• Augustine, J. A., , DeLuisi J. J. , , and Long C. N. , 2000: SURFRAD – A national surface radiation budget network for atmospheric research. Bull. Amer. Meteor. Soc., 81, 23412358, doi:10.1175/1520-0477(2000)081<2341:SANSRB>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Barker, H. W., 1996: A parameterization for computing grid-averaged solar fluxes for inhomogeneous marine boundary layer clouds. Part I: Methodology and homogeneous biases. J. Atmos. Sci., 53, 22892303, doi:10.1175/1520-0469(1996)053<2289:APFCGA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Bergman, J. W., , and Salby M. L. , 1996: Diurnal variations of cloud cover and their relationship to climatological conditions. J. Climate, 9, 28022820, doi:10.1175/1520-0442(1996)009<2802:DVOCCA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Bloom, S., and et al. , 2005: Documentation and validation of the Goddard Earth Observing System (GEOS) Data Assimilation System—Version 4. M. J. Suarez, Ed., Technical Report Series on Global Modeling and Data Assimilation, Vol. 26, NASA Tech. Rep. NASA/TM-2005-104606, 187 pp.

  • Cairns, B., 1995: Diurnal variation of cloud from ISCCP data. Atmos. Res., 37, 133146, doi:10.1016/0169-8095(94)00074-N.

  • Collins, W. D., , Rasch P. J. , , Eaton B. E. , , Khattatov B. V. , , Lamarque J.-F. , , and Zender C. S. , 2001: Simulating aerosols using a chemical transport model with assimilation of satellite aerosol retrievals: Methodology for INDOEX. J. Geophys. Res., 106, 73137336, doi:10.1029/2000JD900507.

    • Search Google Scholar
    • Export Citation
  • Dee, D. P., and et al. , 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553597, doi:10.1002/qj.828.

    • Search Google Scholar
    • Export Citation
  • Dilley, A. C., , and O’Brien D. M. , 1998: Estimating downward clear sky long-wave irradiance at the surface from screen temperature and precipitable water. Quart. J. Roy. Meteor. Soc., 124, 13911401, doi:10.1002/qj.49712454903.

    • Search Google Scholar
    • Export Citation
  • Doelling, D. R., and et al. , 2013: Geostationary enhanced temporal interpolation for CERES flux products. J. Atmos. Oceanic Technol., 30, 10721090, doi:10.1175/JTECH-D-12-00136.1.

    • Search Google Scholar
    • Export Citation
  • Fu, Q., , and Liou K.-N. , 1993: Parameterization of the radiative properties of cirrus clouds. J. Atmos. Sci., 50, 20082025, doi:10.1175/1520-0469(1993)050<2008:POTRPO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Fu, Q., , Lesins G. , , Higgins J. , , Charlock T. , , Chylek P. , , and Michalsky J. , 1998: Broadband water vapor absorption of solar radiation tested using ARM data. Geophys. Res. Lett., 25, 11691172, doi:10.1029/98GL00846.

    • Search Google Scholar
    • Export Citation
  • Henderson, D. S., , L’Ecuyer T. , , Stephens G. , , Partain P. , , and Sekiguchi M. , 2013: A multisensor perspective on the radiative impacts of clouds and aerosols. J. Appl. Meteor. Climatol, 52, 853871, doi:10.1175/JAMC-D-12-025.1.

    • Search Google Scholar
    • Export Citation
  • Hess, M., , Koepke P. , , and Schult I. , 1998: Optical properties of aerosols and clouds: The software package OPAC. Bull. Amer. Meteor. Soc., 79, 831844, doi:10.1175/1520-0477(1998)079<0831:OPOAAC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Jin, Z., , Charlock T. P. , , Smith W. L. Jr., , and Rutledge K. , 2004: A parameterization of ocean surface albedo. Geophys. Res. Lett., 31, L22301, doi:10.1029/2004GL021180.

    • Search Google Scholar
    • Export Citation
  • Kato, S., , Ackerman T. , , Mather J. , , and Clothiaux E. , 1999: The k-distribution method and correlated-k approximation for a shortwave radiative transfer model. J. Quant. Spectrosc. Radiat. Transfer, 62, 109121, doi:10.1016/S0022-4073(98)00075-2.

    • Search Google Scholar
    • Export Citation
  • Kato, S., , Rose F. G. , , and Charlock T. P. , 2005: Computation of domain-averaged irradiance using satellite-derived cloud properties. J. Atmos. Oceanic Technol., 22, 146164, doi:10.1175/JTECH-1694.1.

    • Search Google Scholar
    • Export Citation
  • Kato, S., and et al. , 2011: Improvements of top‐of‐atmosphere and surface irradiance computations with CALIPSO‐, CloudSat‐, and MODIS‐derived cloud and aerosol properties. J. Geophys. Res., 116, D19209, doi:10.1029/2011JD016050.

    • Search Google Scholar
    • Export Citation
  • Kato, S., , Loeb N. G. , , Rose F. G. , , Doelling D. R. , , Rutan D. A. , , Caldwell T. E. , , Yu L. , , and Weller R. A. , 2013: Surface irradiances consistent with CERES-derived top-of-atmosphere shortwave and longwave irradiances. J. Climate,26, 2719–2740, doi:10.1175/JCLI-D-12-00436.1.

  • Kratz, D. P., , and Rose F. G. , 1999: Accounting for molecular absorption within the spectral range of the CERES window channel. J. Quant. Spectrosc. Radiat. Transfer, 61, 8395, doi:10.1016/S0022-4073(97)00203-3.

    • 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, doi:10.1175/2008JCLI2637.1.

    • Search Google Scholar
    • Export Citation
  • Loeb, N. G., , Kato S. , , Su W. , , Wong T. , , Rose F. G. , , Doelling D. R. , , Norris J. R. , , and Huang X. , 2012: Advances in understanding top-of-atmosphere radiation variability from satellite observations. Surv. Geophys., 33, 359–385, doi:10.1007/s10712-012-9175-1.

    • Search Google Scholar
    • Export Citation
  • Long, C. N., , and Ackerman T. P. , 2000: Identification of clear skies from broadband pyranometer measurements and calculation of downwelling shortwave cloud effects. J. Geophys. Res., 105, 15 60915 626, doi:10.1029/2000JD900077.

    • Search Google Scholar
    • Export Citation
  • McPhaden, M. J., 2002: TAO/TRITON tracks Pacific Ocean warming in early 2002. CLIVAR Exchanges, No. 24, International CLIVAR Project Office, Southampton, United Kingdom, 7–9.

  • McPhaden, M. J., and et al. , 2009: RAMA: The Research Moored Array for African–Asian–Australian Monsoon Analysis and Prediction. Bull. Amer. Meteor. Soc., 90, 459480, doi:10.1175/2008BAMS2608.1.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., , and Harrison E. F. , 1984: Diurnal variability of regional cloud and clear-sky radiative parameters derived from GOES data, Part II: November 1978 cloud results. J. Climate Appl. Meteor., 23, 10121031, doi:10.1175/1520-0450(1984)023<1012:DVORCA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., , Smith W. L. Jr., , Garber D. P. , , Ayers J. K. , , and Doelling D. R. , 1995: 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, 59 pp.

  • Minnis, P., and et al. , 2011a: 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
  • Minnis, P., and et al. , 2011b: CERES Edition-2 cloud property retrievals using TRMM VIRS and Terra and Aqua MODIS data—Part II: Examples of average results and comparisons with other data. IEEE Trans. Geosci. Remote Sens., 49, 4401–4430, 10.1109/TGRS.2011.2144602.

    • Search Google Scholar
    • Export Citation
  • Ohmura, A., and et al. , 1998: Baseline Surface Radiation Network (BSRN/WCRP): New precision radiometry for climate change research. Bull. Amer. Meteor. Soc., 79, 21152136, doi:10.1175/1520-0477(1998)079<2115:BSRNBW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Oreopoulos, L., , and Barker H. W. , 1999: Accounting for subgrid-scale cloud variability in a multi-layer 1D solar radiative transfer algorithm. Quart. J. Roy. Meteor. Soc., 125, 301330, doi:10.1002/qj.49712555316.

    • Search Google Scholar
    • Export Citation
  • Pinker, R. T., , and Laszlo I. , 1992: Modeling surface solar irradiance for satellite applications on a global scale. J. Appl. Meteor., 31, 194211, doi:10.1175/1520-0450(1992)031<0194:MSSIFS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Remer, L. A., and et al. , 2005: The MODIS aerosol algorithm, products, and validation. J. Atmos. Sci., 62, 947973, doi:10.1175/JAS3385.1.

    • Search Google Scholar
    • Export Citation
  • Rienecker, M. M., and et al. , 2011: MERRA: NASA's Modern-Era Retrospective Analysis for Research and Applications. J. Climate, 24, 36243648, doi:10.1175/JCLI-D-11-00015.1.

    • Search Google Scholar
    • Export Citation
  • Roemmich, D., , and Gilson J. , 2009: The 2004–2008 mean and annual cycle of temperature, salinity, and steric height in the global ocean from the Argo Program. Prog. Oceanogr.,82, 81–100, doi:10.1016/j.pocean.2009.03.004.

  • Rose, F. G., , Charlock T. , , Fu Q. , , Kato S. , , Rutan D. , , and Jin Z. , 2006: CERES proto-edition_3 radiative transfer: Model tests and radiative closure over surface validation sites. 12th Conf. on Atmospheric Radiation, Madison, WI, Amer. Meteor. Soc., P2.4. [Available online at https://ams.confex.com/ams/Madison2006/techprogram/paper_112358.htm.]

  • Rose, F. G., , Rutan D. A. , , Charlock T. , , Smith G. L. , , and Kato S. , 2013: An algorithm for the constraining of radiative transfer calculations to CERES-observed broadband top-of-atmosphere irradiance. J. Atmos. Oceanic Technol., 30, 10911106, doi:10.1175/JTECH-D-12-00058.1.

    • Search Google Scholar
    • Export Citation
  • Rossow, W. B., , and Schiffer R. A. , 1991: ISCCP cloud data products. Bull. Amer. Meteor. Soc., 72, 220, doi:10.1175/1520-0477(1991)072<0002:ICDP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Rossow, W. B., , and Schiffer R. A. , 1999: Advances in understanding clouds from ISCCP. Bull. Amer. Meteor. Soc., 80, 22612287, doi:10.1175/1520-0477(1999)080<2261:AIUCFI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Rozendaal, M. A., , Leovy C. B. , , and Klein S. A. , 1995: An observational study of diurnal variations of marine stratiform cloud. J. Climate, 8, 17951809, doi:10.1175/1520-0442(1995)008<1795:AOSODV>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Rutan, D., , Rose F. , , Roman M. , , Manalo-Smith N. , , Schaaf C. , , and Charlock T. , 2009: Development and assessment of broadband surface albedo from Clouds and the Earth’s Radiant Energy System clouds and radiation swath data product. J. Geophys. Res., 114, D08125, doi:10.1029/2008JD010669.

    • Search Google Scholar
    • Export Citation
  • Servain, J., , Busalacchi A. J. , , McPhaden M. J. , , Moura A. D. , , Reverdin G. , , Vianna M. , , and Zebiak S. E. , 1998: A Pilot Research Moored Array in the Tropical Atlantic (PIRATA). Bull. Amer. Meteor. Soc., 79, 20192031, doi:10.1175/1520-0477(1998)079<2019:APRMAI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Slingo, A., , Hodges K. I. , , and Robinson G. J. , 2004: Simulation of the diurnal cycle in a climate model and its evaluation using data from Meteosat 7. Quart. J. Roy. Meteor. Soc., 130, 14491467, doi:10.1256/qj.03.165.

