• Abel, P. G., , and Gruber A. , 1979: An improved model for the calculation of longwave flux at 11 μm. NOAA Tech. Rep. NESS 106, 24 pp. [NTIS PB80-119431.].

    • Search Google Scholar
    • Export Citation
  • Cess, R. D., , and Vulis I. L. , 1989: Intercomparison and interpretation of satellite derived directional albedos over deserts. J. Climate, 2 , 393407.

    • Search Google Scholar
    • Export Citation
  • Clough, S. A., , Kneizys F. X. , , and Davies R. W. , 1989: Line shape and the water vapor continuum. Atmos. Res., 23 , 229241.

  • Ellingson, R. G., , Yanuk D. J. , , Lee H-T. , , and Gruber A. , 1989: A technique for estimating outgoing longwave radiation from HIRS radiance observations. J. Atmos. Oceanic Technol., 6 , 706711.

    • Search Google Scholar
    • Export Citation
  • Ellingson, R. G., , Lee H-T. , , Yanuk D. , , and Gruber A. , 1994: Validation of a technique for estimating outgoing longwave radiation from HIRS radiance observations. J. Atmos. Oceanic Technol., 11 , 357365.

    • Search Google Scholar
    • Export Citation
  • Gruber, A., , Ellingson R. G. , , Ardanuy P. , , Weiss M. , , Yang S. K. , , and Oh S. , 1994: A comparison of ERBE and AVHRR longwave flux estimates. Bull. Amer. Meteor. Soc., 75 , 21152130.

    • Search Google Scholar
    • Export Citation
  • Haurwitz, F., , and Kuhn W. R. , 1974: The distribution of tropospheric planetary radiation in the Southern Hemisphere. J. Appl. Meteor., 13 , 417429.

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

    • Search Google Scholar
    • Export Citation
  • Ma, X. L., , Schmit T. J. , , and Smith W. L. , 1999: A nonlinear physical retrieval algorithm—Its application to the GOES-8/9 Sounder. J. Appl. Meteor., 38 , 501513.

    • Search Google Scholar
    • Export Citation
  • Malkmus, R., 1997: Random Lorentz band model with exponential-tailed S−1 line intensity distribution function. J. Opt. Soc. Amer., 57 , 323329.

    • Search Google Scholar
    • Export Citation
  • Menzel, W. P., , Holt F. C. , , Schmit T. J. , , Aune R. M. , , Schreiner A. J. , , Wade G. S. , , and Gray D. G. , 1998: Application of GOES-8/9 soundings to weather forecasting and nowcasting. Bull. Amer. Meteor. Soc., 79 , 20592077.

    • Search Google Scholar
    • Export Citation
  • Morel, P., 1985: Plan for the implementation of the World Climate Research Programme. Adv. Space Res., 5 (6) 1520.

  • NAS/NRC, 1983: Toward an international geosphere–biosphere program. National Research Council Workshop Rep., National Academy Press, Washington, D.C.

    • Search Google Scholar
    • Export Citation
  • Phillips, N., , Susskind J. , , and McMillin L. , 1988: Results of a joint NOAA/NASA sounder simulation study. J. Atmos. Oceanic Technol., 5 , 4456.

    • Search Google Scholar
    • Export Citation
  • Rodgers, C. D., , and Walshaw C. D. , 1966: The computation of infrared cooling rates in planetary atmospheres. Quart. J. Roy. Meteor. Soc., 92 , 6792.

    • Search Google Scholar
    • Export Citation
  • Rothman, L. S., and Coauthors. 1992: The HITRAN molecular database: Editions of 1991 and 1992. J. Quant. Spectrosc. Radiat. Trans., 48 , 496508.

    • Search Google Scholar
    • Export Citation
  • Schmit, T. J., , Feltz W. F. , , Menzel W. P. , , Jung J. , , Noel A. P. , , Heil J. N. , , Nelson J. P. III, , and Wade G. S. , 2002: Validation and use of GOES sounder moisture information. Wea. Forecasting, 17 , 139154.

    • Search Google Scholar
    • Export Citation
  • Schreiner, A. J., , Schmit T. J. , , and Menzel W. P. , 2001: Observations and trends of clouds based on GOES sounder data. J. Geophys. Res., 106 , 2034920363.

    • Search Google Scholar
    • Export Citation
  • SSEC, cited 1997: McIDAS user's guide. Space Science and Engineering Center, University of Wisconsin—Madison. [Available online at www.ssec.wisc.edu/mug.].

    • Search Google Scholar
    • Export Citation
  • Warner, J. X., , and Ellingson R. G. , 2000: A new narrowband radiation model for water vapor absorption. J. Atmos. Sci., 57 , 14811496.

