• Doelling, D. R., , Lukashin C. , , Minnis P. , , Scarino B. , , and Morstad D. , 2012: Spectral reflectance corrections for satellite intercalibrations using SCIAMACHY data. IEEEGeosci. Remote Sens. Lett, 9, 119123, doi:10.1109/LGRS.2011.2161751.

    • Search Google Scholar
    • Export Citation
  • Harries, J. E., and Coauthors, 2005: The Geostationary Earth Radiation Budget Project. Bull. Amer. Meteor. Soc., 86, 945960.

  • Kato, S., , and Loeb N. G. , 2003: Twilight irradiance reflected by the earth estimated from Clouds and the Earth’s Radiant Energy System (CERES) measurements. J. Climate, 16, 26462650.

    • Search Google Scholar
    • Export Citation
  • Kato, S., , and Loeb N. G. , 2005: Top-of-atmosphere shortwave broadband observed radiance and estimated irradiance over polar regions from Clouds and the Earth’s Radiant Energy System (CERES) instruments on Terra. J. Geophys. Res., 110, D07202, doi:10.1029/2004JD005308.

    • Search Google Scholar
    • Export Citation
  • Keyes, D. F., , Doelling D. R. , , Young D. F. , , Rose F. G. , , Rutan D. A. , , Nordeen M. L. , , and Boghosian J. S. , 2006: The validation of the 5-year Terra-based monthly CERES radiative flux and cloud product. Proc. 12th Conf. on Atmospheric Radiation, Madison, WI, Amer. Meteor. Soc., P3.1. [Available online at https://ams.confex.com/ams/Madison2006/techprogram/paper_112731.htm.]

  • Lazzara, M. A., and Coauthors, 1999: The Man Computer Interactive Data Access System: 25 years of interactive processing. Bull. Amer. Meteor. Soc., 80, 271284.

    • Search Google Scholar
    • Export Citation
  • Loeb, N. G., , Priestley K. J. , , Kratz D. P. , , Geier E. B. , , Green R. N. , , Wielicki B. A. , , Hinton P. O. R. , , and Nolan S. K. , 2001: Determination of unfiltered radiances from the Clouds and the Earth’s Radiant Energy System (CERES) instrument. J. Appl. Meteor., 40, 822835.

    • Search Google Scholar
    • Export Citation
  • Loeb, N. G., , Manalo-Smith N. , , Kato S. , , Miller W. F. , , Gupta S. K. , , Minnis P. , , and Wielicki B. A. , 2003: Angular distribution models for top-of-atmosphere radiative flux estimation from the Clouds and the Earth’s Radiant Energy System instrument on the Tropical Rainfall Measuring Mission Satellite. Part I: Methodology. J. Appl. Meteor., 42, 240265.

    • Search Google Scholar
    • Export Citation
  • Loeb, N. G., , Kato S. , , Loukachine K. , , and Smith N. M. , 2005: Angular distribution models for top-of-atmosphere radiative flux estimation from the Clouds and the Earth’s Radiant Energy System instrument on the Terra satellite. Part I: Methodology. J. Atmos. Oceanic Technol., 22, 338351.

    • Search Google Scholar
    • Export Citation
  • Loeb, N. G., , Kato S. , , Loukachine K. , , Smith N. M. , , and Doelling D. R. , 2007: Angular distribution models for top-of-atmosphere radiative flux estimation from the Clouds and the Earth’s Radiant Energy System instrument on the Terra satellite. Part II: Validation. J. Atmos. Oceanic Technol., 24, 564584.

    • Search Google Scholar
    • Export Citation
  • Loeb, N. G., , Wielicki B. A. , , Doelling D. R. , , Smith G. L. , , Keyes D. F. , , Kato S. , , Manalo-Smith N. , , and Wong T. , 2009: Toward optimal closure of the earth’s top-of-atmosphere radiation budget. J. Climate, 22, 748766.

    • Search Google Scholar
    • Export Citation
  • Loeb, N. G., , Lyman J. M. , , Johnson G. C. , , Allen R. P. , , Doelling D. R. , , Wong T. , , Soden B. J. , , and Stephens G. L. , 2012: Observed changes in top-of-the-atmosphere radiation and upper-ocean heating consistent within uncertainty. Nat. Geosci., 5, doi:10.1038/ngeo1375.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., , Smith W. L. Jr., , Garber D. P. , , Ayers J. K. , , and Doelling D. R. , 1994: Cloud properties derived from GOES-7 for spring 1994 ARM intensive observing period using version 1.0.0 of ARM Satellite Data Analysis Program. NASA Reference Publ. 1366, 61 pp.

  • Minnis, P., , Nguyen L. , , Doelling D. R. , , Young D. F. , , and Miller W. , 2002a: Rapid calibration of operational and research meteorological satellite imagers. Part I: Evaluation of research satellite visible channels as references. J. Atmos. Oceanic Technol., 19, 12331249.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., , Nguyen L. , , Doelling D. R. , , Young D. F. , , Miller W. , , and Kratz D. P. , 2002b: Rapid calibration of operational and research meteorological satellite imagers. Part II: Comparison of infrared channels. J. Atmos. Oceanic Technol., 19, 12501266.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., , Doelling D. R. , , Nguyen L. , , Miller W. , , and Chakrapani V. , 2008: Assessment of the visible channel calibrations of the TRMM VIRS and MODIS on Aqua and Terra. J. Atmos. Oceanic Technol., 25, 385400.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., and Coauthors, 2011: CERES edition-2 cloud property retrievals using TRMM VIRS and Terra and Aqua MODIS data—Part I: Algorithms. IEEE Trans. Geosci. Remote Sens., 49, 43744400, doi:10.1109/TGRS.2011.2144601.

    • Search Google Scholar
    • Export Citation
  • Morstad, D. L., , Doelling D. R. , , Bhatt R. , , and Scarino B. , 2011: The CERES calibration strategy of the geostationary visible channels for CERES clouds and flux products. Earth Observing Systems XVI, J. J. Butler, X. Xiong, and X. Gu, Eds., International Society for Optical Engineering (SPIE Proceedings, Vol. 8153), doi:10.1117/12.894650.

  • National Research Council, 2007: Earth Science and Applications from Space: National Imperatives for the Next Decade and Beyond. National Academies Press, 456 pp.

  • Puschell, J. J., and Coauthors, 2003: Design and characterization of the Japanese Advanced Meteorological Imager (JAMI). Optical Spectroscopic Techniques and Instrumentation for Atmospheric and Space Research V, A. M. Larar, J. A. Shaw, and Z. Sun, Eds., International Society for Optical Engineering (SPIE Proceedings, Vol. 5157), 58–74, doi:10.1117/12.509806.

  • Schmit, T. J., , Gunshor M. M. , , Menzel W. P. , , Gurka J. J. , , Li J. , , and Bachmeier A. S. , 2005: Introducing the next-generation Advanced Baseline Imager on GOES-R. Bull. Amer. Meteor. Soc., 86, 10791096.

    • Search Google Scholar
    • Export Citation
  • Stamnes, K., , Tsay S.-C. , , Wiscombe W. , , and Jayaweera K. , 1988: Numerically stable algorithm for discrete-ordinate-method radiative transfer in multiple scattering and emitting layered media. Appl. Opt.,27, 25022509.

  • Suarez, M. J., Ed., 2005: Documentation and validation of the Goddard Earth Observing System (GEOS) data assimilation system—Version 4. Technical Report Series on Global Modeling and Data Assimilation, Vol. 26, NASA Tech. Memo. NASA/TM-2005-104606, 181 pp.

  • Sun, W., , Loeb N. G. , , Davies R. , , Loukachine K. , , and Miller W. F. , 2006: Comparison of MISR and CERES top-of-atmosphere albedo. Geophys. Res. Lett., 33, L23810, doi:10.1029/2006GL027958.

    • Search Google Scholar
    • Export Citation
  • Thompson, S. L., , and Warren S. G. , 1982: Parameterization of outgoing infrared radiation derived from detailed radiative calculations. J. Atmos. Sci., 39, 26672680.

    • Search Google Scholar
    • Export Citation
  • Viollier, M., , Kandel R. , , and Raberanto P. , 2004: Combination of ScaRaB-2 and CERES with Meteosat-5 to remove time sampling bias and improve radiation budget estimations in the INDOEX region. J. Geophys. Res., 109, D05105, doi:10.1029/2003JD003947.

    • Search Google Scholar
    • Export Citation
  • Weinreb, M. P., , Jamieson M. , , Fulton N. , , Chen Y. , , Johnson J. X. , , Bremer J. , , Smith C. , , and Baucom J. , 1997: Operational calibration of Geostationary Operational Environmental Satellite-8 and -9 imagers and sounders. Appl. Opt., 36, 68956904.

    • Search Google Scholar
    • Export Citation
  • Wielicki, B. A., , Barkstrom B. R. , , Harrison E. F. , , Lee R. B. III, , Smith G. L. , , and Cooper J. E. , 1996: Clouds and the Earth’s Radiant Energy System (CERES): An Earth Observing System experiment. Bull. Amer. Meteor. Soc., 77, 853868.

    • Search Google Scholar
    • Export Citation
  • Wu, A., , Xiong X. , , Doelling D. R. , , Morstad D. L. , , Angal A. , , and Bhatt R. , 2013: Characterization of Terra and Aqua MODIS VIS, NIR, and SWIR spectral band calibration stability. IEEE Trans. Geosci. Remote Sens., in press.