    • Search Google Scholar
    • Export Citation
  • Smith, G. L., , Priestley K. J. , , Loeb N. G. , , Wielicki B. A. , , Charlock T. P. , , Minnis P. , , Doelling D. R. , , and Rutan D. A. , 2011: Clouds and Earth Radiant Energy System (CERES), a review: Past, present and future. Adv. Space Res., 48, 254263, doi:10.1016/j.asr.2011.03.009.

    • Search Google Scholar
    • Export Citation
  • Stackhouse, P. W., Jr., , Gupta S. K. , , Cox S. J. , , Zhang T. , , Mikovitz J. C. , , and Hinkelman L. M. , 2011: 24.5-year SRB data set released. GEWEX News, No. 1, International GEWEX Project Office, Silver Spring, MD, 10–12.

  • Stephens, G. L., and et al. , 2012: An update on Earth’s energy balance in light of the latest global observations. Nat. Geosci., 5, 681–696, doi:10.1038/ngeo1580.

    • Search Google Scholar
    • Export Citation
  • Swinbank, W. C., 1963: Long-wave radiation from clear skies. Quart. J. Roy. Meteor. Soc., 89, 339348, doi:10.1002/qj.49708938105.

  • Taylor, K. E., 2001: Summarizing multiple aspects of model performance in a single diagram. J. Geophys. Res., 106, 71837192, doi:10.1029/2000JD900719.

    • Search Google Scholar
    • Export Citation
  • Taylor, P. C., 2012: Tropical outgoing longwave radiation and longwave cloud forcing diurnal cycles from CERES. J. Atmos. Sci., 69, 36523669, doi:10.1175/JAS-D-12-088.1.

    • Search Google Scholar
    • Export Citation
  • Tegen, I., , and Lacis A. A. , 1996: Modeling of particle size distribution and its influence on the radiative properties of mineral dust aerosol. J. Geophys. Res., 101, 19 23719 244, doi:10.1029/95JD03610.

    • Search Google Scholar
    • Export Citation
  • Toon, O. B., , McKay C. P. , , Ackerman T. P. & , and Santhanam K. , 1989: Rapid calculation of radiative heating rates and photodissociation rates in inhomogeneous multiple scattering atmospheres. J. Geophys. Res., 94, 16 28716 301, doi:10.1029/JD094iD13p16287.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., , Fasullo J. T. , , and Kiehl J. , 2009: Earth’s global energy budget. Bull. Amer. Meteor. Soc., 90, 311323, doi:10.1175/2008BAMS2634.1.

    • 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, doi:10.1175/1520-0477(1996)077<0853:CATERE>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Wilber, A. C., , Kratz D. P. , , Gupta S. K. , 1999: Surface emissivity maps for use in satellite retrievals of longwave radiation. NASA Tech. Publ. NASA/TP-1999-209362, 35 pp.

  • Yang, S.-K., , Zhou S. , , and Miller A. J. , 1998: SMOBA: A 3-dimensional daily ozone analysis using SBUV/2 and TOVS measurements. Accessed 6 January 2015. [Available online at http://www.cpc.ncep.noaa.gov/products/stratosphere/SMOBA/smoba_doc.shtml.]

  • Zhang, Y.-C., , Rossow W. B. , , and Lacis A. A. , 1995: Calculation of surface and top of atmosphere radiative fluxes from physical quantities based on ISCCP data sets: 1. Method and sensitivity to input data uncertainties. J. Geophys. Res., 100, 11491165, doi:10.1029/94JD02747.

    • Search Google Scholar
    • Export Citation
  • Zhang, Y.-C., , Rossow W. B. , , Lacis A. A. , , Oinas V. , , and Mishchenko M. I. , 2004: Calculation of radiative fluxes from the surface to top of atmosphere based on ISCCP and other global data sets: Refinements of the radiative transfer model and the input data. J. Geophys. Res.,109, D19105, doi:10.1029/2003JD004457.

  • Zhang, Y.-C., , Rossow W. B. , , and Stackhouse P. W. Jr., 2006: Comparison of different global information sources used in surface radiative flux calculation: Radiative properties of the near-surface atmosphere. J. Geophys. Res., 111, D13106, doi:10.1029/2005JD006873.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Global maps showing the bias in monthly mean downward (a) shortwave and (b) longwave fluxes at the surface when diurnal cycle changes in clouds are not explicitly taken into account throughout the day. The bias is the difference between surface fluxes computed using only the MODIS Terra instrument to derive cloud properties twice daily and then interpolated across time, minus fluxes calculated using cloud properties from MODIS Terra, MODIS Aqua, and calibrated 3-hourly geostationary imager data.

  • View in gallery

    Locations where observations of surface downwelling flux are available during the time span considered in this paper, March 2000–July 2010: 37 land-based site (white) and 48 ocean buoys (blue). Note that not all sites have continuous records of observation for the entire length of time.

  • View in gallery

    Deseasonalized monthly mean downward surface shortwave flux from March 2000 through December 2007. Data are averaged for each month when surface observations at any subset of 37 locations are available. The correlation coefficient of calculated and observed deseasonalized anomalies and the standard deviation of the anomaly differences are shown at the bottom left of each plot. The histogram in (a) shows the number of sites used in each monthly average with a maximum of 37.

  • View in gallery

    Deseasonalized monthly mean downward surface longwave flux from March 2000 through December 2007. Data are averaged for each month when surface observations at any subset of 37 locations are available. The correlation coefficient of calculated and observed deseasonalized anomalies and the standard deviation of the anomaly differences are shown at the bottom left of each plot. The histogram in (a) shows the number of sites used in each monthly average with a maximum of 37.

  • View in gallery

    Taylor diagrams of computed and observed surface downward (a) shortwave and (b) longwave flux at land and ocean sites. The correlation coefficient and standard deviation is computed over time with the time series of deseasonalized anomalies shown in Figs. 3 and 4. The dashed line indicates the standard deviation of observed deseasonalized anomalies at the sites.

  • View in gallery

    (a) Direct comparison of the standard deviation of the 3-hourly transmissions (calculation minus observation) for each of 37 site/months over 8 years plotted against the monthly mean bias for that same site/month. (Red line is the linear fit.) (b) The mean values of the data from (a) for four surface flux data products. Each product name is followed by the correlation coefficient. The relatively high correlations and approximate linear relationship [in (b)] indicate that for the shortwave surface transmission, the smaller the variability in the 3-hourly calculations, the better the monthly mean bias.

  • View in gallery

    Observed monthly mean diurnal cycle of (a) shortwave surface downward flux at Table Mountain and (b) longwave surface downward flux at the WHOI Stratus buoy (20.2°N, 94.8°W) in the eastern Pacific Ocean. Three time averaging scales are shown: 15-min averages, hourly averages, and 3-h averages.

  • View in gallery

    Three-hourly surface downward shortwave flux observations and calculations at Table Mountain, averaged using Julys from 2000 through 2007. (a) Observations averaged to the mid-3-h time period, 0130, 0430, 0730 UTC, etc., to match SYN1deg and MERRA products. (b) Observations averaged to the top of a 3-h time period, 0000, 0300, 0600 UTC, etc., to match ISCCP-FD and SRB3.0 products. Solid lines indicate average downward radiant flux, dashed lines are the difference of calculated minus observed. Bias and standard deviation are listed in the label for each product in parentheses.

  • View in gallery

    Three-hourly surface DLF observations and calculations at Stratus buoy site, S. East tropical Pacific Ocean, averaged using Julys from 2000 through 2007. (a) Observations averaged to the mid-3-h time period—0130, 0430, 0730 UTC, etc.—to match SYN1deg and MERRA products. (b) Observations averaged to the top of a 3-h time period—0000, 0300, 0600 UTC, etc.—to match ISCCP-FD and SRB3.0 products. Solid lines indicate average downward radiant flux, and dashed lines are the difference of calculated minus observed. Bias and standard deviation are listed in the label for each product in parentheses.

  • View in gallery

    (a) Scatterplot of the phase of the diurnal cycle first harmonic of calculated and observed monthly surface downward shortwave flux. Observed fluxes are from 31 land surface sites. Computed fluxes are from SYN1deg. (b) Scatterplot of the difference between computed and observed amplitude of the diurnal cycle first harmonic for surface downward shortwave flux (computed minus observed) and observed amplitude. “Sig” is the standard deviation and r is the correlation coefficient.

  • View in gallery

    As in Fig. 10, but for downward longwave surface flux at the same 31 sites.

  • View in gallery

    Bias distribution of monthly mean longwave downwelling flux from 56 surface sites (black line) and from the Ny Alesund site (red line) and the Alice Springs site (blue line).

  • View in gallery

    Bootstrap distributions (500 realizations) of monthly mean bias (calculation minus observations) where original the distribution of sites is reduced by randomly removing all realizations from 4, 8, 12, etc. surface sites: (a) bias distribution and (b) standard deviation distribution. Each of the 500 samples has a different total n, since each site has a distinct number of monthly mean values available. The standard deviation of each distribution represents the empirical standard error of the mean increasing as more and more sites are removed. The gray dashed line is the original bias and standard deviation of the dataset.

  • View in gallery

    Empirical standard error of the mean derived from LW (solid) and SW (dashed) bias (black) and standard deviation (red) bootstrap distributions. Blue lines represent estimate of a “true” standard error of the mean.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 281 281 46
PDF Downloads 229 229 50

CERES Synoptic Product: Methodology and Validation of Surface Radiant Flux

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 Synoptic (SYN1deg), edition 3, product provides climate-quality global 3-hourly 1° × 1°gridded top of atmosphere, in-atmosphere, and surface radiant fluxes. The in-atmosphere surface fluxes are computed hourly using a radiative transfer code based upon inputs from Terra and Aqua Moderate Resolution Imaging Spectroradiometer (MODIS), 3-hourly geostationary (GEO) data, and meteorological assimilation data from the Goddard Earth Observing System. The GEO visible and infrared imager calibration is tied to MODIS to ensure uniform MODIS-like cloud properties across all satellite cloud datasets. Computed surface radiant fluxes are compared to surface observations at 85 globally distributed land (37) and ocean buoy (48) sites as well as several other publicly available global surface radiant flux data products. Computed monthly mean downward fluxes from SYN1deg have a bias (standard deviation) of 3.0 W m−2 (5.7%) for shortwave and −4.0 W m−2 (2.9%) for longwave compared to surface observations. The standard deviation between surface downward shortwave flux calculations and observations at the 3-hourly time scale is reduced when the diurnal cycle of cloud changes is explicitly accounted for. The improvement is smaller for surface downward longwave flux owing to an additional sensitivity to boundary layer temperature/humidity, which has a weaker diurnal cycle compared to clouds.