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

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

    • Search Google Scholar
    • Export Citation
  • WMO, 1988: Concept of the Global Energy and Water Cycle Experiment: Report of the JSC Study Group on GEWEX. WMO Tech. Doc. 215, WCRP-5, 70 pp.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    (a) CERES TRMM orbital passes remapped onto an earth-referenced 10 km × 10 km fixed grid 1600 UTC on 2 Jul 1998. (b) Spatially interpolated GOES observations on the CERES TRMM orbit

  • View in gallery

    Scatterplot of GOES and TRMM CERES OLR data from 100 km × 100 km homogeneous scenes for Jul 1998 for all scenes.

  • View in gallery

    Scatterplot of GOES and Terra CERES FM1 OLR data from 100 km × 100 km homogeneous scenes for Apr 2000 for all scenes

  • View in gallery

    Scatterplot of GOES and Terra CERES FM2 OLR data from 100 km × 100 km homogeneous scenes for Apr 2000 for all scenes

  • View in gallery

    Scatterplots of CERES OLR and SW fluxes from 100 km × 100 km homogeneous scenes for Jul 1998 over the Pacific Ocean off the coast of Baja California: (a) CERES version-1 fluxes, and (b) CERES version-2 fluxes

  • View in gallery

    Scatterplot of GOES-8 vs TRMM CERES monthly averaged OLR for 50 km × 50 km instantaneous collocated scenes

  • View in gallery

    Scatterplot of GOES-8 and -10 vs Terra CERES FM1 monthly averaged OLR for 50 km × 50 km instantaneous collocated scenes

  • View in gallery

    Scatterplot of GOES-8 and -10 vs Terra CERES FM2 monthly averaged OLR for 50 km × 50 km instantaneous collocated scenes

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 0 0 0
PDF Downloads 0 0 0

Validation of a Technique for Estimating OLR with the GOES Sounder

View More View Less
  • 1 Cooperative Institute for Climate Studies, University of Maryland at College Park, College Park, Maryland
  • 2 NOAA/NESDIS, Camp Springs, Maryland
© Get Permissions
Full access

Abstract

In order to eventually use the capability of the Geostationary Operational Environmental Satellite (GOES) Sounder to capture the diurnal signal of outgoing longwave radiation (OLR), it is necessary to establish its instantaneous accuracy. Error characteristics of OLR derived from the GOES Sounder are analyzed using Clouds and Earth's Radiant Energy System (CERES) observations. The comparisons are based on over 28 000 data collected in July 1998 and April 2000 over the continental United States. The July data correspond to observations from GOES-8 and -9 and the CERES instrument on board the Tropical Rainfall Measurement Mission (TRMM) satellite. The April data correspond to GOES-8 and -10, and two CERES instruments on board the Terra satellite. The comparisons are for instantaneous, homogeneous scenes of 1° × 1° boxes. Comparisons of GOES Sounder with collocated TRMM and Terra CERES OLR show instantaneous rms agreement to within about 7 W m−2 for day and/or night homogeneous scenes. The GOES technique explained over 91% and 96% of the variance of CERES observations for both night and day, and for both land and ocean scenes for July 1998 and April 2000, respectively.

Current affiliation: Raytheon ITSS, Lanham, Maryland

+ Current affiliation: Department of Meteorology, The Florida State University, Tallahassee, Florida

Corresponding author affiliation: Dr. Robert Ellingson, Department of Meteorology, Room 404, Love Building, The Florida State University, Tallahassee, FL 32306. Email: bobe@met.fsu.edu

Abstract

In order to eventually use the capability of the Geostationary Operational Environmental Satellite (GOES) Sounder to capture the diurnal signal of outgoing longwave radiation (OLR), it is necessary to establish its instantaneous accuracy. Error characteristics of OLR derived from the GOES Sounder are analyzed using Clouds and Earth's Radiant Energy System (CERES) observations. The comparisons are based on over 28 000 data collected in July 1998 and April 2000 over the continental United States. The July data correspond to observations from GOES-8 and -9 and the CERES instrument on board the Tropical Rainfall Measurement Mission (TRMM) satellite. The April data correspond to GOES-8 and -10, and two CERES instruments on board the Terra satellite. The comparisons are for instantaneous, homogeneous scenes of 1° × 1° boxes. Comparisons of GOES Sounder with collocated TRMM and Terra CERES OLR show instantaneous rms agreement to within about 7 W m−2 for day and/or night homogeneous scenes. The GOES technique explained over 91% and 96% of the variance of CERES observations for both night and day, and for both land and ocean scenes for July 1998 and April 2000, respectively.

