirm this.

Corresponding author address: Anne G. O’Carroll, Met Office, Fitzroy Road, Exeter, EX1 3PB, United Kingdom. Email: anne.ocarroll@metoffice.gov.uk

Abstract

A near-continuous series of global retrievals of sea surface temperature (SST) has been made from the Along-Track Scanning Radiometer (ATSR) series of instruments from 1991 to 2005. To analyze possible long-term trends in the global or regional SST throughout the period daily anomalies are computed using a 1961–90 daily climatology, averaged into global monthly means, and plotted as a global time series. To evaluate any biases in these anomalies they are compared with other satellite SST datasets that have been computed and compared over the same time period. Global infrared satellite SST data have been received from the Advanced Very High Resolution Radiometer (AVHRR) series, microwave SST data from the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI), and global microwave SST data from the Advanced Microwave Sounding Radiometer (AMSR)-E on Aqua. Additionally, the anomalies have also been compared with the Hadley Centre Global Sea Ice Coverage and Sea Surface Temperature (HadISST1) anomalies. HadISST1 is a globally complete 1° SST analysis compiled from in situ and bias-corrected AVHRR SSTs at the Met Office (UK).

The results of the study show the high accuracy of the Advanced Along Track Scanning Radiometer (AATSR) SSTs, but there are concerns with the NOAA-14 AVHRR data (1996–2000) being biased cold, especially in the Northern Hemisphere, and the AMSR-E SSTs (version 4), which show unexplained biases. Since 1999 TMI SSTs appear to have a consistently warm (∼0.2 K) bias relative to the infrared sensors and HadISST1.

The time series in (A)ATSR SSTs indicate the possibility of warming trends between 0.1 and 0.2 K decade−1, but the remaining ATSR-1 data are required to confirm this.

Corresponding author address: Anne G. O’Carroll, Met Office, Fitzroy Road, Exeter, EX1 3PB, United Kingdom. Email: anne.ocarroll@metoffice.gov.uk

1. Introduction

The extent to which the global oceans are warming or cooling is a key quantity to determine for use in climate change detection studies. The oceans are a huge reservoir of energy allowing the slow release or storage of heat. Sea surface temperature (SST) is a measure of the temperature of the ocean’s surface layer, and once diurnal, seasonal, and other regional anomalies are removed, SST can be used as a robust monitor of the ocean surface climatology. It has been suggested (Allen et al. 1994) that 15 yr is a sufficient period to start to detect climatic trends in the ocean. Results of research on much longer time scales of 20 000 yr (e.g., Sabin and Pisias 1996) show how the changing SSTs significantly affect the climate on nearby continents.

Traditional attempts at monitoring global SST trends are performed using in situ ship and buoy SST data. However, problems with this approach include the scarcity of observations in some regions of the world, especially the southern oceans, and the measurements themselves being only point values and not necessarily representative of a spatially averaged area of the ocean. Additionally, measurements of SST using drifting, moored buoys and ship observations all employ different measurement methods, including observation at different depths causing the SST record to be inconsistent in time and space.

Satellite observations of SST have the advantage of providing a consistent global dataset averaged over areas equivalent to those used in climate model analyses. Their disadvantage is that any residual biases in the dataset must be understood and removed in order to not contaminate the in situ climate data record. This paper presents results from various satellite datasets to monitor global SST changes over the past 15 yr in order better to assess the accuracy of the various satellite SST datasets that could be used for climate monitoring.

Global retrievals of SST have been analyzed from a variety of different satellite sensors to compute SST anomalies. For the infrared sensors the Along-Track Scanning Radiometer (ATSR) series of instruments from 1991 to the present and the multichannel SST (MCSST) data from the Advanced Very High Resolution Radiometer (AVHRR) series have been analyzed. For the microwave sensors tropical microwave SST data from the Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI), and global microwave SST data from the Advanced Microwave Sounding Radiometer (AMSR)-E on Aqua were analyzed. Additionally, the satellite anomalies have also been compared with a climate SST analysis, the Hadley Centre Global Sea Ice Coverage and Sea Surface Temperature (HadISST1), which is a globally complete 1° SST analysis compiled from in situ and AVHRR SSTs.

A similar recent study of ATSR-1 and ATSR-2 SST anomalies using averaged SST (ASST) data is given in Lawrence et al. 2004. ASST data supplies a preretrieved skin SST at 0.5° resolution. The Advanced Along Track Scanning Radiometer (AATSR) data used within this report is from a new dataset—the Cloud Screened Consolidated Averaged Brightness Temperature (CSCABT) data, provided by the Rutherford Appleton Laboratory (RAL), United Kingdom—allowing control over the retrieval of skin SST and at a higher spatial resolution of 1/6° (10 arcmin).