Corresponding author address: David A Rutan, Science Systems and Applications Inc., 1 Enterprise Parkway, Suite 200, Hampton, VA 23666. E-mail: david.a.rutan@nasa.gov

Abstract

The Clouds and the Earth’s Radiant Energy System Synoptic (SYN1deg), edition 3, product provides climate-quality global 3-hourly 1° × 1°gridded top of atmosphere, in-atmosphere, and surface radiant fluxes. The in-atmosphere surface fluxes are computed hourly using a radiative transfer code based upon inputs from Terra and Aqua Moderate Resolution Imaging Spectroradiometer (MODIS), 3-hourly geostationary (GEO) data, and meteorological assimilation data from the Goddard Earth Observing System. The GEO visible and infrared imager calibration is tied to MODIS to ensure uniform MODIS-like cloud properties across all satellite cloud datasets. Computed surface radiant fluxes are compared to surface observations at 85 globally distributed land (37) and ocean buoy (48) sites as well as several other publicly available global surface radiant flux data products. Computed monthly mean downward fluxes from SYN1deg have a bias (standard deviation) of 3.0 W m−2 (5.7%) for shortwave and −4.0 W m−2 (2.9%) for longwave compared to surface observations. The standard deviation between surface downward shortwave flux calculations and observations at the 3-hourly time scale is reduced when the diurnal cycle of cloud changes is explicitly accounted for. The improvement is smaller for surface downward longwave flux owing to an additional sensitivity to boundary layer temperature/humidity, which has a weaker diurnal cycle compared to clouds.

Corresponding author address: David A Rutan, Science Systems and Applications Inc., 1 Enterprise Parkway, Suite 200, Hampton, VA 23666. E-mail: david.a.rutan@nasa.gov

1. Introduction

The earth’s radiation budget describes the exchange of radiant shortwave (SW) and longwave (LW) energy between the earth and space and indicates how much energy is available to drive the hydrological cycle and general circulation (Trenberth et al. 2009). Earth’s response and feedback to radiative forcing directly influences Earth’s radiation budget. At the top of the atmosphere (TOA), satellite instruments such as the Clouds and the Earth’s Radiant Energy System (CERES) provide global observations of radiant energy reflected and emitted by Earth (Wielicki et al. 1996; Loeb et al. 2009; Doelling et al. 2013). CERES instruments on board the Terra and Aqua satellites have been operating since December 1999 and May 2002, respectively, accurately monitoring the earth’s reflected and emitted radiation. A CERES instrument on board the Suomi–National Polar-Orbiting Partnership (SNPP) mission launched in October 2011.

Estimating the surface radiation budget is more complex than at the TOA, as it requires a radiative transfer model and satellite-derived properties of clouds and aerosols and atmospheric state from either satellites or reanalysis. Underlying assumptions in the radiative transfer model calculations and ancillary input data error increases the uncertainty in the surface radiation budget estimates. Furthermore, it is known that the diurnal cycle of clouds and their contribution to the diurnal cycle of surface radiant flux must be taken into account (Bergman and Salby 1996). Other studies (Slingo et al. 2004; Taylor 2012) indicate that including the diurnal cycle of clouds is needed for meaningful comparisons with global climate models (GCMs) and accurate representation of TOA and surface radiation budgets.

Multiple independent estimates of the surface radiation budget are necessary in order to increase our confidence in satellite-derived surface radiation budget. Here we compare CERES surface radiative fluxes with those from the International Satellite Cloud Climatology Project (ISCCP) (Zhang et al. 1995) and the Global Energy and Water Cycle Experiment (GEWEX) (Pinker and Laszlo 1992; Stackhouse et al. 2011). Zhang et al. (2004) improved the computation by including satellite-derived aerosol vertical profiles. Unlike ISCCP, cloud properties used in the CERES synoptic 1° × 1° (SYN1deg) product are derived from both MODIS and geostationary satellite measurements using the approaches described in Minnis et al. (2011a,b) and Minnis et al. (1995), respectively. The different independent datasets help improve our understanding of the impact of different ancillary inputs used to compute surface radiant flux (hereinafter, the terms surface radiant flux and flux will be used interchangeably).

More recently, new observations have emerged that further improve the representation of Earth’s radiation budget. At the TOA, the CERES energy balanced and filled (EBAF) product uses 5 years of Argo in situ ocean heat content tendencies (Roemmich and Gilson 2009) to provide a constraint for global mean energy imbalance (Loeb et al. 2009, 2012). Stephens et al. (2012), Kato et al. (2011), and Henderson et al. (2013) use combined CERES, MODIS, CloudSat, and Cloud–Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) data to provide improved instantaneous surface radiative fluxes. These datasets have also been used to estimate uncertainties in passive-only surface flux estimates (Kato et al. 2011).

The CERES EBAF-surface product, which also integrates multiple satellite observations, provides monthly mean surface upward and downward longwave and shortwave fluxes that are consistent with the TOA EBAF estimates (Kato et al. 2013). The primary input into the EBAF-surface product is the CERES SYN1deg, edition 3, data product. SYN1deg relies on a one-dimensional radiative transfer model to compute 3-hourly surface radiant flux. Inputs to this model include imager data from Terra, Aqua, and geostationary satellites to resolve the diurnal cycle of clouds (Doelling et al. 2013) and reanalysis data for the diurnal cycle of temperature and water vapor amount (Kato et al. 2013).

The purpose of this paper is to describe the SYN1deg data product (section 2) and to evaluate the computed surface fluxes by comparing them with surface measurements over both land and ocean at various sites around the world (section 3). We also compare the SYN1deg data to other surface radiation datasets, focusing on the impact of the diurnal cycle of clouds on surface flux computations. As these data products are often used to assess the quality of global climate models, it is important to assess surface flux products over a range of time scales (3 hourly, hourly, daily, and monthly). Additionally, in appendix C we consider the continued need for globally distributed, well-maintained, long-term surface observation sites.

2. CERES SYN1deg, edition 3a, data product

The SYN1deg product combines Terra and Aqua CERES and MODIS observations and 3-hourly geostationary (GEO) data to provide 3-hourly broadband TOA radiant flux and cloud properties (Doelling et al. 2013). The SYN1deg product also contains computed 3-hourly TOA, in-atmosphere (at three levels), and surface fluxes based upon radiative transfer model calculations. Direct and diffuse shortwave, photosynthetic active radiation (PAR), and ultraviolet (UVA, UVB) radiation are also computed in SYN1deg. The product also contains the aerosol and atmospheric data used as input to compute the flux. Fluxes are provided as a 1° × 1° latitude × longitude gridded product in 3-hourly UTC time. CERES products and associated documentation can be obtained at the project website (http://ceres.larc.nasa.gov).

a. Model inputs

The radiative transfer model calculations require numerous input datasets merged together in a consistent manner. Table 1 provides a detailed list of ancillary inputs used to produce the SYN1deg data product. Cloud properties are derived from MODIS and multiple geostationary satellites imagers. Each GEO instrument is calibrated against well-calibrated Terra-MODIS radiances (Doelling et al. 2013). The CERES single satellite footprint (SSF), edition 2, cloud algorithm (Minnis et al. 2011a,b) derives cloud properties (e.g., fraction, optical depth, top height, particle size, and phase) from narrowband radiances measured by MODIS twice a day from March 2000 through August 2002 (Terra only) and 4 times a day after July 2002 (Terra plus Aqua). The GEO, edition 2, two-channel visible and infrared cloud algorithm (Minnis et al. 1995) provides cloud properties between MODIS observations at 3-hourly temporal resolution. All cloud properties are gridded into 44 012 equal-area grid boxes (1° × 1° between 45°N and 45°S, and larger toward the poles) and interpolated to hourly temporal values (UTC hour boxes centered on the half hour) to fill times with no retrieved cloud properties. Details on how the GEO data are processed can be found in Doelling et al. (2013). Cloud-base height is estimated by an empirical formula described by Minnis et al. (2011a). Pressure, temperature, and water vapor profiles are specified from the Global Modeling and Assimilation Office (GMAO) Goddard Earth Observing System Model, version 4 (GEOS-4), a 1° gridded reanalysis product (Bloom et al. 2005) until 2007, after which GEOS-5.2 was incorporated. Ozone comes from the National Centers for Environmental Prediction Stratosphere Monitoring Ozone Blended Analysis (SMOBA) product (Yang et al. 1998).

Table 1.

SYN1deg inputs.

Table 1.

The radiative transfer code requires spectral albedo and emissivity for the bands shown in appendix A. Spectral albedo is estimated at ⅙° resolution and then integrated over the equal-area grid box. We also derive a broadband surface albedo for each CERES grid box. This broadband surface albedo is based upon observed TOA albedo, meteorology, and surface type (Table 2) and is used to adjust the initial spectral albedo such that the integral of the spectral albedo equals the observationally derived broadband surface albedo value. Solar zenith angle for a grid box is calculated as the integral across an hour at its central latitude unless the box is completely overcast, in which case a value of 55° is used. Initial broadband surface albedo and spectral shape for each grid box are determined through a hierarchy of parameterized models as shown in Table 2. For all-sky ocean albedo with no sea ice, snow-covered land, and ice-covered water, SYN1deg uses a lookup table (LUT) based on the Coupled Ocean–Atmosphere Radiative Transfer (COART) model developed in Jin et al. (2004) that provides broadband albedo and its spectral shape. For snow-free land, spectral albedo comes from a static ⅙° resolution map based on International Geosphere–Biosphere Programme (IGBP) scene type and associated spectral shapes (Rutan et al. 2009). Broadband albedo for clear sky, snow-free land is determined via a parameterized version of the Fu and Liou surface albedo (FLSA) model. This model is run in advance, monthly, using clear-sky data to create a surface albedo history (SAH) map that is used for broadband albedo under cloudy skies. The methodology is similar to that developed in Rutan et al. (2009). The main difference is that the validation approach shown in that paper comparing CERES footprint results to MODIS data does not apply to the gridbox values used in SYN1deg. Land surface albedo is handled in the same manner for grid boxes with less than 10% snow. Initial albedo for grid boxes that contain a combination of snow or sea ice and land or water are derived by integrating results from the previously mentioned models. Various scene-type limits for model selection are found in the lower rows of Table 2.

Table 2.

Surface albedo source for land and ocean. LINT: linear interpolation of albedo weighted by ice or snow fraction. SAH: surface albedo history map (preprocessed before month is run) updated monthly. ASSALUT: all-sky surface albedo LUT for cloud optical depth < 20. SGR: snow grain size retrieval (snow grain size points to one of two different spectral shapes, fresh and old snow, and its associated broadband albedo. COART: Coupled Ocean–Atmosphere Radiative Transfer Model. FLSA: Fu and Liou surface albedo parameterization.

Table 2.

Emissivity over land and water is derived in 12 spectral bands (appendix A) based on surface scene type (Wilber et al. 1999) as defined by the IGBP map of 17 Earth surface types along with fresh snow and ice placed into a ⅙th° resolution grid (Rutan et al. 2009). In addition, three LW bands (bands 5–7, appendix A) over land can be changed based on monthly climatological values derived from imager radiance observations to give a more accurate representation when observations are available. The spectrally integrated value for the emissivity of water does not change and is approximately 0.985. The ⅙th° emissivity values are integrated to provide a value for the CERES grid box.