Current affiliation: Raytheon ITSS, Lanham, Maryland

+ Current affiliation: Department of Meteorology, The Florida State University, Tallahassee, Florida

Corresponding author affiliation: Dr. Robert Ellingson, Department of Meteorology, Room 404, Love Building, The Florida State University, Tallahassee, FL 32306. Email: bobe@met.fsu.edu

1. Introduction

Information on the top of the atmosphere and surface radiation budgets can be used to address different aspects of climate issues, such as developing and improving climate model parameterizations with respect to surface–atmosphere interactions (Morel 1985), separating the radiative impact of clouds into contributions to the surface and to the atmosphere (Cess and Vulis 1989); understanding the global hydrological cycle (WMO 1988), and contributing to the International Geosphere–Biosphere Program (IGBP) effort (NAS/NRC 1983). Furthermore, many different World Climate Research Programme (WCRP) projects [e.g., Tropical Ocean Global Atmosphere (TOGA) and the Global Energy and Water Cycle Experiment (GEWEX)] have required and will require detailed surface and atmospheric radiation budget data.

The National Oceanic and Atmospheric Administration (NOAA) operationally produces monthly estimates of the global distribution of the outgoing longwave radiation (OLR) at 2.5° × 2.5° from the Advanced Very High-Resolution Radiometer (AVHRR) and the High-Resolution Infrared Sounder 2 (HIRS2) instruments onboard the NOAA polar orbiters. The AVHRR OLR data date to 1979 and those from HIRS2 to 1993. Neither the AVHRR or HIRS2 directly measure the OLR. Instead, radiances from the AVHRR and HIRS2 instruments are converted to fluxes through the use of algorithms based on theoretical model calculations (e.g., AVHRR, Abel and Gruber 1979; HIRS2, Ellingson et al. 1989). The AVHRR technique uses data from one spectral interval in the 10-μm window that is primarily sensitive to the temperature of the lowest viewed radiating surface (i.e., surface or cloud top). The HIRS2 technique combines radiances from four different spectral intervals that sense contributions from the radiating surface and the vertical distribution of temperature and moisture above that surface.

OLR data from the HIRS2 technique were compared with NOAA9 Earth Radiation Budget Experiment (ERBE) scanner data (Ellingson et al. 1994), and they were found to have similar accuracy—about 5 W m−2 rms for instantaneous 104 km2 homogeneous scenes. The HIRS2 technique overcomes many of the shortcomings of the AVHRR technique (see Gruber et al. 1994), and OLR data are produced with it along with the AVHRR as part of the NOAA operational OLR product.

When the National Polar-Orbiting Operational Environmental Satellite System (NPOESS) is launched later this decade, NOAA will add broadband flux data from a CERES (Clouds and Earth's Radiant Energy System; Wielicki et al. 1996; Loeb et al. 2001) instrument to their OLR product. The broadband data will help alleviate some of the uncertainties associated with the HIRS2 and AVHRR techniques, but there will be temporal sampling problems associated with each of the instruments. Since many scientific studies have made and are likely to continue to make use of the NOAA, ERBE, and CERES OLR data, it is important to establish the magnitude of uncertainties associated with neglected physics and to account for the deficiencies where possible. The effect of not resolving the diurnal cycle is of particular concern.

With the addition of the Sounder instrument to the NOAA Geostationary Operational Environmental Satellites (GOES)-8, -9, and -10 (see Menzel et al. 1998) came the possibility of having HIRS-like OLR at high temporal resolution, thereby allowing for the study of the diurnal cycle of OLR over many different surface types. OLR at high temporal resolution can also be computed using various combination of temperature and moisture profiles (and skin temperature), parameters that are physically retrieved from sounder data (e.g., Ma et al. 1999). The GOES Sounder has similar spectral characteristics to HIRS2, but it has much higher spatial and temporal resolution (GOES: 10 km at satellite subpoint, 1 h; HIRS2: about 17 km at nadir, 6 h for two-satellite system), and data from it are almost spatially contiguous, whereas those from HIRS2 are not. Furthermore, the GOES always looks at a given location at approximately the same angle of view, whereas the HIRS2 view changes due to the nature of the polar orbit and the cross track scanning.

Due to the instrument and sampling differences between the HIRS2 and the GOES Sounder, one cannot simply transfer the algorithms from the HIRS2 to the Sounder and assume that the HIRS2 OLR error characteristics established by Ellingson et al. (1994) will hold for the Sounder. Fortunately, OLR data collected by the CERES instruments on board the Tropical Rainfall Measuring Mission (TRMM) and Terra satellites offered the possibility for establishing the error characteristics of GOES Sounder OLR estimates. This paper is directed at presenting those characteristics. An additional paper will present results concerning the diurnal cycle of OLR.