The format of the paper is as follows: section 2 describes the SST datasets used, section 3 the methodology for deriving the SST anomalies, section 4 the results, and section 5 the conclusions.

2. Description of SST datasets

A short description of the various satellite and model datasets used is given here. More details are provided in the references.

a. Along-Track Scanning Radiometer measurements

The ATSR has to date been flown on three satellite missions: ATSR-1 on the European Space Agency (ESA) Remote Sensing Satellite (ERS)-1 (1991–95), ATSR-2 on ERS-2 (1995–present), and the (A)ATSR on Envisat (2002–present). Each instrument has three infrared channels centered in the atmospheric windows at 3.7, 10.8, and 12 μm plus channels in the visible/near-infrared part of the spectrum used for cloud detection. ATSR has an inclined conical scan, which enables it to make observations of the surface from two different angles—nadir and a forward view (∼55°)—within a few minutes of each other, allowing for better correction of atmospheric absorption and aerosol than AVHRR. The ATSR series of instruments are designed to retrieve SST to better than 0.30-K absolute accuracy in the derived skin SST and give a long-term stability of better than 0.1 K. More details on the (A)ATSR instruments and retrieval of skin SST from the data can be found in Edwards et al. (1990), Smith et al. (2001), and Mutlow et al. (1994) and from the ATSR Web site (available online at http://www.atsr.rl.ac.uk/). The ATSR-1/-2 datasets used for this study are from the averaged brightness temperature (ABT) files produced at RAL, which provide cloud-free top-of-atmosphere brightness temperatures averaged over 10-arcmin cells. The ATSR-2 data have been consolidated to remove any data gaps. The AATSR data come from the ESA near-real-time meteoproduct, which also contains brightness temperatures. There are small gaps in the latter resulting from occasional problems with the data dissemination.

In this paper two types of dual-view ATSR skin SST retrievals are used—those using all three infrared channels (hereafter D3) during the night only because of the solar contribution to the 3.7-μm channel during the day, and retrievals using just the 10.8- and 12-μm channels (hereafter D2) during both day and night. Results are presented for ATSR-1 SSTs from September 1991 to February 1992 and from August 1992 to November 1994, for ATSR-2 SSTs from 1995 to 2000, and for AATSR SSTs from 2002 to 2004. The gap in 2001–02 is due to the gyro failure on ERS-2, which has resulted in inferior ATSR-2 measurements since January 2001. Work is underway to apply corrections to these data so they can be included in the time series. The early ATSR-1 SSTs presented are for D3 retrievals. However, the 3.7-μm channel failed on ATSR-1 in May 1992, so ATSR-1 SSTs used after this date are D2 SST retrievals only.

The (A)ATSR skin SSTs have been processed to a “pseudobulk” SST, which is the temperature of the ocean at around 1 m in depth, using the Fairall model (Fairall et al. 1996a). This is necessary because the satellite observes a radiative skin temperature that is usually cooler than the subskin by more than 0.1 K on monthly time scales. For obtaining SST for climate purposes we require the bulk temperature, which provides a better comparison to AVHRR bulk SSTs and HadISST1 than (A)ATSR skin SSTs, at the expense of relying on the accuracy of NWP surface wind fields to enable a skin-to-bulk correction to be made. More information on the processing of (A)ATSR skin SSTs to bulk SSTs is contained in Horrocks et al. (2003), O’Carroll et al. (2004, 2006). The AATSR and ATSR-2 plotted data are for the nighttime D3 retrieval, and the ATSR-1 plotted data for the period 1991–92 are for the D3 retrievals. However, from August 1992 to 1994 the plotted data for ATSR-1 are from D2 retrievals, due to the failure of the 3.7-μm channel on ATSR-1. Further time series for all experiments, which include the standard deviations and number of observations for each mean SST time series, are reported in O’Carroll et al. (2005).

Satellite SST retrievals can be subject to systematic biases. The biases in the (A)ATSR retrievals, for which coefficients are derived via radiative transfer modeling, can be summarized as retrieval errors that are state dependent and also as a mean global offset, which varies according to whether a D2 or D3 retrieval has been performed. In this paper, anomaly time series are only presented when the (A)ATSR data have been bias corrected to attempt to correct for the mean global offset with respect to in situ data. The bias corrections were derived by utilizing a database that contains collocated (A)ATSR and quality-controlled buoy SST observations. Bias statistics and validation results using this match-up database are presented in O’Carroll et al. (2004, 2006). To evaluate the global biases, the buoy “bulk” SSTs were first converted to a buoy “skin” SST using the Fairall model (Fairall et al. 1996a). These buoy skin SSTs were then compared to the (A)ATSR skin SSTs during night only to derive the mean global offsets. It was found that these offsets varied annually, especially just after launch, and so for the purposes of correcting the (A)ATSR anomaly plots, offsets that were derived separately for each year are used, and these are presented in Table 1.