    • Search Google Scholar
    • Export Citation
  • Xiong, X., , Sun J. , , Wu A. , , Chiang K. , , Esposito J. , , and Barnes W. , 2005: Terra and Aqua MODIS calibration algorithms and uncertainty analysis. Sensors, Systems, and Next-Generation Satellites IX, R. Meynart, S. P. Neeck and H. Shimoda, Eds., International Society for Optical Engineering (SPIE Proceedings, Vol. 5978), doi:10.1117/12.627631.

  • Young, D. F., , Minnis P. , , Doelling D. R. , , Gibson G. G. , , and Wong T. , 1998: Temporal interpolation methods for the Clouds and the Earth’s Radiant Energy System (CERES) experiment. J. Appl. Meteor., 37, 572590.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Geostationary satellite timelines used in CERES processing listed by satellite longitude location.

  • View in gallery

    SW CERES geostationary temporal interpolation flowchart. NB refers to narrowband and BB refers to broadband.

  • View in gallery

    (a) GOES-12 visible counts (proportional to radiance) and Terra MODIS coincident ray-matched 0.5° latitude × 0.5° longitude gridded 2394 radiance pairs for August 2007. The red line is the linear regression through the space count of 29. (b) GOES-12 monthly gains from GOES-12/Terra MODIS ray-matching fit with a second-order regression. (c) As in (a), but for 2657 pairs of 11-μm temperature. The red line is the linear regression. (d) Monthly-mean temperature bias of GOES-12 minus Terra MODIS.

  • View in gallery

    GEO-derived BB-minus-CERES SW regional flux bias (a) coincident within an hour for January 2001 before normalization. (b) As in (a), but after SW regional normalization.

  • View in gallery

    (a) GEO-derived BB-minus-CERES SW flux bias coincident within an hour as a function of cloud amount, (b) SZA, and (c) view zenith angle before normalization for January 2001 over the Meteosat-7 (black), Meteosat-5 (green), GMS-5 (cyan), GOES-10 (red), and GOES-8 (purple) domains. (d)–(f) Corresponding biases after SW regional normalization.

  • View in gallery

    Terra CO-minus-CG for July 2002–June 2011 mean (a) SW and (b) LW fluxes. Aqua CO-minus-CG (c) SW and (d) LW fluxes. CO is CERES-only and CG is CERES-geostationary.

  • View in gallery

    (top) Monthly-mean RMS flux difference of CG Terra-minus-Aqua from July 2002 to June 2011. (bottom) RMS flux difference of CO Terra-minus-Aqua. (left) TOA SW flux. (right) TOA LW flux. Units: W m−2.

  • View in gallery

    Monthly hourly CERES and GERB flux comparison for January 2005. (left) SW comparisons and (right) LW comparisons. (top) Sahara Desert (30.5°N, 0.5°E), (middle) maritime stratus (20.5°S, 10.5°E), and (bottom) land convection (20.5°S, 20.5°E) region. Black, blue, and red lines represent Terra, Aqua, and Terra+Aqua datasets, respectively. Solid and dotted lines represent the CO and CG methods, respectively. GERB is the thick green line. GERB is truth for this comparison.

  • View in gallery

    (left) Aqua CO-minus-GERB for January 2005, (middle) Terra+Aqua CO-minus-GERB, and (right) Aqua CG-minus-GERB (top) SW and (bottom) LW fluxes. GERB is truth for this comparison.

  • View in gallery

    (left) Aqua CO-minus-GERB for January 2005, (middle) Terra+Aqua CO-minus-GERB, and (right) Aqua CG-minus-GERB (top) SW and (bottom) LW daily flux RMS errors. GERB is truth for this comparison.

  • View in gallery

    Terra CO from July 2002 to June 2011 (a) SW and (b) LW flux anomaly trends. Aqua CO from July 2002 to June 2011 (c) SW and (d) LW flux anomaly trends.

  • View in gallery

    Terra CG-minus-CO from July 2002 to June 2011 (a) SW and (b) LW flux anomaly trends. Aqua CG-minus-CO from July 2002 to June 2011 (c) SW and (d) LW flux anomaly trends.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 359 359 73
PDF Downloads 363 363 65

Geostationary Enhanced Temporal Interpolation for CERES Flux Products

View More View Less
  • 1 NASA Langley Research Center, Hampton, Virginia
  • | 2 SSAI, Hampton, Virginia
  • | 3 NASA Langley Research Center, Hampton, Virginia
  • | 4 SSAI, Hampton, Virginia
© Get Permissions
Full access

Abstract

The Clouds and the Earth’s Radiant Energy System (CERES) instruments on board the Terra and Aqua spacecraft continue to provide an unprecedented global climate record of the earth’s top-of-atmosphere (TOA) energy budget since March 2000. A critical step in determining accurate daily averaged flux involves estimating the flux between CERES Terra or Aqua overpass times. CERES employs the CERES-only (CO) and the CERES geostationary (CG) temporal interpolation methods. The CO method assumes that the cloud properties at the time of the CERES observation remain constant and that it only accounts for changes in albedo with solar zenith angle and diurnal land heating, by assuming a shape for unresolved changes in the diurnal cycle. The CG method enhances the CERES data by explicitly accounting for changes in cloud and radiation between CERES observation times using 3-hourly imager data from five geostationary (GEO) satellites. To maintain calibration traceability, GEO radiances are calibrated against Moderate Resolution Imaging Spectroradiometer (MODIS) and the derived GEO fluxes are normalized to the CERES measurements. While the regional (1° latitude × 1° longitude) monthly-mean difference between the CG and CO methods can exceed 25 W m−2 over marine stratus and land convection, these regional biases nearly cancel in the global mean. The regional monthly CG shortwave (SW) and longwave (LW) flux uncertainty is reduced by 20%, whereas the daily uncertainty is reduced by 50% and 20%, respectively, over the CO method, based on comparisons with 15-min Geostationary Earth Radiation Budget (GERB) data.

Corresponding author address: David Doelling, NASA Langley Research Center, Mail Stop 420, Hampton, VA 23681. E-mail: david.r.doelling@nasa.gov

Abstract

The Clouds and the Earth’s Radiant Energy System (CERES) instruments on board the Terra and Aqua spacecraft continue to provide an unprecedented global climate record of the earth’s top-of-atmosphere (TOA) energy budget since March 2000. A critical step in determining accurate daily averaged flux involves estimating the flux between CERES Terra or Aqua overpass times. CERES employs the CERES-only (CO) and the CERES geostationary (CG) temporal interpolation methods. The CO method assumes that the cloud properties at the time of the CERES observation remain constant and that it only accounts for changes in albedo with solar zenith angle and diurnal land heating, by assuming a shape for unresolved changes in the diurnal cycle. The CG method enhances the CERES data by explicitly accounting for changes in cloud and radiation between CERES observation times using 3-hourly imager data from five geostationary (GEO) satellites. To maintain calibration traceability, GEO radiances are calibrated against Moderate Resolution Imaging Spectroradiometer (MODIS) and the derived GEO fluxes are normalized to the CERES measurements. While the regional (1° latitude × 1° longitude) monthly-mean difference between the CG and CO methods can exceed 25 W m−2 over marine stratus and land convection, these regional biases nearly cancel in the global mean. The regional monthly CG shortwave (SW) and longwave (LW) flux uncertainty is reduced by 20%, whereas the daily uncertainty is reduced by 50% and 20%, respectively, over the CO method, based on comparisons with 15-min Geostationary Earth Radiation Budget (GERB) data.

Corresponding author address: David Doelling, NASA Langley Research Center, Mail Stop 420, Hampton, VA 23681. E-mail: david.r.doelling@nasa.gov

1. Introduction

The Clouds and the Earth’s Radiant Energy System (CERES) mission (Wielicki et al. 1996) is a National Aeronautics and Space Administration (NASA) satellite project dedicated to observe the earth’s top-of-atmosphere (TOA) global energy budget and estimate surface and within atmosphere radiation budgets as well. The first CERES instrument was launched in 1997 aboard the Tropical Rainfall Measuring Mission (TRMM) satellite into a 35° precessing low-latitude orbit and cycled through 24 local hours every 46 days (Wielicki et al. 1996). Two sets of CERES instruments on the Terra (launched in 1999) and Aqua (launched in 2002) satellites are in sun-synchronous orbits with equator crossing times of 1030 and 1330 local time (LT), respectively. In October 2011, a sixth CERES instrument was launched aboard the Suomi National Polar-Orbiting Partnership (NPP) also in a 1330 LT sun-synchronous orbit. While the TRMM CERES instrument ceased to operate after March of 2000, Terra CERES and Aqua CERES instruments continue to provide over a decade of TOA global energy budget data that are critical to cloud and climate feedback studies.

CERES measures filtered broadband radiances in shortwave (SW) (0.3–5 μm), total (TOT) (0.3–200 μm), and window (WN) (8.1–11.8 μm) channels with a spatial resolution of 20 km at nadir (equivalent circle diameter) for both Terra and Aqua satellites. The filtered broadband radiances are converted to unfiltered broadband SW, outgoing longwave (LW), and WN radiances using the approach described in Loeb et al. (2001). The radiances are then converted to a radiative flux using empirical angular directional models (ADMs) (Loeb et al. 2003, 2005; Kato and Loeb 2005; Loeb et al. 2007), which are defined according to various surface, cloud, and atmospheric properties. Cloud property retrievals are determined from coincident Moderate Resolution Imaging Spectroradiometer (MODIS) imager radiances from Terra and Aqua using the methodology described in Minnis et al. (2011) and are not associated with the official Goddard MODIS cloud product. These CERES level 2 cloud and radiation footprint data are spatially averaged and temporally interpolated to all local hours to produce the final CERES level 3 data products. The current CERES level 3 data products provide observed TOA broadband reflected SW, LW, and downward net fluxes along with computed TOA and surface fluxes, consistent with the associated MODIS-derived cloud and aerosol properties at regional, zonal, and global spatial scales, at the 3-hourly, daily, and monthly temporal scales.