In the material that follows, section 2 describes the OLR estimation technique, section 3 summarizes the data sources and selection procedures, section 4 discusses the results of comparisons of CERES and Sounder OLR, and section 5 presents our conclusions.

2. The GOES Sounder OLR estimation technique

Details concerning the multispectral approach for OLR are given in Ellingson et al. (1989, 1994). Here, the OLR is estimated from GOES sounder radiances as the weighted sum of observations in five spectral intervals, given as
i1520-0426-20-1-79-e1
where the as are coefficients determined from a regression analysis of model calculations, and N is the observed radiance in the ith interval at the satellite zenith angle θ. The coefficients used in Eq. (1) may be obtained from the authors electronically.

The GOES spectral intervals and regression coefficients were determined by regression analyses of calculations from a version of the radiation model of Warner and Ellingson (2000) using 1596 soundings from Phillips et al. (1988) as input, assuming no air–skin temperature discontinuity. The model uses an empirical transmission for water vapor fit to line-by-line model calculations in approximately 10 cm−1 intervals from 0 to 3000 cm−1 (see Warner and Ellingson 2000). The transmissions by carbon dioxide, ozone, methane, and nitrous oxide are determined through the use of the Malkmus (1997) transmittance model fit to 1992 High-Resolution Transmission Molecular Absorption Database (HITRAN) spectral data (Rothman et al. 1992). The water vapor continuum is specified with the Clough–Kneizys–Davies (CKD) version 2.2 defined by Clough et al. (1989). The necessary integrations over altitude and zenith angles are performed with trapezoidal and Gaussian quadratures, respectively. The treatment of clouds closely follows the techniques and procedures used to develop the HIRS2 OLR estimation technique (Ellingson et al. 1989). Liquid water clouds are treated as blackbodies. For cirrus clouds, we used a modification of the parameterization of Haurwitz and Kuhn (1974). Briefly, the Haurwitz and Kuhn (1974) flux transmissivity was converted to intensity transmittance by assuming the diffusity approximation and an exponential dependence of transmittance on optical depth.

Radiances and fluxes were computed for completely clear and overcast conditions for each sounding. Following Ellingson et al. (1989), a cloud layer was added randomly in the vertical to each sounding, but the H2O and temperature profiles were not altered when the cloud layer was included. Initially, the overcast conditions were nearly uniformly distributed in low (1000 − 850 hPa), middle (850 − 450 hPa), and high (450 − 240 hPa) layers. To more adequately approximate the range of optical properties of clouds that reach to high levels, the soundings chosen to have high cloud layers were used to perform two sets of calculations, one with the clouds approximated as thin cirrus and a second with the clouds approximated as blackbodies (e.g., cumulonimbi). This results in a total of 2121 overcast cases. The regression analysis is performed using all 3717 (clear and overcast conditions) calculations to determine the constants of Eq. (1) for each of five view angles (0°, 21°, 48°, 53°, 71°), a subset of the Gaussian angles (21°, 48°, and 71°), and two others routinely calculated. Coefficients at other angles are determined through linear interpolation.

The rms OLR error of the regression decreases with the number of GOES sounder channels (predictors) selected in a manner similar to that described by Ellingson et al. (1989) for HIRS2. Typically, the first predictor (channel 6) chosen explains about 97% of the variance, and the incrementally explained variance for the GOES Sounder becomes small after the addition of the fifth channel (channel 9). Although additional predictors were judged to be significant by the F test when noise-free data were used in the regression, they have been neglected because we estimate that the uncertainty caused by the effects of realistic noise in the those channels exceeds the variance they explain in the noise free regression.

The characteristics of the best five GOES channels selected by the regression are summarized in Table 1. As shown in Table 1, channels 10 and 12 are sensitive to lower and upper tropospheric water vapor and temperature, respectively; channel 2 is sensitive to air temperature from a 10-km layer centered approximately near 250 hPa; channel 6 is sensitive to both the lower tropospheric and the surface temperature; and channel 9 is sensitive to the columnar O3 amount.

The regression analysis showed that OLR could be estimated from radiances measured in these five narrow spectral intervals with rms errors between 4.8 and 6.1 W m−2, depending upon the viewing angle. The standard errors are smallest near 50° (4.8 W m−2) and highest at 71° (6.1 W m−2). These rms errors are about three times larger than those of Ellingson et al. (1989) using an equation that includes a constant term but no cold black clouds. However, this study has found that including a constant term in the regression results in high biases for OLR values below 150 W m−2 when the equation is applied to observed radiances. Forcing the constant term to zero in the regression analysis eliminates this bias at low OLR values. Note that to eliminate the bias for OLR values below 150 W m−2, Ellingson et al. (1994) used an empirical adjustment of the cirrus emissivity. That adjustment is not done in this study.