b. Advanced Very High Resolution Radiometer measurements

The AVHRR sensor is on board the operational National Oceanic and Atmospheric Administration (NOAA) polar-orbiting satellites, and data are available from 1982, when NOAA-7 was launched, to the present. Continuity of AVHRR data is likely to be ensured until 2020, because the instrument will also fly on the European polar orbiter, the Meteorological Operational (METOP) weather satellite. AVHRR has an instantaneous field of view of 1.1 km, which scans across the swath perpendicular to the ground track of the satellite, which is in a sun-synchronous orbit. It has two visible channels at 0.6 and 0.9 μm—one short-wavelength infrared channel at 3.7 μm and two long-wavelength infrared channels at 11 and 12, μm respectively. From the 3.7-, 11-, and 12-μm channels, brightness temperatures can be inferred, which are used to calculate SST (Pichel 1991). AVHRR measures top-of-the-atmosphere radiances and then SSTs are retrieved using algorithms that have been tuned by regression of brightness temperatures from the instrument’s infrared channels onto in situ SST data from a set of drifting buoys. Therefore, AVHRR SSTs are equivalent to measurements of the bulk surface temperature of the sea in contrast to (A)ATSR SSTs.

The AVHRR SST data used for this paper are not bias corrected, and take the form of monthly “super observations,” separately estimated for day and night. These superobservations, generated at the National Centers for Environmental Prediction (NCEP), were calculated by taking averages of weekly observations for all weeks falling totally or mainly in the target month. The monthly means of nighttime SST of an area 1° in latitude by 1° in longitude were used for this study. The AVHRR data were obtained from NOAA-7 from 1982 to February 1985, from NOAA-9 up to November 1988, from NOAA-11 up to September 1994, from NOAA-12 up to March 1995, from NOAA-14 up to September 2001, from NOAA-16 up to July 2004, and from NOAA-17 thereafter.

c. TRMM Microwave Imager measurements

The TMI is on board TRMM, a joint National Aeronautics and Space Administration (NASA) and National Space Development Agency (NASDA) mission, which was launched in November 1997. The TRMM satellite moves from west to east in an inclined orbit providing data at various local times between 40°S and 40°N. The TMI is a conical scan microwave radiometer with channels at five separate frequencies (10.7, 19.4, 21.3, 38, and 85 GHz) and a spatial resolution of about 40 km. The 10.7-GHz channel is used to retrieve SST because it is insensitive to water vapor and cloud. This gives an advantage over infrared sensors because cloudy (although nonprecipitating) SST coverage can be obtained with TMI. In addition to precipitation, the TMI does have sensitivities to sea surface roughness, and so the microwave SST retrieval attempts to account for these effects. Details of the physical retrieval algorithm, which aims to derive SST to an accuracy better than 0.5 K, are given in Wentz and Meissner (2000). The TMI SSTs have been retrieved as daily averaged files (version 4) from the Remote Sensing Systems (RSS) Web site (available online at http://www.ssmi.com) between −40° and +40° latitude. The microwave-derived SSTs measure at a depth down to a few millimeters, which is deeper than the infrared measurements. It should be noted that RSS have applied bias corrections to the TMI SSTs (Wentz et al. 2001).

d. Advanced Microwave Sounding Radiometer measurements

The AMSR-E instrument is on board the sun-synchronous Aqua satellite, which was launched in May 2002. The instrument was provided to NASA by the NASDA. Global, daily, version 4 AMSR-E products of SST, wind speed, atmospheric water vapor, cloud water, and rain rate are available at 0.25° spatial resolution (see information online at http://www.ssmi.com). As for TMI, the ability of AMSR-E to view SST through nonprecipitation clouds is a major advantage over traditional cloud-free infrared measurements of SST, although microwave SST retrievals have great sensitivity to surface roughness changes and thus local surface wind speed. Details of the AMSR-E SST algorithm are given in Wentz and Meissner (2000), and the retrieved SST product has the same characteristics as that of TMI.

e. HadISST1 dataset

The HadISST1 (Rayner et al. 2003) dataset is a globally complete 1° spatial resolution sea ice and SST analysis field, produced on a monthly basis at the Hadley Centre. The HadISST1 analyses run from 1870 to the present day and offer a consistent climate record of SST. The SST data that contribute to HadISST1 come from ships, buoys, and AVHRR MCSST observations. The ship and buoy data are averaged into a 1° latitude × 1° longitude globally complete field, from which a nighttime AVHRR SST field is subtracted. The difference is smoothed to create a bias field for which the AVHRR data can be corrected. The corrected AVHRR and in situ fields are then averaged together onto a 2° latitude × 2° longitude grid, which is reconstructed as a 1° spatial resolution field, with gaps covered by reduced space optimal interpolation (RSOI) (Kaplan et al. 1997) and sea ice fields added to produce the final HadISST1 field.