Aerosol optical depths (AODs) come from the MODIS (MOD04) product (Remer et al. 2005) and/or the Model for Atmospheric Transport and Chemistry (MATCH) (Collins et al. 2001). MATCH is a global chemical transport model that assimilates MODIS retrievals of aerosol optical depth using collection 4 prior to May 2006 and collection 5 thereafter. MATCH defines aerosol constituents throughout the vertical atmospheric column using Optical Properties of Aerosols and Clouds (OPAC) from Hess et al. (1998) and desert dust from Tegen and Lacis (1996). Snow and ice information come from National Snow and Ice Data Center daily maps supplemented by maps from NOAA NESDIS and a cryosphere flag derived from MODIS clear-sky imager. All of these input datasets provide consistent parameter quality over the CERES record.

b. CERES SYN1deg flux computation

SYN1deg provides computed surface and profile radiant fluxes and cloud properties that can be compared directly with climate model output. TOA, surface, and fluxes at three pressure levels (70, 200, and 500 hPa) are computed hourly in approximate equal-area grid boxes using the Langley Fu–Liou radiative transfer code (Fu and Liou 1993; Fu et al. 1998; Kratz and Rose 1999; Kato et al. 1999, 2005). Hourly values are subsequently averaged to give 3-h averages at 1330, 1630 UTC, etc., times of the day. Gaseous absorption in the shortwave region is treated by the method described in Kato et al. (1999) and accounts for absorption by water vapor, carbon dioxide, ozone, methane, and oxygen. For carbon dioxide, methane, and nitrous oxide, secularly increasing values are used from the NOAA/ESRL annual greenhouse gas index (AGGI; http://www.esrl.noaa.gov/gmd/aggi/aggi.html). Note that the band structure used in computations is modified from the original Fu–Liou code and different from that described in Kato et al. (1999). A list of the LW and SW bands in the current model is given in appendix A and in Rose et al. (2006). Gaseous absorption in the longwave regions is treated by the method described in Kratz and Rose (1999). A complete description of the code is given in Rose et al. (2013).

To treat horizontal variability of optical thickness within a cloud layer explicitly in radiative transfer computations, both linear and logarithmic means of cloud optical thicknesses are computed for clouds at four levels: low, low–middle, middle, and high. These layers are defined by cloud-top pressure: surface to 700, 700–500, 500–300, and less than 300 hPa. Cloud optical thickness expressed as a gamma distribution is estimated from the linear and logarithmic cloud optical thickness means (Barker 1996; Oreopoulos and Barker 1999; Kato et al. 2005). Once the distribution of cloud optical thickness is estimated for each cloud layer, a gamma-weighted two-stream radiative transfer model (Kato et al. 2005) is used to compute the shortwave flux vertical profile for each cloud layer. The logarithmic mean optical thickness is used in the longwave flux computation with a modified two-stream approximation (Toon et al. 1989; Fu et al. 1998). Though any combination of the four cloud layers and clear conditions can be run for a single grid/hour box, SYN1deg does not include cloud overlap in radiative transfer calculations. Radiant fluxes are then computed for several atmospheric states.

Radiant fluxes are computed for pristine (no aerosol, no clouds), clear-sky (include aerosol, no clouds), all-sky no-aerosol (no aerosol, include clouds), and all-sky conditions with inputs described in section 2a. To achieve consistency between computed and observed TOA flux, small adjustments are made to the input variables within predetermined uncertainties for each variable (Rose et al. 2013). Computed fluxes from original inputs are labeled “untuned” and the adjusted input and computed fluxes are labeled as “tuned.” For example, in cloudy-sky conditions, cloud-top height, cloud optical depth, and cloud fraction are noniteratively adjusted within their uncertainties to simultaneously match calculated shortwave and longwave to TOA observed flux. TOA observations also carry an uncertainty that is used in the tuning algorithm. Uncertainty in the TOA flux outside of CERES overpass times is large due to larger uncertainties in the GEO-derived TOA flux (Doelling et al. 2013). Because of this larger GEO uncertainty, inputs are perturbed little in the tuning process resulting in a nearly negligible impact on untuned monthly regional fluxes. Thus, for the SYN1deg we recommend using untuned surface and TOA fluxes, which are used for comparisons in this paper.

3. Calculated surface flux validation

Doelling et al. (2013) describe the methodology and impact of incorporating calibrated imager data from geostationary satellites to determine TOA fluxes in the SYN1deg product. We begin by examining the impact of this enhanced diurnal cycle of cloud properties on the diurnal cycles of surface downward shortwave and longwave flux by considering results with and without the enhancement. Figure 1 shows differences between monthly 1° × 1° gridded mean shortwave and longwave surface downward fluxes for July 2010 computed with two sets of cloud properties. The first uses cloud properties derived only from the Terra satellite [CERES flight model 1 (FM1) and MODIS imager] interpolated across the day. Terra has a local equatorial crossing time of 1030. The second uses cloud properties derived from MODIS on both the Terra and Aqua satellites as well as TOA fluxes and cloud properties derived using multiple geostationary satellites that cover the 60°N–60°S latitudinal band. Surface downward shortwave flux shows differences up to 10 W m−2 in some regions. Prominent differences are found in the eastern Pacific and Atlantic Oceans, where marine stratocumulus clouds occur. Generally, marine stratocumulus clouds reach their maximum cloud fraction in the early morning (e.g., Rozendaal et al. 1995) and their minimum in the afternoon. A product that relied completely on sampling based on instruments on board the Terra satellite, which has an ~1030 equatorial crossing time, misses the decrease in afternoon cloud cover. However, when cloud properties derived from Terra, Aqua, and geostationary satellites are used, the decrease is captured, which appears as an increase in the surface SW flux. As a consequence monthly mean surface SW (LW) fluxes are larger (smaller) for the combined satellite product than those computed with Terra-only-derived cloud properties. This leads to the negative bias in SW and positive bias in LW in the marine stratocumulus regions found in Figs. 1a and 1b. Similarly, sampling over the course of the day by multiple satellites captures convective clouds that occur in the afternoon over land. This gives a smaller surface downward shortwave flux and larger downward longwave flux than those computed with Terra only. In addition, the low-level cloud amount over land reaches a maximum in early afternoon (Cairns 1995). Mid- and high-level clouds reach maxima in nighttime and early morning for most regions (Cairns 1995). The differences shown in Fig. 1 are consistent with these diurnal cycles of cloud properties. To quantify surface flux improvements, we compare calculations at the surface directly with observed surface fluxes.

Fig. 1.
Fig. 1.

Global maps showing the bias in monthly mean downward (a) shortwave and (b) longwave fluxes at the surface when diurnal cycle changes in clouds are not explicitly taken into account throughout the day. The bias is the difference between surface fluxes computed using only the MODIS Terra instrument to derive cloud properties twice daily and then interpolated across time, minus fluxes calculated using cloud properties from MODIS Terra, MODIS Aqua, and calibrated 3-hourly geostationary imager data.

Citation: Journal of Atmospheric and Oceanic Technology 32, 6; 10.1175/JTECH-D-14-00165.1

To evaluate surface flux calculations, we use observations from 85 surface locations around the globe. Figure 2 shows the locations of 37 land surface sites (white diamonds) and 48 buoy locations (blue diamonds). We use observations taken from March 2000 through December 2007. More detailed descriptions of these sites are given in appendix B. If available, combined observations from a narrow incidence pyrheliometer for direct flux and from a shaded pyranometer for diffuse flux are used for observed shortwave. If not available (as on all buoys), the unshaded pyranometer observation is used. Longwave flux is typically observed on land by a shaded pyrgeometer. Measurements at these sites are independently calibrated under different programs including the Baseline Surface Radiation Network (BSRN; Ohmura et al. 1998), NOAA’s Global Monitoring Division, the Surface Radiation Network (SURFRAD; Augustine et al. 2000), and the U.S. Department of Energy’s Atmospheric System Research (ASR) program. Ocean buoy site data were downloaded from Woods Hole Oceanographic Institution (WHOI) and the NOAA/Pacific Marine Environmental Laboratory (PMEL). Validation sites used in this study are selected based on data availability, radiometer accuracy, and proximity from other surface sites. All land and buoy sites have shortwave flux measurements, while pyrgeometers are included in all land and 19 buoys to observe longwave flux. To avoid undersampled months, an 85% operational time threshold is used to compute a valid monthly mean. Site/months that do not meet this threshold are also filtered from shorter time period comparisons. Further information on these sites can be found on the CERES/ARM Validation Experiment (CAVE) website (http://www-cave.larc.nasa.gov), where one can plot and download both surface observations and the collocated SYN1deg data. We rely primarily on quality flags provided by the surface data products to filter poor-quality data and to perform threshold tests for unreasonable values as well. Since errors in input used in flux computations might depend on geographical region, it is important that validation sites cover a wide range of geographical regions. An analysis of how surface site selection influences the validation of computed fluxes is found in appendix C. We find the bias distribution (calculation minus observation) is impacted by site selection due to persistent biases that can compensate for each other and that removing sites can affect the distribution in a nonrandom manner.

Fig. 2.
Fig. 2.

Locations where observations of surface downwelling flux are available during the time span considered in this paper, March 2000–July 2010: 37 land-based site (white) and 48 ocean buoys (blue). Note that not all sites have continuous records of observation for the entire length of time.

Citation: Journal of Atmospheric and Oceanic Technology 32, 6; 10.1175/JTECH-D-14-00165.1

In addition to the surface radiant flux included in the SYN1deg product, several other surface flux data products are included in the comparison. These are the 40-yr European Centre for Medium-Range Weather Forecasts Re-Analysis, the ECMWF interim reanalysis (ERA-Interim) (Dee et al. 2011), the Modern-Era Retrospective Analysis for Research and Applications (MERRA) (Rienecker et al. 2011), ISCCP flux data (FD) (Rossow and Schiffer 1991, 1999; Zhang et al. 2004) and Global Energy and Water Cycle Experiment Surface Radiation Budget (GEWEX SRB) 3.0 products (Stackhouse et al. 2011). Surface fluxes from ECMWF and MERRA are based on reanalysis, whereas SRB relies on the same GEOS profiles as SYN1deg and ISCCP clouds, which are solely based on GEO data. ISCCP FD computes surface fluxes using ISCCP clouds and TIROS Operational Vertical Sounder (TOVS) profiles. In addition we include surface flux from the edition 2.7 EBAF-surface product (Kato et al. 2013), though as the EBAF-surface product is initially calculated from the SYN1deg product, it is not a completely independent dataset.

a. Evaluation of computed surface flux using surface observations

Tables 3 and 4 show monthly mean biases of shortwave and longwave surface flux for land, ocean, and all 85 surface sites combined as well as the standard deviations at three time scales (3-hourly, daily, and monthly mean) for each product. Statistics are for computed-minus-observed fluxes for nearly 8 years, March 2000 through December 2007. The beginning of the SYN1deg and length of the SRB3.0 products determines this time limit. ERA-Interim and EBAF-surface data were compared for monthly means only. Table 3 shows that the EBAF-surface data have the smallest shortwave bias overall, though four products have biases between 1% and 2%. However, the ISCCP-FD and SRB3.0 biases are low as a result of compensating errors between land and ocean. SYN1deg standard deviation is smaller by approximately 30% (relative to the other three high-temporal-resolution products) for 3-h and daily averages. Standard deviations for SYN1deg and EBAF-surface data are ~45% (relative) lower than those for the other products. Surface shortwave fluxes are primarily dependent on TOA insolation, solar zenith angle, cloud fraction, and aerosol (Long and Ackerman 2000). Since TOA insolation and solar zenith angle are known, the diurnal cycle of cloud fraction becomes the dominant feature to determine the quality of calculated SW flux at shorter time scales. Doelling et al. (2013) show that GEO imager radiances calibrated to the Terra-MODIS imager provide a stable database from which to derive cloud properties across space and time. This improved diurnal cycle of cloud fraction in the SYN1deg product accounts for the lower standard deviations compared to the other products at shorter time scales and is discussed in more detail in section 3b.