We have not studied the angular dependence of the rms error in detail. However, as shown in many seminal papers (e.g., Rodgers and Walshaw 1966), the OLR from a plane parallel atmosphere is approximately π × radiance at about 53° (i.e., the diffusivity approximation). Although the upwelling radiance at other angles is highly correlated with that near 50°, there is not as much OLR information there, for the channels selected, as there is near 50°.

3. Data sources and selection procedures

a. Data sources

Data collected from the GOES-8, -9, and -10 satellites, and from the TRMM and Terra satellites for July 1998 and April 2000, are used in this study. The study region is limited to the area extending from 20° to 53°N and from 60° to 130°W. Three CERES instruments were flown onboard the two National Aeronautics and Space Administration (NASA) satellites. The CERES proto-flight model (PFM) instrument is onboard the TRMM launched in November 1997, and twin instruments, CERES flight models 1 (FM1) and 2 (FM2) are onboard the Terra spacecraft launched in December 1999. The TRMM satellite is a low orbiting spacecraft with an inclination of 35° to the equator that provides good time sampling between 35°N and 35°S. In contrast, Terra is in a 1030 LT, sun-synchronous polar, 705-km orbit that covers all latitudes. From the TRMM spacecraft, the CERES footprint is approximately 10 km in diameter at the nadir. From Terra, the instrument footprint is approximately 20 km at the nadir.

NOAA/National Environmental Satellite, Data, and Information Service (NESDIS) produces hourly data from the GOES Sounder that are used for sounding and cloud-top applications (Schreiner et al. 2001; Schmit et al. 2002) over the continental United States (CONUS) and surrounding oceans. These hourly data correspond to observations made during the 30-min satellite scans across the CONUS and surrounding oceans. GOES-8 starts scanning across the CONUS 14 min before each hour and GOES-9 (10) 1 min after the hour.

The GOES Sounder data consist of radiances in 18 infrared channels and reflected radiance in a visible channel and are archived as Man–Computer Interactive Data Access System (McIDAS) area files (see SSEC 1997). The spatial resolution of these data is approximately 15 km over the study region. For this study, these data are remapped to a common earth-referenced grid. The grid is a simple latitude–longitude mesh over the CONUS and surrounding oceans, extending from 60° to 130°W and from 20° to 53°N. The resolution is 0.10° latitude × 0.112° longitude, yielding 206 250 grid boxes. The GOES-8 and GOES-9 (10) data are merged into a single image with 105°W longitude separating the two sectors. GOES-9 stopped operating on 17 July 1998 and was replaced thereafter by GOES-10.

Before the GOES-8 Sounder radiances are used in Eq. (1), they are corrected for biases using data obtained from NESDIS (T. Schmit 2001, personal communication). The biases are the monthly mean difference between GOES-observed radiances and model-calculated radiances using radiosonde data. The GOES-8 bias-corrected radiances reduce the mean difference between GOES and CERES by about 2 W m−2. No bias values were available for GOES-9 and -10 and their radiances are used with no bias correction.

The CERES data used herein are the ERBE-like product named the ES-8 archival data product. This dataset contains 24-h, instantaneous field-of-view (FOV) scanner fluxes at the top of atmosphere (TOA). The instantaneous shortwave (SW) and longwave (LW) fluxes are obtained using ERBE angular distribution models (ADMs; Wielicki et al. 1996). For each FOV, the ES-8 data file contains the observation time, latitude and longitude, filtered radiances (total LW, shortwave, and window channel measurements), satellite and sun geometry (satellite zenith, solar zenith, and relative azimuth angles), unfiltered radiances calculated from filtered radiances (shortwave and longwave only), TOA shortwave and longwave fluxes, and ERBE scene identification. (The details of ES-8 data product may be found online at http://eosweb.larc.nasa.gov/project/ceres/DPC/index.html.)

The CERES orbital data are remapped to the same earth-referenced GOES grid using latitude and longitude information. Note that the CERES–Terra footprint is approximately 20 km at the nadir. Note that four pixels of CERES–Terra correspond approximately to nine GOES pixels over the study region. However, this will not affect our analysis since the comparisons of the fluxes are based on a 1° × 1° grid box as will be described later. For each observation time, images of OLR, SW, scene identification, and satellite zenith and relative azimuth angles are created.

b. Selection procedures

The use of CERES data to validate the GOES technique is complicated by the difference in viewing geometry and the difference of the observation time of the two satellites. GOES views the study region with satellite zenith angles varying between approximately 30° and 55°, while the CERES instrument views the earth's surface from a wide variety of viewing angles. For the July 1998 data used in comparisons, the TRMM viewing angles range from 0° to 65°, and for the April 2000 data, the Terra viewing angles range from 0° to 65° and 0 to 51° for CERES FM1 and FM2, respectively.