3. Methodology

(A)ATSR anomalies are produced by subtracting a daily 1° 1961–90 Global Sea Ice and Sea Surface Temperature (GISST) climatology (Parker et al. 1995) from daily (A)ATSR skin and bulk (D2 or D3) SST observations averaged to 10-arcmin resolution. These anomalies are then spatially averaged onto a 1° grid and temporally averaged to produce global monthly mean anomalies. The TMI and AMSR-E anomalies were created by using 0.25° SSTs averaged to 1° cells, and then subtracting the GISST climatology. These daily anomalies are then averaged to a global monthly mean anomaly. The AVHRR and HadISST1 SST anomalies were created by the same method, but use observations averaged over 1° cells.

SST anomaly time series from different sources have been plotted together. It is important to note that for these plots global mean anomalies were only computed where coincident observations were available for all the SST data sources in the plot. There are two exceptions to this rule, which are identified and described in the relevant paragraphs in section 4. Plots containing TMI data are only presented for SST observations within ±40° latitude resulting from TMI data only being available within these latitudes. Additionally, time series are included for data within the El Niño region of 5°S–5°N, 120°–170°W.

A number of different anomaly time series were plotted according to different SST sources [e.g., bulk/skin D2/D3 bias-corrected (A)ATSR SSTs], different regions, and whether the SST anomalies from different sources are coincident with each other. A summary of the undertaken plots detailing the SST anomaly–produced time series is shown in Table 2.

4. Results

a. SST time series

Figure 1 shows the time series of global mean (A)ATSR SSTs and HadISST used in this paper. It shows indications of a slight warming in global temperatures over the 10-yr period presented. The seasonal cycle can be clearly observed.

b. Anomaly time series

The global time series presented in Fig. 2 show that for the bias-corrected time series, both ATSR-2 and AATSR SSTs are cooler than HadISST1; however, AVHRR SSTs change from being consistently the coolest up to 2000, to being in close agreement with HadISST1 from 2002 onward. Apart from an offset of up to 0.1 K, (A)ATSR and HadISST1 track each other very closely. Note the presence of the 1998 El Niño event, which increases the global SST anomalies for all sensors in 1997–98. During 1991 and early 1992 the AVHRR SSTs are significantly cooler than those for both ATSR-1 and HadISST, likely due to a cooling effect in the single-view AVHRR SSTs from the 1991 Mount Pinatubo volcanic eruption, where vast amounts of aerosol were injected into the stratosphere. The ATSR-1 SSTs do not display a similar cooling, because the (A)ATSR SST retrieval method used in this study is designed to be robust to stratospheric aerosol (Merchant and Harris 1999) in the field of view.

The global time series presented in Fig. 3 show similar data to those plotted in Fig. 2; however, the (A)ATSR SSTs are not coincident with the other SST sources, which allows AVHRR data for the missing (A)ATSR data period in 2001/02 to be plotted, and AVHRR data to be plotted back to 1982. AVHRR SSTs follow bias-corrected (A)ATSR SSTs more closely than HadISST1. However, the AVHRR line clearly shows the cooling effect of the 1982 El Chichón and 1991 Mount Pinatubo volcanic eruption, where vast amounts of aerosol were injected into the stratosphere. The AVHRR anomalies show at least three cool episodes during 1991–94, which are primarily attributed to problems in the retrieval resulting from Mount Pinatubo aerosols. See Reynolds (1992) for more details of the impact of the Mount Pinatubo aerosols on AVHRR SSTs. The HadISST1 anomalies do not appear to be affected by these eruptions. Only ATSR-1 data from late 1991 and early 1992 are currently presented, but these limited data do indicate that the ATSR-1 SST retrievals are better able to cope with the existence of stratospheric aerosol than the AVHRR SSTs. We would expect (A)ATSR SST retrievals to be less effected by the presence of stratospheric aerosol because of the dual-view capability and the use of aerosol-robust SST retrieval coefficients, as described in Merchant and Harris (1999).

Overall, Fig. 3 shows that HadISST1 had good agreement with the general trend of AVHRR (which is subject to wider fluctuations) up to 1995, and then AVHRR is cooler than HadISST1 until 2002, from which point they are once again similar. The reason for this is that NOAA-14 drifted in its equator-crossing time so that solar radiation was able to enter the instrument and cause it to be thermally unstable, resulting in errors in the calibration (A. Harris 2005, personal communication). The NOAA-16 AVHRR had an improved design to reduce this effect, and this is seen in the much lower biases with respect to the other datasets. The effect is greatest in the Southern Hemisphere.