Table 3.

Statistics for calculated-minus-observed surface shortwave downward radiant flux across 8 years (2000–07) with observed means of 181 (land sites), 236 (ocean sites), and 202 W m−2 (all sites*).

Table 3.
Table 4.

Monthly mean bias and standard deviation of calculated-minus-observed surface longwave downward radiant flux across 8 years (2000–07) with observed means of 319 (land sites), 401 (ocean sites), and 330 W m−2 (all sites*).

Table 4.

Table 4 shows that for longwave flux, the SYN1deg has the lowest standard deviation at the 3-h time average, but that all products are similar at daily and monthly mean time scales except perhaps for the ISCCP-FD product. Monthly mean biases range from −17.6 to +8.1 W m−2 with the SYN1deg product having a bias of −4 W m−2. This −4 W m−2 bias in SYN1deg was estimated to be primarily due to low clouds and is removed in the monthly means as described in Kato et al. (2013) for the EBAF-surface data, whose bias is 0.6 W m−2 (Table 4). Longwave surface fluxes are primarily dependent on atmospheric temperature, water vapor profiles, and cloud-base height. Since both the satellite and reanalysis products use similar meteorological inputs, there is less diversity among the resulting statistics. Thus, improvement in cloud fraction via the CERES-GEO processing has its greatest impact on surface shortwave calculations.

To ensure the SYN1deg fluxes are of climate quality, they must replicate the interannual variability of surface-observed flux. For each dataset we compute longwave and shortwave monthly mean flux anomalies and compare them to anomalies observed by surface measurements. Figures 3 and 4 show deseasonalized time series of observed and calculated shortwave and longwave surface fluxes at 29 land, six island, and two buoy sites. We do not include the majority of the buoy data, as few maintain a complete time series over the 8 years considered. Deseasonalized anomalies ΔF(m) are defined as differences between monthly mean flux F and the climatological monthly mean , such that
e1
where x is shortwave or longwave, m is the month in the time series, and M is the canonical month January, February, etc. For example, for July is the average of monthly mean values from all available sites during July for the years 2000–07. Each point represents a deseasonalized value from the ensemble (up to 37 sites) mean month. The histograms in Figs. 3a and 4a show the total number of sites used for each month. While true not all months have the same number of sites, the calculation of anomalies is valid as a test so long as the products and observations are treated equally and the variable number of sites does not affect the variability significantly. “Truth” for the purpose of this analysis is not an accurate representation of interannual variations of surface flux but instead, given the available surface observations, do the products accurately recreate the anomalies observed? Most products agree well with observed anomalies for both longwave and shortwave fluxes. There are notable differences in the correlation coefficients as well as during periods when the correlation with observations differ more severely, for example, ISCCP-FD data during the first several years of the longwave record. The SYN1deg and EBAF-surface datasets have the lowest standard deviations and best capture the monthly shortwave flux anomalies. A comparison of monthly longwave flux anomalies with surface observations in Fig. 4 indicates the datasets are similar except for the aforementioned ISCCP-FD dataset.
Fig. 3.
Fig. 3.

Deseasonalized monthly mean downward surface shortwave flux from March 2000 through December 2007. Data are averaged for each month when surface observations at any subset of 37 locations are available. The correlation coefficient of calculated and observed deseasonalized anomalies and the standard deviation of the anomaly differences are shown at the bottom left of each plot. The histogram in (a) shows the number of sites used in each monthly average with a maximum of 37.

Citation: Journal of Atmospheric and Oceanic Technology 32, 6; 10.1175/JTECH-D-14-00165.1

Fig. 4.
Fig. 4.

Deseasonalized monthly mean downward surface longwave flux from March 2000 through December 2007. Data are averaged for each month when surface observations at any subset of 37 locations are available. The correlation coefficient of calculated and observed deseasonalized anomalies and the standard deviation of the anomaly differences are shown at the bottom left of each plot. The histogram in (a) shows the number of sites used in each monthly average with a maximum of 37.

Citation: Journal of Atmospheric and Oceanic Technology 32, 6; 10.1175/JTECH-D-14-00165.1

To summarize the statistics, Figs. 5a (shortwave) and Fig. 5b (longwave) show Taylor diagrams (Taylor 2001) demonstrating the standard deviations (of the time series) and correlations between observed and calculated fluxes. Standard deviations between the calculated and observed data are given by the semicircles radiating from the mean-zero standard deviation of the observed time series (shown by the radial dashed purple line). The shortwave diagram (Fig. 5a) indicates that EBAF-surface and SYN1deg surface fluxes have the highest correlation and lowest standard deviations relative to observations. All calculated shortwave fluxes show correlations greater than 0.85 and all have approximately the same standard deviation across time, close to the observed value of ~5.7 W m−2. Standard deviations between calculated and observed flux go from a low of ~1.7 W m−2 for the SYN1deg and EBAF products to a high of ~3.0 W m−2 for the MERRA and ISCCP-FD products. In Fig. 5b, except for the ISCCP-FD dataset, computed longwave flux clusters nicely with ERA-Interim and MERRA reanalyses having the highest correlation and smallest standard deviation with respect to observations. Five out of six products have correlation coefficients near 0.95. As mentioned above the surface downward longwave flux is primarily a function of the input temperature profiles, column water vapor (Swinbank 1963; Dilley and O’Brien 1998), and cloud-base height. Because near-surface air temperature is fairly well characterized in reanalysis and so subsequently the satellite data products (SYN1deg, EBAF surface, and SRB) that use various reanalysis products for meteorological inputs, most of the products show a similar agreement with surface measurements. In addition, although cloud-base height is difficult to retrieve from satellite-based passive sensors, the error in the cloud-base height is often due to missing lower levels of overlapping clouds (Kato et al. 2011). This indicates either that the variability of the cloud-base height has a minor contribution to variability in downward longwave flux or that passive sensors capture part of the variability of lower-level clouds because the upper-layer optical thickness of some midlevel and high-level clouds are small.

Fig. 5.
Fig. 5.

Taylor diagrams of computed and observed surface downward (a) shortwave and (b) longwave flux at land and ocean sites. The correlation coefficient and standard deviation is computed over time with the time series of deseasonalized anomalies shown in Figs. 3 and 4. The dashed line indicates the standard deviation of observed deseasonalized anomalies at the sites.

Citation: Journal of Atmospheric and Oceanic Technology 32, 6; 10.1175/JTECH-D-14-00165.1

Downward surface shortwave flux is highly correlated with the diurnal cycle of cloud fraction, but does an accurate portrayal of the diurnal cycle of clouds improve the computation of monthly mean downward shortwave and longwave flux? This is suggested by the statistics in Table 3, where the SYN1deg has smaller 3-h and daily standard deviations and relatively small biases for both land and ocean sites. Another way to address this question is to look for a direct correlation between computed and measured flux differences at 3-hourly and monthly time scales. Figure 6a shows an 8-yr comparison between SYN1deg and 3-hourly measurements, expressed as the standard deviation of the difference in transmission (calculation minus observation) for each site and month (abscissa) plotted against the monthly mean bias for that same site/month (ordinate). It is not expected that such a relationship would fall along the 1-to-1 line shown, particularly since a mean bias of zero can result from compensating errors as in the land/ocean differences in the SRB3.0 and ISCCP-FD monthly means found in Table 3. Small biases and variances in the SW can also result simply from low sun angles. But if the accuracy of 3-hourly calculations has a significant impact on the monthly mean result, then one would expect the two to be correlated after sufficient sampling. Figure 6b shows the mean for the ordinate versus the mean of the abscissa for each of the four products with 3-hourly resolutions. Correlation coefficients for the two variables are shown in parentheses. The datasets show a correlation between 3-hourly error and monthly mean bias ranging from 0.94 and 0.91, and a linear relationship indicating that as the 3-hourly standard deviation decreases, so too does monthly mean transmission bias. A similar analysis done for downward longwave flux (not shown) had correlations between 0.27 and 0.38 and mean standard deviation versus mean bias showed little if any relationship. Thus, improvements in high-temporal-resolution inputs such as cloud properties (indicated by improved standard deviation between calculated and observed transmission) do positively impact the monthly mean bias in the SW flux down at the surface but do not appear to improve monthly mean bias in downward LW flux.

Fig. 6.
Fig. 6.

(a) Direct comparison of the standard deviation of the 3-hourly transmissions (calculation minus observation) for each of 37 site/months over 8 years plotted against the monthly mean bias for that same site/month. (Red line is the linear fit.) (b) The mean values of the data from (a) for four surface flux data products. Each product name is followed by the correlation coefficient. The relatively high correlations and approximate linear relationship [in (b)] indicate that for the shortwave surface transmission, the smaller the variability in the 3-hourly calculations, the better the monthly mean bias.

Citation: Journal of Atmospheric and Oceanic Technology 32, 6; 10.1175/JTECH-D-14-00165.1

b. Calculation and observation comparison of monthly mean diurnal cycle

We consider the observed monthly mean diurnal cycle of surface downward shortwave and longwave flux at two sites with distinct diurnal cycles: (i) Table Mountain, Colorado, the Boulder site (BOS), in July; and (ii) WHOI Stratus (STR) buoy in October, located at 20.2°S, 94.8°W in the eastern Pacific Ocean off the west coast of Chile. At the Boulder site, the diurnal cycle is strongly influenced by afternoon convection during summertime, while marine stratus clouds dominate at the WHOI site. At the latter site, cloud fraction typically decreases in the afternoon, while lower tropospheric temperature remains fairly constant. Consequently, surface longwave flux decreases in the afternoon (Minnis and Harrison 1984). Figure 7 shows observed monthly mean diurnal cycles at these locations for three temporal resolutions: 15-min averages (gray bars), 1-h averages (black line), and 3-h averages (red line). The decrease in surface downward shortwave flux due to afternoon convection remains distinct in 15-min and 1-h averages but is smoothed out in the 3-h resolution. Failing to model afternoon convective clouds at any time scale would incorrectly increase downward shortwave flux after local noon for this month. The diurnal cycle of longwave flux measured at the Stratus buoy is evident at all three time average resolutions. However, the 3-hourly flux increases the minimum value by ~5 W m−2.

Fig. 7.
Fig. 7.

Observed monthly mean diurnal cycle of (a) shortwave surface downward flux at Table Mountain and (b) longwave surface downward flux at the WHOI Stratus buoy (20.2°N, 94.8°W) in the eastern Pacific Ocean. Three time averaging scales are shown: 15-min averages, hourly averages, and 3-h averages.

Citation: Journal of Atmospheric and Oceanic Technology 32, 6; 10.1175/JTECH-D-14-00165.1

Surface downward radiant fluxes at 3-hourly resolution averaged over 8 years from SYN1deg, MERRA, SRB3.0, and the ISCCP FD are compared to surface observations in Figs. 8 and 9. Shortwave comparisons are shown for BOS in Fig. 8a (SYN1deg, MERRA) and Fig. 8b (ISCCP FD, SRB3.0). Two separate plots are provided because ISCCP FD and SRB3.0 are averaged to the top of a 3-h time period (0000, 0300 UTC, etc.), while SYN1deg and MERRA are averaged to the midpoint of a 3-h time period (0130, 0430 UTC, etc.) For BOS, Fig. 8, except for the SYN1deg, all computed fluxes overestimate the downward shortwave between 1300 and 1800 local time when afternoon convection occurs. As a consequence, standard deviation from the observed diurnal cycle is significantly lower for the SYN1deg (13.2 W m−2).