For comparison purposes, we consider 1° × 1° boxes. First, GOES observations within 30 min of a CERES view are spatially interpolated onto the CERES orbit (Fig. 1). It is assumed that the atmospheric state remains constant over the 30-min interval. Since the two instruments do not view exactly identical scenes, we attempted to design the analysis to select scenes for which the flux field is relatively horizontally homogeneous. We followed the technique of Ellingson et al. (1994) to test for homogeneity. If the standard deviation of the box is less than 2% of the mean flux from the box for both satellite observations, the field was judged to be homogeneous. Boxes containing less than 25% CERES coverage are not considered in the comparison.

Note that the 2% homogeneity test is equivalent to assuming about 66% of the clear scenes in a 1° × 1° box have surface temperatures within about ±4.5°C, assuming midlatitude summer conditions. For overcast conditions, this corresponds to about ±0.5, ±0.4, and ±0.3 km variations in cloud-top heights for low- (3 km), middle- (7 km), and high- (10 km) level (black) clouds, respectively. In terms of cloud fraction, the 2% homogeneity test corresponds to about ±24, ±7, and ±4% variations about 50% cloud cover for black clouds at 3, 7, and 10 km, respectively. Thus, the criteria will generally exclude 1° × 1° cases where there are large scene-to-scene variations in cloud amount and/or cloud top heights. We suspect that areas with multi-layer clouds will be particularly under sampled. Although these may well be the “easiest” scenes, they are the ones for which the ERBE angular models, used to derive the CERES fluxes, should have the highest accuracy.

The selection procedures resulted in approximately 28 000, 10 000, and 17 000 instantaneous, homogenous scenes for CERES–TRMM (July 1998), and CERES–Terra FM1, and FM2 (April 2000), respectively. Note that this includes a significant number of observations for most of the 12 ERBE scene types. The continental sector located west of 105°W is mostly desert area while the continental sector located east of 105°W is mainly vegetated land.

4. Results and discussion

a. Analysis of comparisons of collocated instantaneous, homogeneous scenes

Figures 2–4 present scatterplots of the comparisons between CERES and GOES instantaneous OLR for July 1998 (Fig. 2) and April 2000 (Figs. 3 and 4) for all scenes and for both night and day conditions. In general, the spread between the CERES and GOES estimates tends to be largest in the middle range, a result likely due to time varying broken cloudiness and/or cloud-top heights. There is very little bias in the mean between the CERES and GOES fluxes. A linear regression with no constant term is used in the analysis. Such an analysis preserves the bias between the variables and does not suffer from the problems with a small number of data points. The slope (β) of the regression line is very close to 1, which is consistent with a very small bias between the GOES and TRMM and Terra CERES OLR. The GOES technique explains 93% and 97% of CERES variance in July 1993 and April 2000, respectively.

Tables 2–4 present summary statistics on the agreement between GOES and CERES for land and ocean scenes and for night and day conditions. These statistics show a maximum positive GOES − CERES average difference of about 2 W m−2 for night ocean and day land scenes for July 1998. For April 2000, the maximum GOES − CERES bias of about 1 W m−2 is observed for day ocean scenes. Table 5 and 6 show rms and average CERES − GOES differences sorted by GOES satellites and sky conditions. The clear and cloudy scenes are determined using the CERES scene identification, and land and ocean scenes are determined using land mask data. The CERES cloud mask uses high-resolution imager pixel radiances within CERES footprints to derive an estimate of cloud cover (f) over the footprint. The following relationship between f and ERBE-like scene type is assumed: clear–0 ≤ f < 5%, partly cloudy–5 ≤ f < 50%, mostly cloudy–50 ≤ f < 95%, and overcast–f > 95%.

Overall, there are relatively small average and rms differences between the GOES and CERES OLR estimates (about 0.5 and 7 W m−2, respectively). However, the overall statistics are clouded somewhat by the differences between GOES-8 and -9 (Table 5) and GOES-8 and -10 (Table 6). In general, the CERES − GOES bias changes sign from GOES-8 to -9 in July 1998 and this change reduces the overall bias. The CERES − GOES bias does not change sign from GOES-8 to -10. Since we do not have the empirical calibrations for GOES-9 and -10, it is difficult to make conclusions regarding the mean differences and the scene-to-scene differences for GOES-9 and -10. Since we do have an empirical calibration of GOES-8, most of our comments will be directed at those results.