The time series in Fig. 4 show global (A)ATSR, HadISST1, AVHRR, and AMSR-E SSTs, where they are all coincident observations. The AMSR-E SSTs seem to have a lag in the anomaly pattern and reduced amplitude compared to the other datasets, which is hard to explain. The AVHRR SSTs are much warmer than the other SSTs from mid-2004 onward, which may be due to the switch to NOAA-17. HadISST1 is also slightly warmer, because of its assimilation of AVHRR data.

Over the TMI region Fig. 5 shows that within latitude bounds of 40°N–40°S bias-corrected (A)ATSR and HadISST1 are very similar throughout the period, although the AATSR SSTs are slightly cooler than HadISST than for ATSR-2, suggesting that the AATSR bias correction may be slightly too great. AVHRR SSTs are close to bias-corrected AATSR SSTs and HadISST1 from 2002 onward, but are significantly cooler from 1998 to 2000. This supports the observation of inconsistency in AVHRR observed in the global anomalies. Looking at the TMI anomalies, TMI SSTs are similar to the bias-corrected ATSR-2 SSTs and HadISST1 from 1998 to 1999, but are warmer than ATSR and HadISST1 from 1999 onward by ∼0.2 K, which could suggest a consistent warm bias for the TMI SSTs. TMI anomaly data are plotted where the observations are both coincident and noncoincident with the other data. In the latter case this enables TMI data to be plotted in the ATSR-2 to AATSR missing data period (2001/02), and therefore acts as a reference dataset to bridge the gap. Looking at the gap in Fig. 5, a similar pattern is seen as for the periods before and after where coincident data were used.

Figures 6a and 6b show anomaly time series for the Northern and Southern Oceans (i.e., areas of 20°–60°N/20°–60°S, 180°–180°, respectively). For the Northern Hemisphere time series (Fig. 6a), the AVHRR SST anomalies are seen to be much cooler than HadISST1 by up to 0.5 K and those for (A)ATSR are slightly cooler (∼0.15 K). For the Southern Ocean (Fig. 6b), HadISST1 anomalies are shown to be in between bias-corrected ATSR-2 and AVHRR before 2001. However, HadISST1 anomalies are very similar to the AATSR SST anomalies from 2002 onward.

Figure 7 shows a time series for the tropical Pacific (5°S–5°N, 120°–170°W) for all coincident observations (hence, the missing data period in 2001/02), and it starts from 1998 when TMI SST data became available. The effect of the 1997/98 El Niño on the anomaly is clearly seen at the start of the period. Additionally, there is a smaller warming observed in late 2002. (A)ATSR follows HadISST1 most closely with the exception of late 1998, where it is believed that the HadISST1 analysis did not capture the decline of the El Niño accurately, as shown by Saunders et al. (2004), and the AATSR/TMI data are probably more accurate. In agreement with the global anomalies AVHRR has a cool bias during 2000. The TMI data for this region do not have a warm bias, unlike the full TMI statistics.

c. Global maps of SST anomalies for July 2003

July 2003 is noted for the event of a significant Saharan dust plume outbreak over the tropical Atlantic (Haywood et al. 2005). Mineral dust can significantly affect satellite retrievals of SST, making July 2003 an interesting month to study global SST anomalies from different satellite sensors and other SST datasets. Figure 8 shows the global SST anomalies for non-bias- and bias-corrected AATSR, TMI, HadISST1, nighttime non-bias-corrected AVHRR, and AMSR-E for July 2003. These anomaly maps have been constructed using differences of monthly mean SSTs over 1° spatial resolution. July 2003 was chosen because a major Saharan dust plume is known to have extended over the Atlantic during this month (see Haywood et al. 2005). Figure 8 shows clearly how AVHRR SST anomalies are cooled by the aerosol in the mid-Atlantic during July 2003, while the AATSR SST anomalies are better able to cope because of the instruments’ dual-view capability.

All images in Fig. 8 show similar characteristics in the distribution of maximum and minimum anomaly differences, although there are more subtle differences between the plots. The AVHRR SSTs display cooler SSTs in the tropical Atlantic at a latitude of 20°N, likely to be due to the presence of Saharan dust, while the HadISST1, TMI, and AATSR plots show neutral or warm SST anomalies in this region. The dual-view capability of AATSR enables improved SST retrieval in such regions.