Fig. 8.
Fig. 8.

Three-hourly surface downward shortwave flux observations and calculations at Table Mountain, averaged using Julys from 2000 through 2007. (a) Observations averaged to the mid-3-h time period, 0130, 0430, 0730 UTC, etc., to match SYN1deg and MERRA products. (b) Observations averaged to the top of a 3-h time period, 0000, 0300, 0600 UTC, etc., to match ISCCP-FD and SRB3.0 products. Solid lines indicate average downward radiant flux, dashed lines are the difference of calculated minus observed. Bias and standard deviation are listed in the label for each product in parentheses.

Citation: Journal of Atmospheric and Oceanic Technology 32, 6; 10.1175/JTECH-D-14-00165.1

Fig. 9.
Fig. 9.

Three-hourly surface DLF observations and calculations at Stratus buoy site, S. East tropical Pacific Ocean, averaged using Julys from 2000 through 2007. (a) Observations averaged to the mid-3-h time period—0130, 0430, 0730 UTC, etc.—to match SYN1deg and MERRA products. (b) Observations averaged to the top of a 3-h time period—0000, 0300, 0600 UTC, etc.—to match ISCCP-FD and SRB3.0 products. Solid lines indicate average downward radiant flux, and dashed lines are the difference of calculated minus observed. Bias and standard deviation are listed in the label for each product in parentheses.

Citation: Journal of Atmospheric and Oceanic Technology 32, 6; 10.1175/JTECH-D-14-00165.1

Figures 9a and 9b show similar plots for surface downward longwave flux at the STR site. Downward longwave flux can be large relative to the diurnal cycle differences, so to focus on those differences, we remove the bias difference by subtracting the monthly mean value from the corresponding 3-hourly flux FLW(t),
e2
Term δFLW(t) is plotted in Fig. 9 and by definition it equals 0.0 over the day. Deviations in δFLW from surface observations are shown as dashed lines. At the STR buoy, all but the MERRA-calculated fluxes appear to show some decrease in afternoon LW flux though of varying magnitude and time of day. MERRA (Fig. 9a) captures little of the diurnal cycle of clouds, while ISCCP FD (Fig. 9b) appears slightly out of phase. The SYN1deg again has the smallest standard deviation across the day, indicating relatively good representation of the nighttime increase and afternoon decrease in low clouds.

These results demonstrate the importance of accounting for the diurnal cycle of clouds in computations of downward shortwave and longwave fluxes (Smith et al. 2011; Taylor 2012). Cloud properties derived from both Terra and Aqua MODIS (twice a day for each MODIS), supplemented with those inferred from 3-hourly geostationary satellites, reduce the error in the monthly mean surface shortwave and longwave flux. To determine if the SYN1deg maintains its advantage at more than just the two sites considered, we investigate the amplitude and phase of surface shortwave and longwave fluxes at more surface sites.

The amplitude (A) and phase (ϕ) of the diurnal cycles are determined by expanding 3-hourly surface shortwave and longwave fluxes in a Fourier series:
e3
where t is local time, a = A cosϕ, b = −A sinϕ, and tan(2ϕ/24) = −b/a. Note that 1.5 h is subtracted from ϕ because the calculation is registered at the center of the 3-h bin for the SYN1deg and MERRA calculations. In addition, ϕ is multiplied by −1,
e4
so that positive ϕ′ indicates a positive shift in time.

Figures 10a and 10b are scatterplots of phase and amplitude, respectively, for SYN1Deg computed and observed surface downward shortwave flux at 31 land surface sites. Six polar sites are excluded, since cloud properties derived from geostationary satellites are only available between 60°N and 60°S. Figure 10b plots the amplitude difference between monthly mean diurnal cycles against the observed amplitude. Figure 11 shows the same but for surface longwave flux. We remove sites where the amplitude is less than 5 W m−2. The correlation of shortwave phase is large (r = 0.94 in Fig. 10a), as the diurnal cycle is primarily caused by solar zenith angle change. The correlation of the longwave phase is lower (r = 0.39 in Fig. 11a) because the downward longwave flux varies with temperature, humidity, and cloud responses to the solar diurnal cycle. The standard deviation of the longwave amplitude relative to the mean is significantly larger, ~5 out of ~14 W m−2, than the standard deviation of shortwave amplitude relative to the mean 17 out of ~300 W m−2.

Fig. 10.
Fig. 10.

(a) Scatterplot of the phase of the diurnal cycle first harmonic of calculated and observed monthly surface downward shortwave flux. Observed fluxes are from 31 land surface sites. Computed fluxes are from SYN1deg. (b) Scatterplot of the difference between computed and observed amplitude of the diurnal cycle first harmonic for surface downward shortwave flux (computed minus observed) and observed amplitude. “Sig” is the standard deviation and r is the correlation coefficient.

Citation: Journal of Atmospheric and Oceanic Technology 32, 6; 10.1175/JTECH-D-14-00165.1

Fig. 11.
Fig. 11.

As in Fig. 10, but for downward longwave surface flux at the same 31 sites.

Citation: Journal of Atmospheric and Oceanic Technology 32, 6; 10.1175/JTECH-D-14-00165.1

The same analysis is done for all products with 3-hourly temporal resolution and the statistics are shown in Tables 5 and 6 for the shortwave and longwave flux, respectively. The tables show only standard deviations of calculated values from observed for land and ocean buoy sites separately. For the shortwave flux (Table 5), the standard deviations of phase are approximately the same for all four products though SYN does slightly better. The error in the SW amplitude for SYN1Deg is approximately half that of the other products. For longwave, Table 6, the SYN1Deg performs better over land than the other products in terms of both phase and amplitude. Over ocean the ISCCP-FD and SRB3.0 LW products are slightly better in terms of phase, but the amplitude over the buoys is nearly identical for all products except ISSCP FD. This smaller standard deviation in the amplitude of the first shortwave harmonic (Table 5) indicates the improvement in SYN1Deg SW flux diurnal cycle found at BOS and the STR site (Figs. 8 and 9) was not unique, but that the inclusion of CERES GEO clouds enhances the SYN1Deg results at the sites compared here and by inference around the globe.

Table 5.

Standard deviation between calculated and observed shortwave surface downward radiant flux phase (h) and amplitude (W m−2) of monthly mean diurnal cycles.

Table 5.
Table 6.

Standard deviation between calculated and observed longwave surface downward radiant flux phase (h) and amplitude (W m−2) of monthly mean diurnal cycles.

Table 6.

4. Summary and conclusions

The CERES SYN1deg, edition 3, data product provides profiles of calculated upward and downward radiant fluxes at five levels in the atmosphere over a 1° × 1° latitude–longitude grid globally at 3-hourly temporal resolution. Determining surface radiative fluxes requires a radiative transfer model and ancillary inputs of cloud, aerosol, and surface properties as well as atmospheric profiles of temperature, humidity, and ozone. Many of the ancillary inputs used in the calculations are also included as part of the SYN1deg data product. Computed surface radiative fluxes are evaluated with surface observations from 85 land and ocean sites. Over all sites, both land and ocean, the SYN1Deg bias in shortwave and longwave surface downwelling flux is 3.0 and −4 W m−2, respectively. The standard deviation of the difference between SYN1deg computed and observed SW fluxes is 55.5, 31.0, and 11.6 W m−2 for 3-hourly, daily, and monthly time scales, respectively. These errors are smaller on average than 3-hourly, daily, and monthly errors in ISCCP FD, SRB3.0, and MERRA by 29%, 26%, and 44% (relative to the average of the standard deviations of the ISCCP-FD, SRB3.0, and MERRA values), respectively. Since the various products considered here use similar temperature and water vapor inputs (except ISCCP FD, which uses TOVS), the variability in the downward longwave flux (DLF) comparisons to surface observations is relatively small.

An analysis at two surface locations with distinct monthly mean diurnal cycles of longwave and shortwave flux shows that over the 8 years considered, the SYN1deg monthly mean diurnal cycle provides a better match with surface observations compared to the three other data products considered. A harmonic analysis was applied to 31 land surface sites and 48 ocean buoys. Based upon the variance between the phase and amplitude of the first harmonic in the monthly mean diurnal cycles for those groupings of sites, the SYN1deg product retrieved better monthly mean diurnal cycles in shortwave surface flux. Results are only slightly better for the longwave diurnal cycle phase, with little difference in the various products’ longwave diurnal cycle amplitude. This is likely because longwave surface fluxes are more dependent on atmospheric input data, especially near-surface air temperature and humidity, while the cloud contribution to longwave surface flux is less pronounced. Using estimates of water vapor uncertainty from Zhang et al. (2006), Kato et al. (2011) suggest a value of 5.2 W m−2 for the global uncertainty of DLF due to column water vapor and a value of 4.5 W m−2 for the uncertainty in DLF due to surface air temperature. Improvements in the reanalyses that produce the atmospheric inputs are ongoing, and we will use a newer GEOS-5 product for SYN1deg, edition 4.

We attribute the improvement in accuracy in the shortwave surface calculations in SYN1deg to the inclusion of cloud properties derived from multiple 3-hourly GEO imager radiances calibrated with coincident MODIS radiance measurements. The statistics in Tables 3 and 4 show that of the four products with high temporal resolution, the SYN1deg has the smallest overall land and ocean bias and the lowest standard deviations with respect to surface observations. An analysis of the first harmonic of the diurnal cycle at 31 surface sites shows that the SYN1deg has significantly improved the amplitude with respect to surface observations in the shortwave flux compared to the other products (Table 5). The amplitude of the downward shortwave flux is primarily dependent on solar zenith angle and cloud properties (fraction and optical depth). Since all models are able to calculate solar zenith angle well (Table 5), we conclude the improved amplitude relative to the other products is due to improved cloud properties in SYN1deg. Finally, we show in Fig. 6b that the monthly mean bias of shortwave transmission decreases as the standard deviation between observed and calculated 3-hourly transmission decreases. Again, this implies that improving the diurnal cycle of clouds, which in turn improves the diurnal cycle of the downward shortwave flux, reduces the monthly mean bias. This is not the case for the longwave downward flux, as it is more dependent on the quality of the temperature and water vapor vertical profiles used in the radiative transfer models.

Acknowledgments

This work was funded by the NASA CERES project. The products and the validation could not have been accomplished without the help of the CERES TISA team. Data were obtained from the NASA Langley Research Center EOSDIS Distributed Active Archive Center. We also wish to acknowledge the hard work by the many dedicated scientists maintaining surface instrumentation in diverse climates to obtain high-quality observations of downwelling shortwave and longwave surface flux. Those groups are noted in appendix B. We also thank the reviewers, whose close read of the paper greatly improved the end result.

APPENDIX A

Spectral Band Limits for Langley Fu and Liou Radiative Transfer Model

Tables A1 and A2 show the spectral limits for the 18 shortwave and 12 longwave spectral bands along with gases treated in each band for the Langley Fu and Liou radiative transfer model used in calculations for the SYN1deg data product. The correlated-k method in the Langley Fu and Liou model groups absorption lines of similar strength within each band into subbands as shown in the second column of each table. Two additional thermal LW bands are added (between 2200 and 2500, and between 2500 and 2850 cm−1) to treat thermal radiation from hot surfaces, such as daytime deserts, where the Planck curve begins to have nonnegligible energy in that wavenumber range.