As regards GOES-8, the CERES − GOES difference tends to increase from night to day on average by almost 2 W m−2 for July 1998. The only scenes not showing this increase are the mostly cloud ones. However, the number of mostly cloudy and overcast scenes is small. Similar results also are observed between GOES-8 and CERES FM2 in April 2000. We have not yet determined the cause for these differences, but they could be related to solar contamination in CERES instruments, diurnal variations in the CERES ADMs, improper modeling of the temperature structure near the surface, and/or improper specification of the surface emissivity.

It must be emphasized that there are relatively few mostly cloudy and overcast cases in the data studied to date, and the mean and rms differences for these categories are the largest. Thus, the overall statistics are heavily weighted by the generally good results for clear and partly cloudy conditions. Clearly, additional overcast and mostly cloudy cases must be studied in order to be more confident of the range of uncertainty of the GOES Sounder OLR estimates for these conditions.

In an early version of CERES-TRMM products, there is evidence of solar contamination in CERES OLR fluxes. The CERES team estimated that there was an inconsistency of about 1% between CERES OLR version 1 and OLR estimated from the CERES window channel during daylight hours. To evaluate this contamination, shortwave fluxes are directly plotted against longwave fluxes using a sample of version-1 and -2 data from clear ocean and low bright cloud scenes off the coast of Baja California (21°–30°N). This location is used because of the generally uniform cover by highly reflective stratocumulus clouds occurring there.

Figure 5a shows a tendency of increasing OLR with reflected solar flux, which accounts for 11% of version 1 OLR variance. If there were no shortwave contamination of CERES OLR, one should expect no correlation with the shortwave. Indeed, this is shown in Fig. 5b where the plot is repeated with CERES version-2 data. Our analysis shows no correlation between OLR and SW fluxes obtained from both CERES instruments on board Terra for April 2000.

Ellingson et al. (1994) have shown that ERBE OLR were high biased relative to HIRS2 OLR for clear desert scenes at large angles (40°–60°) during the night and low biased at the same viewing angles during daylight hours. However, this is not easy to verify with GOES/CERES data because the two satellites observe a particular scene with different viewing angles. When the comparison is limited to cases where the view angle difference between CERES and GOES is within 10°, the biases remain nearly the same for both nighttime and daytime data.

A plot of CERES − GOES values corresponding to clear desert scenes (not shown here) also shows no evidence of dependence on the viewing angles. This suggests that the ADMs are not major contributors to the differences. However, both techniques may suffer from errors due to 3D cloud structure since longwave anisotropy varies with 3D cloud structure, a parameter that was not taken into account in establishing CERES ADM or in GOES OLR technique.

b. Comparisons for monthly averages

Important considerations for use of the GOES OLR data for climate analyses are their biases relative to CERES for all cloud conditions, the land-ocean differences of these biases, and the month-to-month consistency of the data. We do not yet have a long history to compare with CERES, so it is difficult to test the month-to-month consistency of the data. However, shown in Figs. 6–8 are scatterplots of GOES verses CERES monthly averaged OLR for 50 × 50 km instantaneous collocated scenes, independent of flux homogeneity. We limited the monthly comparisons to GOES-8 data since no GOES-9 data were available after 17 July 1998. The monthly means are obtained by averaging all collocated observations of CERES and GOES regardless of the time of day.

The GOES fluxes are, on average, about the same as both TRMM and Terra CERES fluxes for both July 1998 and April 2000. The CERES − GOES mean differences are less than 2 W m−2 and the rms differences are about 3 and 2 W m−2 for July 1998 and April 2000, respectively for both land and ocean scenes. The GOES technique explains about 98% of CERES OLR variance for land scenes for both July 1998 and April 2000. It explains 92% and about 98% of CERES variance for the ocean scenes for July 1998 and April 2000, respectively.

5. Summary and conclusions

This paper has outlined a technique for estimating OLR from the GOES Sounder following the procedures employed in a similar technique using the HIRS2. The GOES technique differs from the HIRS2 technique in that it uses an extra channel in the 9.6 μm O3 region, it has a better representation of optically thick high clouds, and it removes an unsubstantiated empirical adjustment for cirrus clouds. Overall, the regression analysis of model calculated data indicates that it should be possible to obtain instantaneous OLR with an rms error of about 5 W m−2 using radiances from only five GOES Sounder spectral intervals.