Young et al. (1998) provides a detailed description of the CERES prelaunch time–space averaging algorithm and the associated uncertainties using synthetic CERES data. The CERES temporal interpolation is performed using two methodologies to accommodate different science objectives. Between CERES observation times, the Single Scanner Footprint 1° latitude × 1° longitude gridded (SSF1deg) data product assumes cloud properties are invariant; SW TOA fluxes account for albedo changes with solar zenith angle assuming the scene at the CERES observation time remains constant. The LW TOA fluxes are determined using a half-sine fit over land, with a peak at local solar noon and using a constant nightly flux, and linear interpolation over ocean (Young et al. 1998). This methodology of time interpolation is referred to as the CERES-only (CO) temporal interpolation method and is similar to the Earth Radiation Budget Experiment (ERBE) temporal averaging algorithm (Young et al. 1998). In the synoptic 1° latitude × 1° longitude gridded (SYN1deg) data product, cloud and radiation changes between CERES observations are inferred from 3-hourly geostationary (GEO) visible and infrared imager measurements between 60°S and 60°N. This approach is called the CERES geostationary (CG) temporal interpolation method. While the CO temporal interpolation method remains relatively unchanged from that in Young et al. (1998), significant modifications to the CERES geostationary method have been implemented in the CERES processing to address calibration traceability and data quality issues. Additional validation activities for both temporal interpolation methods have been performed using actual CERES data and other available broadband datasets.

The goal of this paper is to describe and validate the postlaunch time interpolation techniques used to produce the TOA radiative fluxes in the CERES SSF1deg and SYN1deg level 3 gridded daily and monthly data products available at the CERES website (http://ceres.larc.nasa.gov). The organization of this paper is as follows. Section 2 provides a brief description of the observational datasets used in this study. This is followed by a step-by-step description of the two CERES temporal interpolation methods in section 3. Section 4 provides the validation study results. The conclusions of this study are given in section 5.

2. Observations

a. CERES

We use the level 3 CERES edition 2.6 SSF1deg-lite and SYN1deg-lite data products from Terra and Aqua from July 2002 to June 2011. Both of these data products contain temporally interpolated daily 1° equal-area, zonal, and global averages of TOA fluxes and imager-derived cloud and aerosol properties.

b. Geostationary imager measurements

The 3-hourly geostationary 8-km nominal-resolution full-disc visible and IR window data are obtained from the Man Computer Interactive Data Access System (McIDAS) (Lazzara et al. 1999) in a consistent data format and near-real-time data download. At any time, five geostationary satellites provide contiguous coverage between ±60° latitude. These satellites are located at a longitude of 135°W [Geostationary Operational Environmental Satellite-10 (GOES-10) and GOES-11], 75°W (GOES-8, GOES-12, and GOES-13), 0°E [Meteorological Satellite-7 (Meteosat-7), Meteosat-8, and Meteosat-9], ~60°E (Meteosat-5 and Meteosat-7), and ~140°E [Geostationary Meteorological Satellite-5 (GMS-5), GOES-9, Multi-Functional Transport Satellite-1R (MTSAT-1R and MTSAT-2)]. Figure 1 shows the timeline of geostationary satellites used in the CERES processing for each of the 13 GEO satellites listed above. The imagery used in the analysis are the full-disc images collected at the 3-hourly intervals that are closest to being centered at 0000 UTC. An individual GEO satellite longitude domain is bounded by the bisecting longitude.

Fig. 1.
Fig. 1.

Geostationary satellite timelines used in CERES processing listed by satellite longitude location.

Citation: Journal of Atmospheric and Oceanic Technology 30, 6; 10.1175/JTECH-D-12-00136.1

c. Geostationary Earth Radiation Budget (GERB)

GERB instruments (Harries et al. 2005) aboard the Meteosat-8 and Meteosat-9 satellites provide top-of-atmosphere broadband SW and LW TOA fluxes every 15 min over the Meteosat domain (60°S–60°N, 60°W–60°E). The GERB record began in 2004 with GERB-2, followed by GERB-1 in 2007. GERB SW and LW broadband fluxes have a nominal footprint resolution of 50 km. In this study, GERB-2 edition 1 average rectified grid (ARG) footprint fluxes for January 2005 are averaged into the CERES spatial grid (Table 1).

Table 1.

Latitude and longitude intervals for CERES equal-area grid.

Table 1.

3. CERES temporal interpolation

The goal of the CERES temporal interpolation algorithm is to determine the 24-h average fluxes, accounting for the diurnal changes in radiation, with limited temporal sampling. This is especially important for data collected from sun-synchronous orbit where satellite observations are locked into specific local times. For example, the Terra and Aqua observations at the equator are only available at 1030/2230 and 1330/0130 LT each day, respectively. The CERES temporal interpolation algorithm provides a way to fill in the missing local times between CERES observation times in order to provide a diurnally complete representation of the cloud and radiation field prior to temporally averaging. This is especially critical in the SW, due to the strong diurnal cycle of solar insolation.

a. CO temporal interpolation method

In the CO temporal interpolation method, MODIS imager–derived cloud properties at the CERES observation times are assumed invariant throughout the day. Temporal interpolation of TOA fluxes follows an approach similar to that used in the ERBE experiment (Young et al. 1998). Between CERES observations, reflected SW TOA fluxes are inferred by accounting for changes in the TOA solar insolation and albedo dependence on the solar zenith angle, assuming the scene observed at the CERES observation time remains invariant. We use empirical scene–dependent diurnal albedo models for 600 different scene types, defined as a function of surface type, wind speed (clear ocean), and imager-derived cloud fraction, cloud phase, and cloud optical depth (Loeb et al. 2003). In nonpolar regions, the diurnal albedo models are based on CERES TRMM measurements, as TRMM samples the full diurnal cycle due to its precessing orbit and are the same models used in the CERES TRMM SSF product. In polar regions, diurnal albedo models from CERES Terra and Aqua are used (Kato and Loeb 2005). To account for the increase in the SW flux due to regional refracted light contributions that occurs when the solar zenith is angle is greater than 90°, an a priori twilight flux is added to the daily SW flux. The monthly regional twilight flux can be as much as 1.0 W m−2 over polar regions, but the global SW flux is increased by only 0.2 W m−2 (Kato and Loeb 2003). To compute the daily mean LW flux, linear interpolation between LW measurements is assumed, except over land, where a half-sine model, with a noon peak, is used to account for diurnal heating (Young et al. 1998). This provides hourly fluxes at the regional level, which can be averaged into daily and monthly means.

b. CG temporal interpolation method

To explicitly account for cloud and radiation changes between CERES observation times, the CERES observations are supplemented by five geostationary imagers around the globe that provide 3-hourly visible and infrared radiances. While the current versions of CERES data products (editions 2 and 3) incorporate 3-hourly GEO data, it is anticipated that the future version (edition 4) will use 1-hourly GEO data. Because the GEO imager channels are narrow band and have a greater radiometric uncertainty than MODIS and CERES, fusion of GEO imager data with CERES and MODIS involves unique challenges. For example, the GEO visible channels are uncalibrated and their sensitivity degrades with time. Further, the channel bandwidths, scan schedules, equatorial satellite positions, navigation quality, and sensor noise are unique to each satellite. If they are not identified and removed, then these limitations can cause temporal and spatial artifacts across the GEO satellite spatial domains. It is expected that hourly GEO sampling and the use of future next-generation GEO satellites will reduce these artifacts.

To minimize the GEO artifacts, several steps are performed (Fig. 2). First, the GEO imager radiances are intercalibrated against the well-calibrated MODIS 0.65-μm radiances. Monthly linear regressions of coincident ray-matched MODIS and GEO radiances within 0.5° latitude × 0.5° longitude regions determine the adjustments needed so that the GEO radiances equal the MODIS radiances. Next, an 8-km nominal-pixel-level cloud retrieval algorithm similar to that used by the CERES team to derive cloud properties from MODIS data is applied to determine 3-hourly GEO cloud properties. The GEO visible and infrared instantaneous 1° latitude × 1° longitude gridded radiances are then converted to broadband SW and LW TOA fluxes using observation-based narrowband (NB)-to-broadband (BB) radiance and radiance-to-flux conversion algorithms. This provides a 3-hourly gridded regional representation of SW and LW radiative flux changes. These are then temporally interpolated into hourly fluxes using the same diurnal albedo models and twilight flux correction in the SW and linearly interpolated in the LW. At this point, the GEO-derived broadband fluxes contain some residual regional and angular dependencies, when compared with instantaneous regional CERES-observed fluxes, which could bias the overall monthly-mean flux. To maintain the CERES instrument calibration and to mitigate any GEO artifacts resulting from this procedure, the GEO-based broadband fluxes are then carefully normalized to CERES fluxes using coincident measurements. Last, the GEO calibration is artificially modified to test the normalization technique. The following subsections summarize each of the key steps involved in merging the GEO data into the CERES processing stream.

Fig. 2.
Fig. 2.

SW CERES geostationary temporal interpolation flowchart. NB refers to narrowband and BB refers to broadband.

Citation: Journal of Atmospheric and Oceanic Technology 30, 6; 10.1175/JTECH-D-12-00136.1

Outside of the GEO domain, poleward of 60° in latitude, the CG fluxes are identical to the CO fluxes. Given that over the polar domain, a sun-synchronous satellite has multiple regional overpass times, up to 14 times per day, suggests that the CO method sufficiently captures the polar diurnal variability.