Table A1.

Shortwave spectral bands for Langley Fu and Liou model used in this work.

Table A1.
Table A2.

Longwave spectral bands for Langley Fu and Liou model used in this work.

Table A2.

APPENDIX B

Surface Observation Databases

Several important sources of surface-observed downwelling radiant flux included in this study are outlined in Table B1 (land) and Table B2 (buoys). For land and island sites, we utilize data from the Baseline Surface Radiation Network (Ohmura et al. 1998), NOAA’s Global Monitoring Division (GMD) (Augustine et al. 2000), and the U.S. Department of Energy’s Atmospheric Radiation Measurement (ARM) Program as well as NOAA’s GMD data, which are made available through the NOAA/GMD Solar and Thermal Radiation (STAR) group. SURFRAD data are made available through NOAA’s Air Resources Laboratory/Surface Radiation Research Branch. Buoy observations come from two sources through four separate projects. The Upper Ocean Processes (UOP) group at Woods Hole Oceanographic Institution has maintained the Stratus, the North Tropical Atlantic Site (NTAS), and the Hawaii Ocean Time Series (HOTS) buoys for a decade or more, providing valuable time series of radiation observations in climatically important regions of the ocean. We would also like to acknowledge the project office of NOAA’s Pacific Marine Environmental Laboratory, where three groups of buoy data were downloaded: In the Pacific, the Tropical Atmosphere Ocean/Triangle Trans-Ocean Buoy Network (TAO/TRITON) (McPhaden 2002) data; from the tropical Atlantic Ocean, the Prediction and Research Moored Array in the Tropical Atlantic (PIRATA) (Servain et al. 1998); and the Research Moored Array for African–Asian–Australian Monsoon Analysis and Prediction (RAMA) (McPhaden et al. 2009) in the Indian Ocean.

Table B1.

Surface observation sites for land/island locations. WCRP: World Climate Research Programme. Penn State: The Pennsylvania State University.

Table B1.
Table B2.

Surface observation sites for ocean buoy locations.

Table B2.

APPENDIX C

Sample Representation of Statistics

Surface-observed radiant fluxes are commonly used to assess the quality of radiative transfer calculations from GCMs, reanalyses, and satellite data products that produce surface flux estimates. However, resultant statistics may not always be as “normal” as they appear, as different surface sites often have unique bias characteristics with respect to calculated surface flux. Thus, should a site be added to or removed from a study, the effect is not a simple random sampling of the overall bias distribution. Figure C1 shows the distribution of monthly mean bias (SYN1deg minus observations) of longwave surface flux down at 56 surface sites (black line) that appears to be a near-normal distribution with a bias of −4.1 W m−2. Red and blue lines show the bias distributions for two sites that make up the larger curve from Ny Alesund, Norway, and Alice Springs, Australia, respectively. It is self-evident that removing or adding such individual sites impacts the bias distribution in a nonrandom manner. General statistical theory states that as a population is randomly subsampled, the standard error of the mean is represented as . But, removing a site is not randomly sampling the original distribution, so this theory is not directly applicable for this type of comparison.

Fig. C1.
Fig. C1.

Bias distribution of monthly mean longwave downwelling flux from 56 surface sites (black line) and from the Ny Alesund site (red line) and the Alice Springs site (blue line).

Citation: Journal of Atmospheric and Oceanic Technology 32, 6; 10.1175/JTECH-D-14-00165.1

Here the standard error of the mean is instead derived empirically. To do this we progressively remove sites, resampling the original distribution 500 times with the new number of sites. We use a simple bootstrap statistic where we rerun the comparisons of the SYN1deg observation to calculate across the same 8 years of data, for four test cases, where for each case we remove four more sites. Thus, for each case we remove 4, 8, 12, and 16 sites, randomly, 500 times, ensuring that none of the 500 samples removes the identical set of sites. We plot the final distributions for these four tests (at monthly mean time resolution) with the expectation that as more surface sites are removed, we are less and less likely to retrieve our original statistics.

Figure C2 shows the probability distributions for longwave biases (Fig. C2a) and standard deviations (Fig. C2b) for the test cases. The original bias (sigma) using all sites is shown as a vertical dashed line. The distribution of the monthly mean bias is the most robust statistic. However, it is clear that by removing 8–16 sites, the distributions become increasingly flat. The distribution of standard deviations also falls off quickly as sites are removed, though their spread (as measured by the variance of the distribution) does not increase as rapidly at the monthly mean time scale.

Fig. C2.
Fig. C2.

Bootstrap distributions (500 realizations) of monthly mean bias (calculation minus observations) where original the distribution of sites is reduced by randomly removing all realizations from 4, 8, 12, etc. surface sites: (a) bias distribution and (b) standard deviation distribution. Each of the 500 samples has a different total n, since each site has a distinct number of monthly mean values available. The standard deviation of each distribution represents the empirical standard error of the mean increasing as more and more sites are removed. The gray dashed line is the original bias and standard deviation of the dataset.

Citation: Journal of Atmospheric and Oceanic Technology 32, 6; 10.1175/JTECH-D-14-00165.1

Figure C3 shows the empirical standard error of the mean (sigma) calculated from the bootstrapped distributions. The blue lines (hypothetical) are based on the idea that if all site/months were available and if one could sample the bias distribution randomly, then these lines represent the values. The black lines show the empirical results approaching the hypothetical as the number of sites increases. Figure C3 indicates that increasing surface sites from 30 to 55 (LW) or from 60 to 80 (SW) dramatically reduces the standard error of the monthly mean fluxes. As demonstrated in Fig. C1, the bias error of computed flux depends on geographical location. Subsampling 55 (LW) or 80 (SW) sites systematically excludes some sites that have a larger or smaller bias error compared to the mean bias error. This leads to a larger standard error than the standard error when the bias error is distributed randomly across all sites. If validation sites are concentrated in one geographical location, then the bias error might be distributed randomly across the sites. Another implication of this result is, therefore, that the validation sites included in this study cover sufficiently different geographical regions that give a range of bias errors.

Fig. C3.
Fig. C3.

Empirical standard error of the mean derived from LW (solid) and SW (dashed) bias (black) and standard deviation (red) bootstrap distributions. Blue lines represent estimate of a “true” standard error of the mean.

Citation: Journal of Atmospheric and Oceanic Technology 32, 6; 10.1175/JTECH-D-14-00165.1

Since individual sites act on the overall distribution nonrandomly, it is important to include as wide a variety of geographical regions as possible for robust results. Figure C3 shows that reducing the number of sites moves the statistics in the wrong direction. While satellite-derived, reanalyses, and atmospheric modeling of radiative fluxes are pushed to shorter time scales (3-hourly and hourly comparisons), this analysis shows that to validate such results requires long-term surface site observations at a wide range of geographical regions.

REFERENCES

  • Augustine, J. A., , DeLuisi J. J. , , and Long C. N. , 2000: SURFRAD – A national surface radiation budget network for atmospheric research. Bull. Amer. Meteor. Soc., 81, 23412358, doi:10.1175/1520-0477(2000)081<2341:SANSRB>2.3.CO;2.

    • Search Google Scholar
    • Export Citation
  • Barker, H. W., 1996: A parameterization for computing grid-averaged solar fluxes for inhomogeneous marine boundary layer clouds. Part I: Methodology and homogeneous biases. J. Atmos. Sci., 53, 22892303, doi:10.1175/1520-0469(1996)053<2289:APFCGA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Bergman, J. W., , and Salby M. L. , 1996: Diurnal variations of cloud cover and their relationship to climatological conditions. J. Climate, 9, 28022820, doi:10.1175/1520-0442(1996)009<2802:DVOCCA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Bloom, S., and et al. , 2005: Documentation and validation of the Goddard Earth Observing System (GEOS) Data Assimilation System—Version 4. M. J. Suarez, Ed., Technical Report Series on Global Modeling and Data Assimilation, Vol. 26, NASA Tech. Rep. NASA/TM-2005-104606, 187 pp.

  • Cairns, B., 1995: Diurnal variation of cloud from ISCCP data. Atmos. Res., 37, 133146, doi:10.1016/0169-8095(94)00074-N.

  • Collins, W. D., , Rasch P. J. , , Eaton B. E. , , Khattatov B. V. , , Lamarque J.-F. , , and Zender C. S. , 2001: Simulating aerosols using a chemical transport model with assimilation of satellite aerosol retrievals: Methodology for INDOEX. J. Geophys. Res., 106, 73137336, doi:10.1029/2000JD900507.

    • Search Google Scholar
    • Export Citation
  • Dee, D. P., and et al. , 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553597, doi:10.1002/qj.828.

    • Search Google Scholar
    • Export Citation
  • Dilley, A. C., , and O’Brien D. M. , 1998: Estimating downward clear sky long-wave irradiance at the surface from screen temperature and precipitable water. Quart. J. Roy. Meteor. Soc., 124, 13911401, doi:10.1002/qj.49712454903.

    • Search Google Scholar
    • Export Citation
  • Doelling, D. R., and et al. , 2013: Geostationary enhanced temporal interpolation for CERES flux products. J. Atmos. Oceanic Technol., 30, 10721090, doi:10.1175/JTECH-D-12-00136.1.

    • Search Google Scholar
    • Export Citation
  • Fu, Q., , and Liou K.-N. , 1993: Parameterization of the radiative properties of cirrus clouds. J. Atmos. Sci., 50, 20082025, doi:10.1175/1520-0469(1993)050<2008:POTRPO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Fu, Q., , Lesins G. , , Higgins J. , , Charlock T. , , Chylek P. , , and Michalsky J. , 1998: Broadband water vapor absorption of solar radiation tested using ARM data. Geophys. Res. Lett., 25, 11691172, doi:10.1029/98GL00846.

    • Search Google Scholar
    • Export Citation
  • Henderson, D. S., , L’Ecuyer T. , , Stephens G. , , Partain P. , , and Sekiguchi M. , 2013: A multisensor perspective on the radiative impacts of clouds and aerosols. J. Appl. Meteor. Climatol, 52, 853871, doi:10.1175/JAMC-D-12-025.1.

    • Search Google Scholar
    • Export Citation
  • Hess, M., , Koepke P. , , and Schult I. , 1998: Optical properties of aerosols and clouds: The software package OPAC. Bull. Amer. Meteor. Soc., 79, 831844, doi:10.1175/1520-0477(1998)079<0831:OPOAAC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Jin, Z., , Charlock T. P. , , Smith W. L. Jr., , and Rutledge K. , 2004: A parameterization of ocean surface albedo. Geophys. Res. Lett., 31, L22301, doi:10.1029/2004GL021180.

    • Search Google Scholar
    • Export Citation
  • Kato, S., , Ackerman T. , , Mather J. , , and Clothiaux E. , 1999: The k-distribution method and correlated-k approximation for a shortwave radiative transfer model. J. Quant. Spectrosc. Radiat. Transfer, 62, 109121, doi:10.1016/S0022-4073(98)00075-2.

    • Search Google Scholar
    • Export Citation
  • Kato, S., , Rose F. G. , , and Charlock T. P. , 2005: Computation of domain-averaged irradiance using satellite-derived cloud properties. J. Atmos. Oceanic Technol., 22, 146164, doi:10.1175/JTECH-1694.1.