All of the SST sources show a warm anomaly around the Gulf Stream. However, HadISST1 and AVHRR view a cool anomaly around 55°N, which is barely seen in the AATSR plot because of the lack of data. It may be that the stringent AATSR cloud-detection scheme is wrongly assigning pixels in this region to cloud because of the high spatial variation of SSTs here.

One area seen as a cool anomaly in all the satellite data but not by HadISST1 is off the east coast of Africa, and because both microwave sensors see this it cannot be because of atmospheric dust or cloud contamination. Cool anomalies in the North and South Pacific are observed by all SST sources, although the extent seems to be reduced in HadISST1.

5. Conclusions

SST anomaly plots from four independent satellite datasets have been presented, along with the anomalies computed from the HadISST1, analyses in order to assess the biases associated with each of the satellite datasets and HadISST1 analyses. The magnitude and variability of these biases are important to be quantified if the satellite SSTs are to be used for climate monitoring. Based on the analysis presented we can conclude the following about the various satellite datasets analyzed.

Using HadISST1 SST anomalies as a reference dataset, AVHRR global SST anomalies show occasional cool episodes (up to 0.5 K relative to HadISST1) in the period of 1991–2001 due partly to stratospheric aerosols (1991–92) and partly to the AVHRR instrument problems and orbital drift. The magnitude of these cool anomalies are larger than those seen by (A)ATSR. Post-2001, when NOAA-16 data replaced NOAA-14 data, the anomalies become similar to HadISST1 and other satellite datasets, except for mid-2004 when a positive bias in the AVHRR became evident when NOAA-17 data started being used. The regional studies show that in the Northern Hemisphere AVHRR SST anomalies are much cooler (∼0.5 K) than the HadISST1 and (A)ATSR SST anomalies prior to 2001, because of NOAA-14 problems.

For the (A)ATSR SSTs small bias corrections are required to remove residual biases from the skin SST retrieval and conversion to bulk SST. ATSR-1 and ATSR-2 bulk SSTs are cooler than those of HadISST1 (∼0.1 K globally), but AATSR SSTs are closer (∼0.05 K cool). ATSR SST anomalies are, in general, closer to the HadISST1 anomalies both globally and locally and are much less affected by stratospheric aerosols during 1991–92 than AVHRR. Figure 8 also shows that AATSR data are not affected by desert dust outbreaks in July 2003, whereas the AVHRR data are negatively biased by up to 1 K. The coverage of (A)ATSR data is not as complete as that of AVHRR data, however, because of the narrower swath width.

Standard deviations for global HadISST1 anomaly differences (not shown) are consistently lower than for AVHRR and (A)ATSR anomalies because of the GISST climatology being compiled in the same way as that of HadISST. The AVHRR and (A)ATSR global standard deviations are both similar and about 30% greater than the corresponding HadISST1 values.

TMI SST anomalies are similar to ATSR-2 anomalies up to early 1999, but are warmer (∼0.2 K) than ATSR-2/AATSR anomalies thereafter. There is some indication the TMI SSTs are contaminated by land in some areas, which causes a warm bias in the TMI data. The AMSR-E SST anomalies appear to display a lag in their pattern behind the other SST sources (AATSR, AVHRR, and HadISST1), indicating problems with the AMSR-E SSTs from version 4 of the processing, which was the latest available version at the time of writing. No similar lag is observed in the TMI SST anomalies. There are known problems with the calibration of AMSR-E data.

The analysis in the tropical Pacific clearly shows the warming effect of the 1998 El Niño seen by all satellites, which warms SST anomalies by up to 3 K in this region. It is of interest to note that the ATSR-2 and TMI data both show the transition to cool anomalies in mid-1998, which is more rapid compared to that of the AVHRR and HadISST1 data.

The inference of climate warming trends from these SST anomaly series is difficult over such a short period. However, the combination of five different estimates of SST anomalies does start to provide confidence in the SST climate record and also allow more meaningful error bars to be assigned to the analyses. The error characteristics of each satellite dataset can also be better understood through comparisons such as this. The time series from 1991, which include the (A)ATSR data, do support the possibility of warming trends between 0.1 and 0.2 K decade−1.

Some recommendations from this study can be made as follows.

  • (i) AVHRR data provide good coverage and continuity of SST data back to 1982. Biases are introduced in the data in regions of high aerosol optical depth, both in the stratosphere and troposphere. There are also issues with the AVHRR calibration resulting from solar intrusion and drift of the local crossing time of the orbit. The data should be used for SST retrievals only where the above problems are not evident or corrected.

  • (ii) (A)ATSR data are of good quality and are not affected by aerosols or calibration issues. The coverage is more limited than AVHRR but is sufficient for monthly means. All the data can be used with confidence after a small bias correction is applied.