1) Geostationary calibration

Because GEO imagers have onboard blackbody calibration sources, the infrared channels are well calibrated (to within 0.5 K) and are temporally stable (Minnis et al. 2002b). In contrast, the GEO imager visible channels do not have onboard calibration and therefore must be continually vicariously calibrated, as the sensors degrade with time. It is projected that the next-generation GEO satellites will have onboard visible calibration (Schmit et al. 2005).

Following an initial quality control check of the GEO imager data by visually inspecting images and removing bad scan lines, the GEO visible (0.65 μm) and IR window (11 μm) channels are calibrated against MODIS Terra bands 1 (0.65 μm) and 31 (11.0 μm) using a similar ray-matching technique as Minnis et al. (2002a). MODIS has an on board solar diffuser and a solar diffuser monitor, which provide excellent visible channel stability (Minnis et al. 2008). MODIS absolute calibration uncertainty is 2% (1σ) (Xiong et al. 2005). Coincident MODIS and GEO measurements within 15 min of each other are averaged onto a 0.5° latitude × 0.5° longitude grid. Only bore-sighted or ray-matched GEO and MODIS gridded radiance pairs are linearly regressed on a monthly basis. To minimize errors due to differences between GEO and MODIS visible channel spectral response functions, only glint-free ocean scenes are considered. Outliers are removed by applying an additional spatial uniformity threshold test. In the visible, the GEO count is regressed against MODIS radiance, while GEO and MODIS infrared brightness temperatures are regressed against one another.

Figure 3a shows the GOES-12 visible count and MODIS Terra radiance regression for August 2007. The space count is 29 and the gain derived from linear regression of 2400 matched pairs is 0.84 W m−2 sr−1 μm−1 per count, with a standard error of ~4%. The visible GOES-12 monthly gains are then plotted over the lifetime of the satellite (Fig. 3b), and the GEO visible channel degradation rate is derived. The standard error for the second-order regression is 0.5%, and the correlation coefficient is 0.99. We find that every GEO instrument has its own degradation rate. The GOES-8–12 imagers typically degrade rapidly initially and then level off (Fig. 3b). The other GEO satellites degrade more linearly with time.

Fig. 3.
Fig. 3.

(a) GOES-12 visible counts (proportional to radiance) and Terra MODIS coincident ray-matched 0.5° latitude × 0.5° longitude gridded 2394 radiance pairs for August 2007. The red line is the linear regression through the space count of 29. (b) GOES-12 monthly gains from GOES-12/Terra MODIS ray-matching fit with a second-order regression. (c) As in (a), but for 2657 pairs of 11-μm temperature. The red line is the linear regression. (d) Monthly-mean temperature bias of GOES-12 minus Terra MODIS.

Citation: Journal of Atmospheric and Oceanic Technology 30, 6; 10.1175/JTECH-D-12-00136.1

The ray-matching calibration technique is validated using a multitiered approach (Morstad et al. 2011) involving both Terra MODIS/GEO and Aqua MODIS/GEO ray-matched radiances, deep convective clouds, and desert targets. All methods use Aqua MODIS 0.65-μm band as the calibration reference, as it has been demonstrated to be more stable and better characterized than Terra MODIS (Wu et al. 2013). For ray-matching purposes only, each method applies a spectral band adjustment factor based on Scanning Imaging Absorption Spectrometer for Atmospheric Chartography (SCIAMACHY) hyperspectral radiances (Doelling et al. 2012). Results show that for GOES-13, the absolute calibration difference among four independent calibration transfer methods is 1.4%. The maximum monthly standard error of all methods assuming a linear trend over 15 months is 0.7%, providing confidence in the stability of the GEO radiances.

The IR gridded radiances are regressed monthly in units of temperature. Separate analyses are performed for daytime and nighttime conditions. Figure 3c shows the daytime GOES-12 IR linear regression for August 2007. In this case, the standard error is 0.1%. The MODIS and GEO temperature bias is due to both spectral response function and calibration differences. Figure 3d shows the GOES-12 and MODIS Terra bias at the 300-K reference temperature for both daytime and nighttime. The day, night, and seasonal bias cycle is indicative of the variation in humidity and cloud properties impacting the resulting radiance from the sensor spectral response function differences. For CERES editions 2 and 3, the GEO cloud property retrieval code expects MODIS-like radiances with no spectral response function adjustment factor applied. For future releases a multichannel GEO cloud code will take into account the individual GEO channel spectral response functions, thereby requiring spectral-response-function-adjusted radiances that have been calibrated against the reference imager.

All of the CERES-era geostationary imager calibrations show a linear response when compared with MODIS, except the visible imager on board MTSAT-1R, located at 140°E and operational between 2005 and 2010. The MTSAT-1R was built to replace the MTSAT-1, where the rocket malfunctioned during launch in 1999. After the failure of the existing GMS-5 satellite in 2003, the GOES-9 satellite was moved to 160°E until MTSAT-1R was operational. MSTAT-1R has a very different design (Puschell et al. 2003) compared to other GEOs. It was built with 336 visible 1D detectors and an off-axis telescope with a two focal plane design, borrowing from the technology of MODIS. By comparison, the GOES imager utilizes eight visible detectors with a simple east–west scan (Weinreb et al. 1997). The follow-on MTSAT-2 imager has a similar design as GOES. The MTSAT-1R visible imager overestimates dark scenes that neighbor very bright clouds, as evident from visual inspection of coincident MTSAT-1R and MTSAT-2 imager radiance differences. Large areas of clear sky are not impacted. Comparisons against the Visible and Infrared Scanner (VIRS) imager on board the TRMM satellite shows the effect to also depend upon solar zenith angle (SZA). To compensate for this, we apply a SZA-dependent nonlinear correction, which improves the quality of visible radiances over the linear approach, but it does not remove all spurious artifacts over the MTSAT domain. In contrast, the MTSAT-1R IR imager showed excellent linear temperature scaling with MODIS.

2) Geostationary cloud retrievals

CERES GEO cloud properties are derived from a two-channel (visible 0.65 and IR 11 μm) retrieval algorithm, similar to the NASA Langley Layered Bispectral Threshold Model (LBTM) cloud retrieval algorithm (Minnis et al. 1994) over the GEO domain at an 8-km nominal pixel resolution every 3 h. While a multichannel cloud retrieval algorithm would be preferable, there are still some GEOs that only have a single visible and IR window channel. Therefore, it was decided that for CERES editions 2 and 3 products, a two-channel code would be used to ensure consistent GEO cloud properties throughout the CERES record. Meteorological data and skin temperatures in the GEO cloud retrieval algorithm are from the NASA Global Modeling and Assimilation Office’s (GMAO’s) Goddard Earth Observing System Data Assimilation System, level 4 (GEOS-4), product (Suarez 2005). The GEO cloud algorithm uses the monthly MODIS-derived clear-sky reflectance and surface emissivity maps that were generated during the CERES SSF MODIS pixel-level cloud retrieval processing. The daytime cloud retrieval parameterization assumes an effective particle radius of 10- and 30-μm radii for water and ice, respectively. The GEO phase is determined from the cloud radiative (effective) center temperature (Minnis et al. 1994) with a threshold of 253 K separating liquid and ice clouds. During the daytime, cloud height is determined using the cloud center temperature adjusted by the IR emissivity for thin clouds. At night, all clouds are considered to be black and therefore no cloud height corrections are performed. Because the GEO cloud algorithm is best suited for cloud retrievals from bright clouds over a dark surface, spurious GEO-derived cloud properties are possible over snow-covered regions where there is little radiative contrast between clear and cloudy conditions.

For each 3-hourly GEO image and for each CERES orbit, the pixel-level cloud retrievals are gridded into 1° latitude × 1° longitude regions and stratified by four static pressure layers, separated at 700-, 500-, and 300-hpa levels. For each region, all MODIS and GEO cloud measurements over the month are placed into hour boxes—MODIS taking precedence if both fall within the same hour box— and are temporally interpolated between measurements independently within each of the four layers. The cloud properties are linearly interpolated between cloudy measurements to fill the remaining hour boxes. However, they are not interpolated across clear-sky events, but are considered constant from the cloud measurement to the clear-sky event. The liquid and ice water path are not temporally interpolated but computed from the interpolated optical depth and particle size to be consistent with the observed MODIS cloud retrieval algorithm water paths. Daily and monthly-mean cloud properties are determined from the cloud fraction–weighted means from the corresponding hour boxes.

The time series of GEO monthly-mean cloud properties is compared with MODIS in order to check for any GEO calibration anomalies. A drift in the GEO calibration can be misinterpreted as temporal changes in cloud properties. The 9-yr global CG cloud amount mean was 65.3% and 65.1% for Terra and Aqua, respectively, and the corresponding CO is 63.3% and 62.1%, respectively. The sampling of clouds over the diurnal cycle has made the Terra- and Aqua-based cloud amounts more consistent. The standard deviation of the Terra-minus-Aqua CG monthly cloud difference was 0.8% and 0.9%, respectively, verifying that the monthly global cloud amounts closely track one another and that GEO clouds are not trending away from MODIS.

3) Geostationary narrowband-to-broadband SW flux conversion

The GEO visible radiances are first converted to MODIS-like band 1 radiances by applying angular and scene-dependent theoretically derived spectral correction coefficients that account for differences between GEO visible and MODIS band 1 spectral response functions. The radiative transfer model calculations are based upon discrete ordinate radiative transfer (DISORT) (Stamnes et al. 1998) for the scene and angular convention used for the CERES TRMM ADMs (Loeb et al. 2003). The clear radiances are based on a midlatitude summer over six surface types. Overcast radiances assumed a water cloud layer between 1 and 3 km, and an ice cloud between 5 and 7 km to define the liquid water and ice cloud scenes. For overcast scenes, the radiance conversion is derived through a linear regression of simulated GEO visible and MODIS band 1 radiance pairs derived from eight optical depth increments. Partly cloudy radiances are a combination of clear-sky and overcast radiances weighted by cloud fraction. The regression coefficients are invariant with season over each GEO domain.