    • Search Google Scholar
    • Export Citation
  • Kato, S., and et al. , 2011: Improvements of top‐of‐atmosphere and surface irradiance computations with CALIPSO‐, CloudSat‐, and MODIS‐derived cloud and aerosol properties. J. Geophys. Res., 116, D19209, doi:10.1029/2011JD016050.

    • Search Google Scholar
    • Export Citation
  • Kato, S., , Loeb N. G. , , Rose F. G. , , Doelling D. R. , , Rutan D. A. , , Caldwell T. E. , , Yu L. , , and Weller R. A. , 2013: Surface irradiances consistent with CERES-derived top-of-atmosphere shortwave and longwave irradiances. J. Climate,26, 2719–2740, doi:10.1175/JCLI-D-12-00436.1.

  • Kratz, D. P., , and Rose F. G. , 1999: Accounting for molecular absorption within the spectral range of the CERES window channel. J. Quant. Spectrosc. Radiat. Transfer, 61, 8395, doi:10.1016/S0022-4073(97)00203-3.

    • 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, doi:10.1175/2008JCLI2637.1.

    • Search Google Scholar
    • Export Citation
  • Loeb, N. G., , Kato S. , , Su W. , , Wong T. , , Rose F. G. , , Doelling D. R. , , Norris J. R. , , and Huang X. , 2012: Advances in understanding top-of-atmosphere radiation variability from satellite observations. Surv. Geophys., 33, 359–385, doi:10.1007/s10712-012-9175-1.

    • Search Google Scholar
    • Export Citation
  • Long, C. N., , and Ackerman T. P. , 2000: Identification of clear skies from broadband pyranometer measurements and calculation of downwelling shortwave cloud effects. J. Geophys. Res., 105, 15 60915 626, doi:10.1029/2000JD900077.

    • Search Google Scholar
    • Export Citation
  • McPhaden, M. J., 2002: TAO/TRITON tracks Pacific Ocean warming in early 2002. CLIVAR Exchanges, No. 24, International CLIVAR Project Office, Southampton, United Kingdom, 7–9.

  • McPhaden, M. J., and et al. , 2009: RAMA: The Research Moored Array for African–Asian–Australian Monsoon Analysis and Prediction. Bull. Amer. Meteor. Soc., 90, 459480, doi:10.1175/2008BAMS2608.1.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., , and Harrison E. F. , 1984: Diurnal variability of regional cloud and clear-sky radiative parameters derived from GOES data, Part II: November 1978 cloud results. J. Climate Appl. Meteor., 23, 10121031, doi:10.1175/1520-0450(1984)023<1012:DVORCA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Minnis, P., , Smith W. L. Jr., , Garber D. P. , , Ayers J. K. , , and Doelling D. R. , 1995: 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, 59 pp.

  • Minnis, P., and et al. , 2011a: 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
  • Minnis, P., and et al. , 2011b: CERES Edition-2 cloud property retrievals using TRMM VIRS and Terra and Aqua MODIS data—Part II: Examples of average results and comparisons with other data. IEEE Trans. Geosci. Remote Sens., 49, 4401–4430, 10.1109/TGRS.2011.2144602.

    • Search Google Scholar
    • Export Citation
  • Ohmura, A., and et al. , 1998: Baseline Surface Radiation Network (BSRN/WCRP): New precision radiometry for climate change research. Bull. Amer. Meteor. Soc., 79, 21152136, doi:10.1175/1520-0477(1998)079<2115:BSRNBW>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Oreopoulos, L., , and Barker H. W. , 1999: Accounting for subgrid-scale cloud variability in a multi-layer 1D solar radiative transfer algorithm. Quart. J. Roy. Meteor. Soc., 125, 301330, doi:10.1002/qj.49712555316.

    • Search Google Scholar
    • Export Citation
  • Pinker, R. T., , and Laszlo I. , 1992: Modeling surface solar irradiance for satellite applications on a global scale. J. Appl. Meteor., 31, 194211, doi:10.1175/1520-0450(1992)031<0194:MSSIFS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Remer, L. A., and et al. , 2005: The MODIS aerosol algorithm, products, and validation. J. Atmos. Sci., 62, 947973, doi:10.1175/JAS3385.1.

    • Search Google Scholar
    • Export Citation
  • Rienecker, M. M., and et al. , 2011: MERRA: NASA's Modern-Era Retrospective Analysis for Research and Applications. J. Climate, 24, 36243648, doi:10.1175/JCLI-D-11-00015.1.

    • Search Google Scholar
    • Export Citation
  • Roemmich, D., , and Gilson J. , 2009: The 2004–2008 mean and annual cycle of temperature, salinity, and steric height in the global ocean from the Argo Program. Prog. Oceanogr.,82, 81–100, doi:10.1016/j.pocean.2009.03.004.

  • Rose, F. G., , Charlock T. , , Fu Q. , , Kato S. , , Rutan D. , , and Jin Z. , 2006: CERES proto-edition_3 radiative transfer: Model tests and radiative closure over surface validation sites. 12th Conf. on Atmospheric Radiation, Madison, WI, Amer. Meteor. Soc., P2.4. [Available online at https://ams.confex.com/ams/Madison2006/techprogram/paper_112358.htm.]

  • Rose, F. G., , Rutan D. A. , , Charlock T. , , Smith G. L. , , and Kato S. , 2013: An algorithm for the constraining of radiative transfer calculations to CERES-observed broadband top-of-atmosphere irradiance. J. Atmos. Oceanic Technol., 30, 10911106, doi:10.1175/JTECH-D-12-00058.1.

    • Search Google Scholar
    • Export Citation
  • Rossow, W. B., , and Schiffer R. A. , 1991: ISCCP cloud data products. Bull. Amer. Meteor. Soc., 72, 220, doi:10.1175/1520-0477(1991)072<0002:ICDP>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Rossow, W. B., , and Schiffer R. A. , 1999: Advances in understanding clouds from ISCCP. Bull. Amer. Meteor. Soc., 80, 22612287, doi:10.1175/1520-0477(1999)080<2261:AIUCFI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Rozendaal, M. A., , Leovy C. B. , , and Klein S. A. , 1995: An observational study of diurnal variations of marine stratiform cloud. J. Climate, 8, 17951809, doi:10.1175/1520-0442(1995)008<1795:AOSODV>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Rutan, D., , Rose F. , , Roman M. , , Manalo-Smith N. , , Schaaf C. , , and Charlock T. , 2009: Development and assessment of broadband surface albedo from Clouds and the Earth’s Radiant Energy System clouds and radiation swath data product. J. Geophys. Res., 114, D08125, doi:10.1029/2008JD010669.

    • Search Google Scholar
    • Export Citation
  • Servain, J., , Busalacchi A. J. , , McPhaden M. J. , , Moura A. D. , , Reverdin G. , , Vianna M. , , and Zebiak S. E. , 1998: A Pilot Research Moored Array in the Tropical Atlantic (PIRATA). Bull. Amer. Meteor. Soc., 79, 20192031, doi:10.1175/1520-0477(1998)079<2019:APRMAI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Slingo, A., , Hodges K. I. , , and Robinson G. J. , 2004: Simulation of the diurnal cycle in a climate model and its evaluation using data from Meteosat 7. Quart. J. Roy. Meteor. Soc., 130, 14491467, doi:10.1256/qj.03.165.

    • Search Google Scholar
    • Export Citation
  • Smith, G. L., , Priestley K. J. , , Loeb N. G. , , Wielicki B. A. , , Charlock T. P. , , Minnis P. , , Doelling D. R. , , and Rutan D. A. , 2011: Clouds and Earth Radiant Energy System (CERES), a review: Past, present and future. Adv. Space Res., 48, 254263, doi:10.1016/j.asr.2011.03.009.

    • Search Google Scholar
    • Export Citation
  • Stackhouse, P. W., Jr., , Gupta S. K. , , Cox S. J. , , Zhang T. , , Mikovitz J. C. , , and Hinkelman L. M. , 2011: 24.5-year SRB data set released. GEWEX News, No. 1, International GEWEX Project Office, Silver Spring, MD, 10–12.

  • Stephens, G. L., and et al. , 2012: An update on Earth’s energy balance in light of the latest global observations. Nat. Geosci., 5, 681–696, doi:10.1038/ngeo1580.

    • Search Google Scholar
    • Export Citation
  • Swinbank, W. C., 1963: Long-wave radiation from clear skies. Quart. J. Roy. Meteor. Soc., 89, 339348, doi:10.1002/qj.49708938105.

  • Taylor, K. E., 2001: Summarizing multiple aspects of model performance in a single diagram. J. Geophys. Res., 106, 71837192, doi:10.1029/2000JD900719.

    • Search Google Scholar
    • Export Citation
  • Taylor, P. C., 2012: Tropical outgoing longwave radiation and longwave cloud forcing diurnal cycles from CERES. J. Atmos. Sci., 69, 36523669, doi:10.1175/JAS-D-12-088.1.

    • Search Google Scholar
    • Export Citation
  • Tegen, I., , and Lacis A. A. , 1996: Modeling of particle size distribution and its influence on the radiative properties of mineral dust aerosol. J. Geophys. Res., 101, 19 23719 244, doi:10.1029/95JD03610.

    • Search Google Scholar
    • Export Citation
  • Toon, O. B., , McKay C. P. , , Ackerman T. P. & , and Santhanam K. , 1989: Rapid calculation of radiative heating rates and photodissociation rates in inhomogeneous multiple scattering atmospheres. J. Geophys. Res., 94, 16 28716 301, doi:10.1029/JD094iD13p16287.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., , Fasullo J. T. , , and Kiehl J. , 2009: Earth’s global energy budget. Bull. Amer. Meteor. Soc., 90, 311323, doi:10.1175/2008BAMS2634.1.

    • 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, doi:10.1175/1520-0477(1996)077<0853:CATERE>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Wilber, A. C., , Kratz D. P. , , Gupta S. K. , 1999: Surface emissivity maps for use in satellite retrievals of longwave radiation. NASA Tech. Publ. NASA/TP-1999-209362, 35 pp.

  • Yang, S.-K., , Zhou S. , , and Miller A. J. , 1998: SMOBA: A 3-dimensional daily ozone analysis using SBUV/2 and TOVS measurements. Accessed 6 January 2015. [Available online at http://www.cpc.ncep.noaa.gov/products/stratosphere/SMOBA/smoba_doc.shtml.]

  • Zhang, Y.-C., , Rossow W. B. , , and Lacis A. A. , 1995: Calculation of surface and top of atmosphere radiative fluxes from physical quantities based on ISCCP data sets: 1. Method and sensitivity to input data uncertainties. J. Geophys. Res., 100, 11491165, doi:10.1029/94JD02747.

    • Search Google Scholar
    • Export Citation
  • Zhang, Y.-C., , Rossow W. B. , , Lacis A. A. , , Oinas V. , , and Mishchenko M. I. , 2004: Calculation of radiative fluxes from the surface to top of atmosphere based on ISCCP and other global data sets: Refinements of the radiative transfer model and the input data. J. Geophys. Res.,109, D19105, doi:10.1029/2003JD004457.

  • Zhang, Y.-C., , Rossow W. B. , , and Stackhouse P. W. Jr., 2006: Comparison of different global information sources used in surface radiative flux calculation: Radiative properties of the near-surface atmosphere. J. Geophys. Res., 111, D13106, doi:10.1029/2005JD006873.

    • Search Google Scholar
    • Export Citation
Save