  • (iii) TMI SSTs have a warm bias (∼0.2 K) globally, but in the tropical Pacific are close to the infrared datasets and the analyses.

  • (iv) AMSR-E still has some issues with its calibration leading to an unexplained lag in the SST anomaly. The microwave SSTs should be used with caution for input to climate datasets, but are valuable for validation thereof.

Acknowledgments

This work was carried out in support of the Climate Predication Programme, funded by the United Kingdom Department for Environment, Food and Rural Affairs (DEFRA). ATSR-1 and ATSR-2 data are ESA/NERC/CCLRC. TMI and AMSR-E data are produced by Remote Sensing Systems and sponsored by the NASA Earth Science REASoN DISCOVER Project and the AMSR-E Science Team. Data are available online at http://www.remss.com. The authors thank Diane Stokes from the National Centers for Environmental Prediction (NCEP) for the supply of the AVHRR data.

REFERENCES

  • Allen, M. R., Mutlow C. T. , Blumberg G. M. C. , Christy J. R. , McNider R. T. , and Llewellyn-Jones D. T. , 1994: Global change detection. Nature, 370 , 182.

    • Search Google Scholar
    • Export Citation
  • Edwards, T., and Coauthors, 1990: The Along Track Scanning Radiometer measurements of sea surface temperature from ERS-1. J. Br. Interplanet. Soc., 43 , 160180.

    • Search Google Scholar
    • Export Citation
  • Fairall, C. W., Bradley E. F. , Godfrey J. S. , Wick G. A. , Edson J. B. , and Young G. S. , 1996a: Cool-skin and warm layer effects on sea surface temperature. J. Geophys. Res., 101 , C1. 12951308.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Haywood, J. M., Allan R. P. , Culverwell I. , Slingo T. , Milton S. , Edwards J. , and Clerbaux N. , 2005: Can desert dust explain the outgoing longwave radiation anomaly over the Sahara during July 2003? J. Geophys. Res., 110 .D05105, doi:10.1029/2004JD005232.

    • Search Google Scholar
    • Export Citation
  • Horrocks, L. A., Candy B. , Nightingale T. , Saunders R. W. , O’Carroll A. G. , and Harris A. R. , 2003: Parameterisations of the skin effect and implications for satellite-based measurement of sea-surface temperature. J. Geophys. Res., 108 .3096, doi:10.1029/2002JC001503.

    • Search Google Scholar
    • Export Citation
  • Kaplan, A., Kushnir Y. , Cane M. A. , and Blumenthal M. B. , 1997: Reduced space optimal analysis for historical data sets: 136 years of Atlantic sea surface temperatures. J. Geophys. Res., 102 , C13. 2783527860.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Lawrence, S. P., Llewellyn-Jones D. T. , and Smith S. J. , 2004: The measurement of climate change using data from the Advanced Very High Resolution and Along Track Scanning Radiometers. J. Geophys. Res., 109 .C08017, doi:10.1029/2003JC002104.

    • Search Google Scholar
    • Export Citation
  • Merchant, C. J., and Harris A. R. , 1999: Toward the elimination of bias in satellite retrievals of sea surface temperature 1. Theory, modelling and interalgorithm comparison. J. Geophys. Res., 104 , C10. 2356523578.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mutlow, C. T., Zavody A. M. , Barton I. J. , and Llewellyn-Jones D. T. , 1994: Sea surface temperature measurements by the along-track scanning radiometer on the ERS-1 satellite: Early results. J. Geophys. Res., 99 , C11. 2257522588.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • O’Carroll, A. G., Saunders R. W. , Watts J. G. , and Horrocks L. A. , 2004: Reprocessing of the ATSR-2 clear-sea consolidated averaged brightness temperature dataset to skin and bulk sea surface temperatures; including a cross-comparison with ATSR-1 SSTs from 1995. Met Office NWP Forecasting Research Tech. Rep. 430, 26 pp. [Available online at http://www.metoffice.gov.uk/research/nwp/publications/papers/technical_reports/2004/FRTR430/FRTR430.pdf.].

  • O’Carroll, A. G., Saunders R. W. , and Watts J. G. , 2005: Comparison of (A)ATSR SST anomalies with other SST datasets. Met Office NWP Tech. Rep. 462, 30 pp.

  • O’Carroll, A. G., Watts J. G. , Horrocks L. A. , Saunders R. W. , and Rayner N. A. , 2006: Validation of the AATSR Meteo product sea surface temperature. J. Atmos. Oceanic Technol., 23 , 711726.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Parker, D. E., Jackson M. , and Horton E. B. , 1995: The GISST2.2 sea surface temperature and sea-ice climatology. Hadley Centre Climate Research Tech. Note CRTN63, 16 pp.