GEO broadband SW radiances are derived from the MODIS-like band 1 GEO radiances by applying empirical relationships developed from instantaneous Terra CERES and MODIS band 1 radiances. This technique was also used to convert Multiangle Imaging Spectroradiometer (MISR) 0.67- and 0.87-μm radiances to broadband albedo in Sun et al. (2006). They found the RMS flux error to be <3% for SZA less than 80°. The GEO narrowband-to-broadband conversions are extended beyond the sun–Earth–satellite viewing domain observed by Terra using radiances obtained from a radiative transfer model. Following Loeb et al. (2003), these calculated radiances are scaled so that their average for the domain observed by the Terra MODIS matches the average of the MODIS radiances. Once the GEO imager visible regional radiances are converted into broadband radiances, the CERES TRMM ADMs (Loeb et al. 2003) are used to convert the broadband radiances into broadband fluxes using the GEO-derived cloud properties for scene identification.

4) Geostationary narrowband-to-broadband LW flux conversion

The GEO IR (11 μm) radiance to broadband LW flux conversion is described in Young et al. (1998). The IR radiance is first converted to a narrowband flux using a limb-darkening function that depends only on viewing zenith angle. The narrowband flux is then converted to a broadband flux using a quadratic relationship with a column-weighted relative humidity (colRH) term to account for the water vapor absorption not observed by the GEO window channel (Thompson and Warren 1982). The colRH, which increases the weight of the upper-tropospheric humidity, is the height-mean RH above the emitting surface based on the GEOS-4 atmospheric profiles. The global coefficients are based upon coincident MODIS band 31 (11 μm) and CERES LW footprint data, stratified according to ocean and land. Since the coefficients show little seasonal dependence, only a single set of coefficients is used. While this approach works adequately for most regions, a small systematic day–night bias is observed over desert. The instantaneous GEO SW and LW flux estimates have an RMS error of 10%–15% and 3%–5%, respectively, depending on GEO satellite. Similarly, Viollier et al. (2004) observed an RMS error of 15%–20% and 6%–8% between Scanner for Radiation Budget (ScaRaB) or CERES and Meteosat-5 in the SW and LW, respectively.

5) Geostationary LW flux normalization with CERES

Unlike the SW normalization, the GEO LW fluxes are instantaneously normalized. The gridded instantaneous 3-hourly GEO LW measurements are first placed into their respective hour boxes during the month and linearly interpolated to fill in the nonobserved hour boxes. For every CERES-observed hour-box flux, a CERES/GEO flux ratio is computed and the GEO flux is multiplied by the ratio in order to return the CERES-observed flux. In between CERES observations, the GEO fluxes and CERES/GEO flux ratios are linearly weighted according to their time difference from the two nearest CERES-observed hour-box times. This procedure ensures the GEO LW observations are on the same radiometric scale as CERES.

For clear-sky LW, it was discovered that the global monthly-mean clear-sky LW flux was ~2 W m−2 lower than the corresponding LW derived from the CERES-only method (section 3a). Further investigation revealed that the daytime GEO cloud retrieval algorithm occasionally misidentifies dark cold pixels as clear sky, which usually only occurs in cloud shadows over ocean surfaces. In very cloudy regions, where the correctly identified clear-sky observations are sparse, even a few cold pixels misidentified as clear sky can have a significant regional impact. Over land, the GEO clear-sky mask is also affected by topography. For this reason, the CERES SYN1deg-lite product uses the same clear-sky LW TOA flux values as those produced from the CERES SSF1deg product. The clear-sky LW monthly-mean (CO based) flux may not be diurnally consistent with the all-sky (CG based) flux over mostly clear land regions in the SYN1deg-lite product.

6) Geostationary SW flux normalization with CERES

The GEO SW-derived broadband fluxes were compared with coincident CERES fluxes and systematic biases were found both regionally (Fig. 4a) and as a function of cloud amount and sun–Earth–satellite viewing geometry (Figs. 5a–c), indicating the need to remove these artifacts. Initially, the LW instantaneous normalization approach was attempted for SW. However, rapid changes in clouds, such as passing frontal boundaries or during convective instability, within an hour in a given 1° latitude × 1° longitude region can cause significant SW TOA flux errors. For example, if an overcast GEO flux observation and a clear-sky CERES flux observation occur in the same hour, the instantaneous normalization approach produces an unrealistic albedo.

Fig. 4.
Fig. 4.

GEO-derived BB-minus-CERES SW regional flux bias (a) coincident within an hour for January 2001 before normalization. (b) As in (a), but after SW regional normalization.

Citation: Journal of Atmospheric and Oceanic Technology 30, 6; 10.1175/JTECH-D-12-00136.1

Fig. 5.
Fig. 5.

(a) GEO-derived BB-minus-CERES SW flux bias coincident within an hour as a function of cloud amount, (b) SZA, and (c) view zenith angle before normalization for January 2001 over the Meteosat-7 (black), Meteosat-5 (green), GMS-5 (cyan), GOES-10 (red), and GOES-8 (purple) domains. (d)–(f) Corresponding biases after SW regional normalization.

Citation: Journal of Atmospheric and Oceanic Technology 30, 6; 10.1175/JTECH-D-12-00136.1

These rapid changes in clouds produced erroneous albedos. Instead, the current approach uses a regional monthly normalization technique. Regional coincident within 90-min GEO and CERES SW flux pairs within a moving 5° × 5° domain are regressed in order to pair all CERES observations with 3-hourly GEO fluxes. Only regions free of sun glint with the same GEO domain and TRMM ADM surface type are included in the regression. Regions with insufficient sampling (<50) from a 5° × 5° domain use matches from all regions inside a 5° latitude zone and mainly occur over glint regions. A single slope and offset factor for each region is then used to normalize all of the regional GEO-derived SW measurements over the month before they are temporally interpolated using the same TRMM-based diurnal albedo models used in the CO approach. In this case the hourly interpolated cloud properties select the proper diurnal albedo model. To validate the GEO SW normalization, the GEO and CERES coincident fluxes are compared again. Most of the monthly 1° × 1° regional biases are less than 3%, except for those in which the CERES/GEO normalization time difference approaches 1.5 h (Fig. 4b). This represents the total narrowband to broadband, ADM, and regional normalization error budget. The SW regional normalization ensures that the monthly 5° × 5° regional bias is zero. In addition, cloud and viewing-geometry-dependent biases are significantly reduced (Figs. 5d–f).

As in the case for clear-sky LW TOA flux, GEO-derived clear-sky SW TOA fluxes are unreliable over most surface land types. For example, while SW TOA clear-sky flux is expected to be approximately symmetric about noon, this is not always observed. This is partly due to cloud contamination of the clear-sky identified pixels, which increases the clear-sky albedo. The GEO clear-sky visible threshold is based on the MODIS band 1 monthly regional clear-sky albedo multiplied by a static a priori land adjustment factor specific to each GEO satellite, which takes into account the mean GEO domain land surface spectral signature as a function of the GEO visible sensor spectral response. Over oceans the GEO clear-sky visible threshold is equivalent to the MODIS threshold, since the GEO radiances are calibrated against MODIS over oceans only. It is also partly due to the GEO narrowband-to-MODIS band 1 radiance conversion, since the simulated model radiances rely on a very limited number of land spectral surface reflectance models, which may not be representative of all regions. Both the adjustment factor and radiance conversion are less accurate because they do not properly account for the region’s unique broadband spectral signature. Thus, the CERES SYN1deg-lite product uses the same clear-sky SW TOA flux values as those produced in the CERES SSF1deg-lite product. Similarly, all-sky SW fluxes over regions with snow coverage greater than 10% for a given month are replaced by values in CERES SSF1deg.

7) GEO calibration sensitivity

To verify the robustness of the GEO broadband flux normalization technique, we artificially perturb the GEO visible and IR radiances by ±5% and repeat each of the steps described in the previous sections. This perturbation exceeds the uncertainty associated with intercalibrating GEO with MODIS, which is ~3% in the visible and ~1% in the IR. If the normalization is successful, then there should be no regional TOA flux difference between the perturbed and unperturbed fluxes. The regional SW RMS error in the recomputed GEO broadband fluxes resulting from the GEO visible radiance perturbations of ±5% is 0.70 W m−2, and the overall mean bias is 0.1 W m−2. In the LW, the regional RMS error and global bias remain within 0.1 W m−2. When the same analysis is performed after modifying the GEO IR calibration ±5%, the regional SW RMS flux error is 0.81 W m−2, and the global bias is within 0.1 W m−2. The LW RMS error and bias were less than 0.1 W m−2. Thus, the CERES/GEO normalization ensures that spurious GEO calibration errors do not impact the normalized CERES GEO broadband fluxes.