  • Pichel, W. G., 1991: Operational production of multichannel sea surface temperature from NOAA polar satellite AVHRR data. Global Planet. Change, 90 , 173177.

    • Search Google Scholar
    • Export Citation
  • Rayner, N. A., Parker D. E. , Horton E. B. , Folland C. K. , Alexander L. V. , Rowell D. P. , Kent E. C. , and Kaplan A. , 2003: Global analyses of sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. J. Geophys. Res., 108 .4407, doi:10.1029/2002JD002670.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., 1992: Impact of Mount Pinatubo aerosols on satellite-derived sea surface temperatures. J. Climate, 6 , 768774.

  • Sabin, A. L., and Pisias N. G. , 1996: Sea surface temperature changes in the north eastern Pacific Ocean during the past 20, 000 years and their relationship to climate change in north-western North America. Quat. Res., 46 , 4861.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Saunders, R. W., Roberts T. , O’Carroll A. G. , and Horrocks L. A. , 2004: Comparison of ATSR-2/AATSR and TMI sea surface temperatures with the Hadley Centre analyses. Met Office NWP Tech. Rep. 426, 22 pp. [Available online at http://www.metoffice.com/research/nwp/publications/papers/technical_reports/2004/FRTR426/FRTR426.pdf.].

  • Smith, D. L., Delderfield J. , Drummond D. , Edwards T. , Mutlow C. T. , Read P. D. , and Toplis G. M. , 2001: Calibration of the AATSR instrument. Adv. Space Res., 28 , 3139.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wentz, F., and Meissner T. , 2000: Algorithm Theoretical Basis Document (ATBD) version 2: AMSR ocean algorithm. RSS Tech. Proposal 121599A-1, 66 pp. [Available online at http://www.ssmi.com/papers/AMSR_Ocean_Algorithm_Version_2.pdf.].

  • Wentz, F., Ashcroft P. , and Gentemann C. , 2001: Post launch calibration of the TRMM Microwave Imager. IEEE Trans. Geosci. Remote Sens., 39 , 415422.

Fig. 1.
Fig. 1.

Global mean (A)ATSR SSTs and HadISST1 from 1995 to 2005.

Citation: Journal of Atmospheric and Oceanic Technology 23, 11; 10.1175/JTECH1934.1

Fig. 2.
Fig. 2.

Global bias-corrected (A)ATSR, HadISST1, and nighttime AVHRR SST anomalies, all coincident.

Citation: Journal of Atmospheric and Oceanic Technology 23, 11; 10.1175/JTECH1934.1

Fig. 3.
Fig. 3.

Global bias-corrected (A)ATSR, HadISST1, and nighttime AVHRR SST anomalies, all coincident except for (A)ATSR.

Citation: Journal of Atmospheric and Oceanic Technology 23, 11; 10.1175/JTECH1934.1

Fig. 4.
Fig. 4.

Global bias-corrected (A)ATSR, HadISST1, nighttime AVHRR, and AMSR-E SST anomalies, all coincident.

Citation: Journal of Atmospheric and Oceanic Technology 23, 11; 10.1175/JTECH1934.1

Fig. 5.
Fig. 5.

Tropical bias-corrected (A)ATSR, HadISST1, nighttime AVHRR, and TMI, all coincident.

Citation: Journal of Atmospheric and Oceanic Technology 23, 11; 10.1175/JTECH1934.1

Fig. 6.
Fig. 6.

Bias-corrected (A)ATSR, HadISST1, and nighttime AVHRR, all coincident, for the (a) Northern Ocean (20°–60°N, 180°–180°) and (b) Southern Ocean (20°–60°S, 180°–180°).

Citation: Journal of Atmospheric and Oceanic Technology 23, 11; 10.1175/JTECH1934.1

Fig. 7.
Fig. 7.

Tropical Pacific (5°S–5°N, 120°–170°W), bias-corrected (A)ATSR, HadISST1, nighttime AVHRR, and TMI, all coincident.

Citation: Journal of Atmospheric and Oceanic Technology 23, 11; 10.1175/JTECH1934.1

Fig. 8.
Fig. 8.

Global SST anomalies for July 2003 for different SST sources: (top left) non-bias-corrected (three-channel night) AATSR, (top right) bias-corrected (three-channel night) AATSR, (middle left) nighttime non-bias-corrected AVHRR, (middle right) HadISST1, (bottom left) TMI, and (bottom right) AMSR-E.

Citation: Journal of Atmospheric and Oceanic Technology 23, 11; 10.1175/JTECH1934.1

Table 1.

(A)ATSR bias corrections: yearly mean global offsets applied to (A)ATSR SSTs.

Table 1.
Table 2.

Details of anomaly time series produced.

Table 2.
Save