4. TOA flux validation

a. CG-minus-CO flux comparison

The 9-yr July 2002–June 2011 global annual mean CG and CO SW, LW, and net flux are shown in Table 2. Instrument calibration differences between CERES Terra and Aqua should not impact the comparison, as the Aqua CERES fluxes have been radiometrically scaled to Terra at the beginning of the Aqua mission. The Terra and Aqua CG global means are consistent, since they take into account the diurnal cycle. Not accounting for the diurnal cycle decreases the global mean SW flux by ~1% from either the Terra and Aqua sun-synchronous satellites. Including the diurnal cycle variations reduces the net radiation imbalance by 0.7 W m−2 for Terra and 1.2 W m−2 for Aqua. Loeb et al. (2009, 2012) provide an error analysis of the global mean fluxes and introduce the Energy Balance and Filled (EBAF) CERES product where the net flux has been constrained by the net imbalance derived from the Argo-based upper-ocean heating rate. Figure 6 shows the corresponding regional 9-yr mean Terra and Aqua CO-minus-CG SW flux difference. The marine stratus and land afternoon convective regions exhibit the strongest diurnal cycles and show the largest flux differences. Because marine stratus clouds exhibit a maximum cloud fraction in the morning and a minimum in the afternoon, the Terra CO marine stratus SW flux is overestimated. In contrast, because land convection peaks in the afternoon, the Terra CO SW flux is biased low in regions of land convection. The opposite holds true for the Aqua satellite, which has a 1330 LT sampling time. The LW flux minimum during peak convection over land is not captured by either Terra or Aqua. Since stratus cloud-top temperatures are similar to the underlying ocean, their diurnal cycle has a weak effect in the LW. The 9-yr CG and CO regional differences can be greater than 25 and 8 W m−2 in the SW and LW (Fig. 6), respectively, for regions that have large diurnal cycles. The regional RMS differences are 3.4 (1.1) W m−2 for SW (LW). However, these strong diurnal cycle regional flux differences cancel globally, as shown in Table 2.

Table 2.

July 2002–June 2011 global mean TOA fluxes as a function of satellite and temporal averaging method. CO is CERES-only, CG is CERES-geostationary.

Table 2.
Fig. 6.
Fig. 6.

Terra CO-minus-CG for July 2002–June 2011 mean (a) SW and (b) LW fluxes. Aqua CO-minus-CG (c) SW and (d) LW fluxes. CO is CERES-only and CG is CERES-geostationary.

Citation: Journal of Atmospheric and Oceanic Technology 30, 6; 10.1175/JTECH-D-12-00136.1

b. Terra-minus-Aqua flux comparisons

If the diurnal cycle were perfectly accounted for between CERES observations, then there would be no difference between the Terra- or Aqua-derived fluxes. Figure 7 displays the 9-yr monthly Terra-minus-Aqua monthly-mean RMS flux differences for CG and CO. The largest Terra-minus-Aqua CO RMS flux differences can exceed 30 W m−2 over the Southern Hemisphere, since the Terra and Aqua time difference is 3 h at the equator and 6 h at 60°S. Most of the CG SW and LW RMS flux differences are under 10 and 4 W m−2, respectively, even over the marine stratus, land convective, and ITCZ domains. The monthly Terra-minus-Aqua CG SW and LW RMS flux differences (Table 3) were ~50% and 30% smaller than the CO method, respectively. For daily means the Terra-minus-Aqua differences for CG are a factor of 4 smaller than for the CO method for both SW and LW. This comparison does not necessarily validate that the CG reflects the true diurnal cycle, but it shows that the Terra- and Aqua-based GEO fluxes are consistent with one another.

Fig. 7.
Fig. 7.

(top) Monthly-mean RMS flux difference of CG Terra-minus-Aqua from July 2002 to June 2011. (bottom) RMS flux difference of CO Terra-minus-Aqua. (left) TOA SW flux. (right) TOA LW flux. Units: W m−2.

Citation: Journal of Atmospheric and Oceanic Technology 30, 6; 10.1175/JTECH-D-12-00136.1

Table 3.

Terra-minus-Aqua during July 2002–June 2011 60°S–60°N domain monthly and daily regional RMS flux difference as a function of the temporal averaging method. CO is CERES-only, CG is CERES-geostationary.

Table 3.

Since both temporal interpolation methods are processed at the regional hourly level, the hourly Terra-based CG-interpolated fluxes can be compared directly with fluxes at CERES Aqua observation time. Keyes et al. (2006) computed the regional hourly (instantaneous) RMS difference between the Terra temporally interpolated flux minus the Aqua-observed flux and vice versa. The reduction in CG hourly RMS difference over CO interpolated was 57% in the SW and 43% in the LW.

c. GERB comparisons

The GERB instrument provides 15-min fluxes over the Meteosat domain (section 2c), which encompasses a wide range of climate regimes with diurnal variations that are representative of those observed over the globe. To compare the diurnal cycle between the two datasets, the CERES and GERB calibration must be consistent. It is assumed that the relative calibration of GERB and CERES is comparable and that only the absolute calibration difference needs to be removed. To normalize the GERB fluxes with the CERES fluxes, coincident within 7.5-min CERES and GERB flux pairs are regressed linearly to remove the absolute calibration difference between GERB and CERES. The linear regressions are performed for each 1° × 1° region in order to remove any scene-dependent spectral differences between CERES and GERB.

To illustrate the monthly hour improvement of the CG method, we compare CO, CG, and GERB fluxes for the Sahara Desert, marine stratus, and African land afternoon convection regions during January 2005 (Fig. 8). The normalized GERB fluxes are considered as truth. Both the CG and CO methods faithfully estimate the SW diurnal flux symmetry about noon. The CO SW fluxes over the stratus region for Aqua assume the clouds at the time of measurement are constant throughout the day and underestimate the morning flux. The opposite holds true over land afternoon convective regions. In contrast, the stratus and land CG SW fluxes closely follow the GERB observations, confirming that GEO fluxes and clouds are correctly sampling the diurnal cycle. Even the Terra+Aqua CO method does not capture the SW diurnal signal as well as the single-satellite CG method.

Fig. 8.
Fig. 8.

Monthly hourly CERES and GERB flux comparison for January 2005. (left) SW comparisons and (right) LW comparisons. (top) Sahara Desert (30.5°N, 0.5°E), (middle) maritime stratus (20.5°S, 10.5°E), and (bottom) land convection (20.5°S, 20.5°E) region. Black, blue, and red lines represent Terra, Aqua, and Terra+Aqua datasets, respectively. Solid and dotted lines represent the CO and CG methods, respectively. GERB is the thick green line. GERB is truth for this comparison.

Citation: Journal of Atmospheric and Oceanic Technology 30, 6; 10.1175/JTECH-D-12-00136.1

The GERB outgoing longwave radiation (OLR) indicates that the Sahara Desert solar heating maximum occurs 1–2 h after local solar noon. This is confirmed by overlaying the Terra and Aqua LW measurements (not shown). The Terra CO land half-sine fit assumes symmetry about noon and underestimates the LW daytime maxima, while the Aqua CO captures the LW maximum. The CG resembles the diurnal shape of the GERB LW flux, but all CG methods underestimate the morning flux. This is likely due to the improper application of global narrowband-to-broadband parameterization factors over dry desert conditions [section 3b(5)]. For the stratus and afternoon convection regions, the CG LW fluxes are in agreement with GERB. The CO linear interpolation and half-sine fit do not adequately resolve the diurnal cycle in these regions. Remarkably, all CO and CG methods, except Aqua CO, are within 1 W m−2 of the GERB LW monthly mean of 274.0 and 240.1 W m−2 for stratus and land convection, respectively. All CG SW and LW methods are very consistent compared with the corresponding CO methods.

To quantify the improvement of the CG over the CO methods over the GERB domain, the monthly, daily, and 3-hourly regional RMS differences are computed using the GERB fluxes as truth (Table 4). The corresponding regional bias and daily RMS errors are shown in Figs. 9 and 10, respectively. Figure 9 shows that the Aqua CO SW regional flux is underestimated over stratus and overestimated over land convection regions, similar to the examples in Fig. 8. The Aqua CG SW removes most of the regional biases in areas with strong diurnal cycles, but unfortunately it also introduces some artifacts over other regions, especially East Africa. The single-satellite monthly SW RMS error is reduced by ~20% using the CG method (Table 4). The dual-satellite case has reduced the SW RMS error over the single-satellite case by 50% and 25% for the CO and CG methods, respectively. The Terra+Aqua SW CO method slightly outperforms the CG method because the CG method still contains residual GEO artifacts and Terra and Aqua observe the peak of the diurnal solar cycle.

Table 4.

January 2005 dataset–minus–GERB flux domain bias, monthly, regional, and 3-hourly RMS error as a function of satellite and temporal averaging method.

Table 4.
Fig. 9.
Fig. 9.

(left) Aqua CO-minus-GERB for January 2005, (middle) Terra+Aqua CO-minus-GERB, and (right) Aqua CG-minus-GERB (top) SW and (bottom) LW fluxes. GERB is truth for this comparison.

Citation: Journal of Atmospheric and Oceanic Technology 30, 6; 10.1175/JTECH-D-12-00136.1

Fig. 10.
Fig. 10.

(left) Aqua CO-minus-GERB for January 2005, (middle) Terra+Aqua CO-minus-GERB, and (right) Aqua CG-minus-GERB (top) SW and (bottom) LW daily flux RMS errors. GERB is truth for this comparison.

Citation: Journal of Atmospheric and Oceanic Technology 30, 6; 10.1175/JTECH-D-12-00136.1

Figure 9 shows that the Aqua CO LW regional flux bias is positive over northern Africa and negative over land convective regions located in Brazil in southern Africa. These regional biases are mostly eliminated in the Aqua CO LW method, except over Brazil. This results in a ~20% reduction in the CG monthly regional LW RMS error over the CO method (Table 4). Since stratus regions do not have large diurnal cycle amplitudes, the monthly LW means are not sensitive to the temporal interpolation method, even though the LW CO method does not model the diurnal cycle accurately compared with GERB (Fig. 8). The dual-satellite CO and CG methods have similar monthly LW RMS errors, but after recomputing the LW RMS with the bias removed, there is a 10% reduction in the CG monthly LW RMS error over the CO method. Although the CO diurnal shape may not be correct according to Fig. 8, the hourly flux difference compensates when deriving the monthly mean, especially in the LW (see Fig. 9).

The greatest impact of the CG method is the reduction in SW and LW RMS errors for daily and 3-hourly fluxes. Compared to RMS errors from the CO method, we see 50% (50%) and ~20% (~40%) reductions for the CG method for SW and LW, respectively. For Terra+Aqua the daily and 3-hourly SW and LW RMS flux error reductions were ~25% (~40%) and 20% (~20%), respectively, compared to the CO method. The CO method benefits from using Terra+Aqua measurements, but it still underperforms a single-satellite-based CG daily and 3-hourly dataset. There is also at least a 10% reduction in the daily and 3-hourly CG SW or LW RMS error for the Terra+Aqua dataset over the single-satellite datasets.

d. 9-yr flux anomaly trends

Every effort is made to remove GEO artifacts in the CERES data products, for example, through intercalibration with MODIS and SW regional normalization. Residual GEO spurious signals do remain as noted in the GERB comparison section. Obtaining the correct diurnal cycle is critical to obtain the mean of the regional flux, but the added noise in the temporal trend is unavoidable owing to the poor quality of some GEO instruments. GEO satellite instruments were designed for weather forecasting, where imaging reliability is critical; they do not have well-calibrated data. Each GEO is unique in terms of channels, spectral bands, scan schedules, GEO satellite position, navigation quality, and sensor noise, with little coordination between international satellite centers.

Figure 11 shows 9-yr trends in the SW and LW TOA flux anomalies (CO method) for both Aqua and Terra. The anomalies are deseasonalized from regional monthly fluxes, and regressions are linear least squares fits through the data. The similarity between Terra and Aqua trends indicates that the diurnal cycle is a weak component of the resulting trend. The large trends are associated with the natural oscillations of the 9-yr dataset, such as from the El Niño–Southern Oscillation (ENSO), North Atlantic Oscillation (NAO), Arctic Oscillation, and others. The CG-minus-CO anomaly flux trends are shown in Fig. 12. The SW trend difference reveals meridional GEO artifacts for both the Terra and Aqua datasets, which are similar to the 3-hourly or 45°-longitude striping spatial patterns of the SW regional GEO normalization RMS flux error displayed in Fig. 4. The Fig. 4 color bar range highlights the trend difference. The LW trend differences are much smaller than for the SW, except for the hourly or 15°-longitude striping pattern mainly over desert regions, due to the discretization of the 3-hourly GEO data.

Fig. 11.
Fig. 11.

Terra CO from July 2002 to June 2011 (a) SW and (b) LW flux anomaly trends. Aqua CO from July 2002 to June 2011 (c) SW and (d) LW flux anomaly trends.

Citation: Journal of Atmospheric and Oceanic Technology 30, 6; 10.1175/JTECH-D-12-00136.1

Fig. 12.
Fig. 12.

Terra CG-minus-CO from July 2002 to June 2011 (a) SW and (b) LW flux anomaly trends. Aqua CG-minus-CO from July 2002 to June 2011 (c) SW and (d) LW flux anomaly trends.

Citation: Journal of Atmospheric and Oceanic Technology 30, 6; 10.1175/JTECH-D-12-00136.1

To estimate the GEO artifact noise in the CG trend over the CO trend, a ratio was computed by dividing the CG-minus-CO standard deviation with the CO standard deviation of all the regional trends over the GEO domain. The ratio was 0.3 and 0.1 for SW and LW, respectively. Zonally, the GEO noise ratio in the CG trend was 20% and 6% for SW and LW, respectively. Thus, the CG flux anomaly trends are influenced by GEO artifacts and do not add diurnal information in the long-term trend. For evaluation of long-term regional trends, the CO products are more reliable.

5. Conclusions

To compute accurate CERES level 3 spatially and temporally averaged TOA monthly and daily fluxes, temporal interpolation is required to remove errors that are tied to specific temporal sampling of the satellite orbit. CERES employs two temporal interpolation methods, the CERES-only (CO) and CERES geostationary (CG) methods. The CO method assumes the scene at the CERES observation time remains constant and models the diurnal cycle of flux using empirical models. The SW TOA fluxes account for albedo changes with solar zenith angle assuming the cloud properties are constant. The CO LW method is similar to ERBE, where the LW flux is linearly interpolated between measurements, except over land where the half-sine diurnal model is used.

The CG method employs 3-hourly GEO-derived broadband fluxes in addition to the CERES-observed fluxes to more accurately model the flux and cloud variability between CERES measurements. To maintain the CERES calibration, the GEO-derived broadband fluxes are carefully normalized to the CERES measurements using regional normalization. The robustness of the normalization technique was tested by perturbing the GEO imager IR and visible radiances by ±5%. The perturbed regional monthly RMS error remained less than 1 W m−2, and it did not impact the global mean flux. Note, that the GEO cloud properties are not normalized in the SYN1deg dataset.

The regional 9-yr mean flux difference between the CG and CO methods can be as large as 25 and 8 W m−2 in the SW and LW, respectively, in regions with strong diurnal cycles (e.g., marine stratus and land convection). Explicitly accounting for the diurnal cycle results in a 1% increase in the 9-yr TOA global mean SW flux, whether sampled from Terra or Aqua. The large regional differences have a compensating effect globally.

The 9-yr monthly regional 60°N–60°S Terra-minus-Aqua RMS flux difference for CG is reduced by 50% and 30% compared to CO for SW and LW, respectively. The daily Terra-minus-Aqua RMS error is reduced by a factor of 4, indicating that the Terra- or Aqua-based CG fluxes are diurnally consistent.

To validate the accuracy of the CERES CG diurnal cycle, comparisons with GERB 15-min broadband fluxes were performed. Since both temporal interpolation methods resolve fluxes hourly, the monthly hourly SW and LW flux differences between the two methods were compared over marine stratus, land afternoon convection, and deserts. Results clearly show that the CG fluxes capture the subtle diurnal variations in the GERB fluxes. The CG fluxes show smaller regional biases compared to CO fluxes, but CG fluxes also introduce artifacts in other regions. Although the CO diurnal shape may not be correct, the hourly flux errors largely cancel when deriving the monthly mean for the GERB domain, especially in the LW. Overall, the single-satellite regional monthly SW and LW RMS error was reduced by ~20% using the CG method compared to the CO method. The greatest benefit of GEO data is manifested in the higher-order temporal-resolution flux products, where the CG method reduces the daily and 3-hourly SW flux uncertainty by half, and the daily and 3-hourly LW uncertainty by 20% and 40%, respectively. There is also a 10% further improvement in the monthly regional CG flux uncertainty when combining Terra and Aqua compared to the single-satellite case.

The fusion of 3-hourly GEO fluxes into the CERES flux data stream does not come without consequences. Each of the 13 GEOs used in the 10-yr CG dataset have unique sensor characteristics with varying degrees of quality, which may have the potential for causing GEO artifacts. The 9-yr CG-minus-CO regional SW monthly anomaly trends show significant GEO artifacts, in some regions, resulting in unnatural spatial patterns. Users interested in long-term regional trends are advised to use the CO (SSF1deg) or EBAF product fluxes. The level 4 EBAF monthly product removes GEO artifacts by using diurnal corrections based on CG-to-CO flux ratios dependent on MODIS cloud retrievals and surface type (see Loeb et al. 2012 supplement; available online at http://www.nature.com/ngeo/journal/v5/n2/extref/ngeo1375-s1.pdf).

Increasing the GEO temporal resolution to 1-hourly should greatly reduce the artifacts associated with the GEO flux normalization with CERES and regional trend analysis, especially over the GOES-West (GOES-10 and GOES-11 at 135°W), GOES-East (GOES-8 and GOES-12 at 75°W), and Meteosat Second Generation (Meteosat-8 and Meteosat-9 at 0°E) GEO domains, which currently have the highest-quality GEO instruments and are successively replaced by nearly identical instruments. Combining Terra and Aqua observations also reduces GEO artifacts. It is expected that the GEO data quality will greatly improve when most satellite centers launch their third-generation GEO satellites beginning in 2015. These new GEO data will have higher spatial and temporal resolution and 16 MODIS-like imager channels with onboard calibration (Schmit et al. 2005). This will allow almost seamless integration of hourly GEO flux and cloud parameters across GEO domains and with either NPP or Joint Polar Satellite System (JPSS) CERES fluxes and imager cloud properties. It is anticipated that with the launch of the Climate Absolute Radiance and Refractivity Observatory (CLARREO) (National Research Council 2007) mission, extremely well-calibrated visible and IR hyperspectral radiances will be available to uniformly calibrate all GEO instruments.

Acknowledgments

This work has been funded by the NASA CERES project. The products and the validation could not have been accomplished without the help of the CERES TISA team. The authors also thank the CERES, GERB, Megha-Tropiques, and CLARREO science teams for their insightful temporal averaging discussions. These data were obtained from the NASA Langley Research Center EOSDIS Distributed Active Archive Center.

REFERENCES

  • Doelling, D. R., , Lukashin C. , , Minnis P. , , Scarino B. , , and Morstad D. , 2012: Spectral reflectance corrections for satellite intercalibrations using SCIAMACHY data. IEEEGeosci. Remote Sens. Lett, 9, 119123, doi:10.1109/LGRS.2011.2161751.

    • Search Google Scholar
    • Export Citation
  • Harries, J. E., and Coauthors, 2005: The Geostationary Earth Radiation Budget Project. Bull. Amer. Meteor. Soc., 86, 945960.

  • Kato, S., , and Loeb N. G. , 2003: Twilight irradiance reflected by the earth estimated from Clouds and the Earth’s Radiant Energy System (CERES) measurements. J. Climate, 16, 26462650.

    • Search Google Scholar
    • Export Citation
  • Kato, S., , and Loeb N. G. , 2005: Top-of-atmosphere shortwave broadband observed radiance and estimated irradiance over polar regions from Clouds and the Earth’s Radiant Energy System (CERES) instruments on Terra. J. Geophys. Res., 110, D07202, doi:10.1029/2004JD005308.

    • Search Google Scholar
    • Export Citation
  • Keyes, D. F., , Doelling D. R. , , Young D. F. , , Rose F. G. , , Rutan D. A. , , Nordeen M. L. , , and Boghosian J. S. , 2006: The validation of the 5-year Terra-based monthly CERES radiative flux and cloud product. Proc. 12th Conf. on Atmospheric Radiation, Madison, WI, Amer. Meteor. Soc., P3.1. [Available online at https://ams.confex.com/ams/Madison2006/techprogram/paper_112731.htm.]

  • Lazzara, M. A., and Coauthors, 1999: The Man Computer Interactive Data Access System: 25 years of interactive processing. Bull. Amer. Meteor. Soc., 80, 271284.

    • Search Google Scholar
    • Export Citation
  • Loeb, N. G., , Priestley K. J. , , Kratz D. P. , , Geier E. B. , , Green R. N. , , Wielicki B. A. , , Hinton P. O. R. , , and Nolan S. K. , 2001: Determination of unfiltered radiances from the Clouds and the Earth’s Radiant Energy System (CERES) instrument. J. Appl. Meteor., 40, 822835.

    • Search Google Scholar
    • Export Citation
  • Loeb, N. G., , Manalo-Smith N. , , Kato S. , , Miller W. F. , , Gupta S. K. , , Minnis P. , , and Wielicki B. A. , 2003: Angular distribution models for top-of-atmosphere radiative flux estimation from the Clouds and the Earth’s Radiant Energy System instrument on the Tropical Rainfall Measuring Mission Satellite. Part I: Methodology. J. Appl. Meteor., 42, 240265.

    • Search Google Scholar
    • Export Citation
  • Loeb, N. G., , Kato S. , , Loukachine K. , , and Smith N. M. , 2005: Angular distribution models for top-of-atmosphere radiative flux estimation from the Clouds and the Earth’s Radiant Energy System instrument on the Terra satellite. Part I: Methodology. J. Atmos. Oceanic Technol., 22, 338351.

    • Search Google Scholar
    • Export Citation
  • Loeb, N. G., , Kato S. , , Loukachine K. , , Smith N. M. , , and Doelling D. R. , 2007: Angular distribution models for top-of-atmosphere radiative flux estimation from the Clouds and the Earth’s Radiant Energy System instrument on the Terra satellite. Part II: Validation. J. Atmos. Oceanic Technol., 24, 564584.

    • Search Google Scholar
    • Export Citation
  • Loeb, N. G., , Wielicki B. A. , , Doelling D. R. , , Smith G. L. , , Keyes D. F. , , Kato S. , , Manalo-Smith N. , , and Wong T. , 2009: Toward optimal closure of the earth’s top-of-atmosphere radiation budget. J. Climate, 22, 748766.

    • Search Google Scholar
    • Export Citation
  • Loeb, N. G., , Lyman J. M. , , Johnson G. C. , , Allen R. P. , , Doelling D. R. , , Wong T. , , Soden B. J. , , and Stephens G. L. , 2012: Observed changes in top-of-the-atmosphere radiation and upper-ocean heating consistent within uncertainty. Nat. Geosci., 5, doi:10.1038/ngeo1375.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., , Smith W. L. Jr., , Garber D. P. , , Ayers J. K. , , and Doelling D. R. , 1994: Cloud properties derived from GOES-7 for spring 1994 ARM intensive observing period using version 1.0.0 of ARM Satellite Data Analysis Program. NASA Reference Publ. 1366, 61 pp.

  • Minnis, P., , Nguyen L. , , Doelling D. R. , , Young D. F. , , and Miller W. , 2002a: Rapid calibration of operational and research meteorological satellite imagers. Part I: Evaluation of research satellite visible channels as references. J. Atmos. Oceanic Technol., 19, 12331249.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., , Nguyen L. , , Doelling D. R. , , Young D. F. , , Miller W. , , and Kratz D. P. , 2002b: Rapid calibration of operational and research meteorological satellite imagers. Part II: Comparison of infrared channels. J. Atmos. Oceanic Technol., 19, 12501266.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., , Doelling D. R. , , Nguyen L. , , Miller W. , , and Chakrapani V. , 2008: Assessment of the visible channel calibrations of the TRMM VIRS and MODIS on Aqua and Terra. J. Atmos. Oceanic Technol., 25, 385400.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., and Coauthors, 2011: CERES edition-2 cloud property retrievals using TRMM VIRS and Terra and Aqua MODIS data—Part I: Algorithms. IEEE Trans. Geosci. Remote Sens., 49, 43744400, doi:10.1109/TGRS.2011.2144601.

    • Search Google Scholar
    • Export Citation
  • Morstad, D. L., , Doelling D. R. , , Bhatt R. , , and Scarino B. , 2011: The CERES calibration strategy of the geostationary visible channels for CERES clouds and flux products. Earth Observing Systems XVI, J. J. Butler, X. Xiong, and X. Gu, Eds., International Society for Optical Engineering (SPIE Proceedings, Vol. 8153), doi:10.1117/12.894650.

  • National Research Council, 2007: Earth Science and Applications from Space: National Imperatives for the Next Decade and Beyond. National Academies Press, 456 pp.

  • Puschell, J. J., and Coauthors, 2003: Design and characterization of the Japanese Advanced Meteorological Imager (JAMI). Optical Spectroscopic Techniques and Instrumentation for Atmospheric and Space Research V, A. M. Larar, J. A. Shaw, and Z. Sun, Eds., International Society for Optical Engineering (SPIE Proceedings, Vol. 5157), 58–74, doi:10.1117/12.509806.

  • Schmit, T. J., , Gunshor M. M. , , Menzel W. P. , , Gurka J. J. , , Li J. , , and Bachmeier A. S. , 2005: Introducing the next-generation Advanced Baseline Imager on GOES-R. Bull. Amer. Meteor. Soc., 86, 10791096.

    • Search Google Scholar
    • Export Citation
  • Stamnes, K., , Tsay S.-C. , , Wiscombe W. , , and Jayaweera K. , 1988: Numerically stable algorithm for discrete-ordinate-method radiative transfer in multiple scattering and emitting layered media. Appl. Opt.,27, 25022509.

  • Suarez, M. J., Ed., 2005: Documentation and validation of the Goddard Earth Observing System (GEOS) data assimilation system—Version 4. Technical Report Series on Global Modeling and Data Assimilation, Vol. 26, NASA Tech. Memo. NASA/TM-2005-104606, 181 pp.

  • Sun, W., , Loeb N. G. , , Davies R. , , Loukachine K. , , and Miller W. F. , 2006: Comparison of MISR and CERES top-of-atmosphere albedo. Geophys. Res. Lett., 33, L23810, doi:10.1029/2006GL027958.

    • Search Google Scholar
    • Export Citation
  • Thompson, S. L., , and Warren S. G. , 1982: Parameterization of outgoing infrared radiation derived from detailed radiative calculations. J. Atmos. Sci., 39, 26672680.

    • Search Google Scholar
    • Export Citation
  • Viollier, M., , Kandel R. , , and Raberanto P. , 2004: Combination of ScaRaB-2 and CERES with Meteosat-5 to remove time sampling bias and improve radiation budget estimations in the INDOEX region. J. Geophys. Res., 109, D05105, doi:10.1029/2003JD003947.

    • Search Google Scholar
    • Export Citation
  • Weinreb, M. P., , Jamieson M. , , Fulton N. , , Chen Y. , , Johnson J. X. , , Bremer J. , , Smith C. , , and Baucom J. , 1997: Operational calibration of Geostationary Operational Environmental Satellite-8 and -9 imagers and sounders. Appl. Opt., 36, 68956904.

    • Search Google Scholar
    • Export Citation
  • Wielicki, B. A., , Barkstrom B. R. , , Harrison E. F. , , Lee R. B. III, , Smith G. L. , , and Cooper J. E. , 1996: Clouds and the Earth’s Radiant Energy System (CERES): An Earth Observing System experiment. Bull. Amer. Meteor. Soc., 77, 853868.

    • Search Google Scholar
    • Export Citation
  • Wu, A., , Xiong X. , , Doelling D. R. , , Morstad D. L. , , Angal A. , , and Bhatt R. , 2013: Characterization of Terra and Aqua MODIS VIS, NIR, and SWIR spectral band calibration stability. IEEE Trans. Geosci. Remote Sens., in press.

    • Search Google Scholar
    • Export Citation
  • Xiong, X., , Sun J. , , Wu A. , , Chiang K. , , Esposito J. , , and Barnes W. , 2005: Terra and Aqua MODIS calibration algorithms and uncertainty analysis. Sensors, Systems, and Next-Generation Satellites IX, R. Meynart, S. P. Neeck and H. Shimoda, Eds., International Society for Optical Engineering (SPIE Proceedings, Vol. 5978), doi:10.1117/12.627631.

  • Young, D. F., , Minnis P. , , Doelling D. R. , , Gibson G. G. , , and Wong T. , 1998: Temporal interpolation methods for the Clouds and the Earth’s Radiant Energy System (CERES) experiment. J. Appl. Meteor., 37, 572590.

    • Search Google Scholar
    • Export Citation
Save