Comparisons of GOES with collocated TRMM and Terra CERES OLR show instantaneous rms agreement to within about 7 W m−2 for day and/or night homogeneous scenes. This is slightly larger than the 5 W m−2 rms agreement shown for similar instantaneous comparisons between NOAA-9 HIRS2 and ERBE by Ellingson et al. (1994). It should be noted that the HIRS–ERBE comparisons were helped by both instruments being on the same satellite and simultaneously viewing nearly identical scenes. The surface footprint was much larger for ERBE than for CERES, and this along with the differences in sampling time may account for the differences in the rms errors. In both cases, however, the rms differences are close to the value expected from the model calculations. It should be noted, however, that about 90% of the cases come from clear and partly cloudy regions. Additional research must be done for mostly cloudy and overcast cases.

Our analysis does indicate about a 2 W m−2 day to night difference in the GOES − CERES bias that were not evident in the HIRS2 comparisons. However, it should be remembered that the HIRS comparisons covered regions different from those considered here. Thus it is difficult to ascertain the cause for these differences at this time. We leave a detailed examination of these differences to further study when more data have been gathered and the CERES data have been modified to include the effects of more advanced angular models.

The differences between the comparisons on GOES-8 and -9 (10) highlight the importance on calibration (for information on GOES imager and sounder calibration, see Weinreb et al. 1997) when using these and similar instruments for application to climate problems. Currently, the Sounder uses empirical calibrations based on minimizing the differences between observed and model calculated radiances using radiosonde data as input. Although such a calibration insures consistency in inferred temperature and water vapor soundings, there are potential climate consequences should there be errors in the radiative transfer models and/or the soundings. Furthermore, such calibrations are difficult in data poor oceanic regions. Nonetheless, these problems can be examined through extended analyses of simultaneous CERES and GOES data.

Finally, the small differences between the CERES and GOES for monthly averages in 50 km × 50 km areas encourages us to believe that the GOES Sounders may be an effective tool for gathering accurate data for climatological studies of the longwave radiation budget of extended areas and for increased understanding of the diurnal nature of the planetary radiation budget. The GOES has an advantage over the polar orbiters—high temporal resolution—and thus does not need to incorporate diurnal models for estimating the radiation budget. This is particularly important over land regions when large changes in the near surface temperature result in large diurnal changes in radiation budget. Examples of the effects of different sampling intervals and the diurnal variation of OLR for different surface types will be the subject of a future publication.

Acknowledgments

We acknowledge Dr. Hai-Tien Lee, our colleague in the Department of Meteorology, University of Maryland, for his help in modifying the radiation model code; Messrs. T. Schmit of the Advanced Satellite Products Team (NESDIS), and T. Schreiner of Space Science and Engineering Center, University of Wisconsin, Madison, Wisconsin; and Jaime Daniels of the National Environmental Satellite Data and Information Service, National Oceanic Atmospheric Administration, Camp Springs, Maryland, for their help in providing GOES data. CERES data were obtained from Langley Atmospheric Sciences Data Center. This research was funded in part by NASA under Grant NASA.NAG 56483.

REFERENCES

  • Abel, P. G., , and Gruber A. , 1979: An improved model for the calculation of longwave flux at 11 μm. NOAA Tech. Rep. NESS 106, 24 pp. [NTIS PB80-119431.].

    • Search Google Scholar
    • Export Citation
  • Cess, R. D., , and Vulis I. L. , 1989: Intercomparison and interpretation of satellite derived directional albedos over deserts. J. Climate, 2 , 393407.

    • Search Google Scholar
    • Export Citation
  • Clough, S. A., , Kneizys F. X. , , and Davies R. W. , 1989: Line shape and the water vapor continuum. Atmos. Res., 23 , 229241.

  • Ellingson, R. G., , Yanuk D. J. , , Lee H-T. , , and Gruber A. , 1989: A technique for estimating outgoing longwave radiation from HIRS radiance observations. J. Atmos. Oceanic Technol., 6 , 706711.

    • Search Google Scholar
    • Export Citation
  • Ellingson, R. G., , Lee H-T. , , Yanuk D. , , and Gruber A. , 1994: Validation of a technique for estimating outgoing longwave radiation from HIRS radiance observations. J. Atmos. Oceanic Technol., 11 , 357365.

    • Search Google Scholar
    • Export Citation
  • Gruber, A., , Ellingson R. G. , , Ardanuy P. , , Weiss M. , , Yang S. K. , , and Oh S. , 1994: A comparison of ERBE and AVHRR longwave flux estimates. Bull. Amer. Meteor. Soc., 75 , 21152130.

    • Search Google Scholar
    • Export Citation
  • Haurwitz, F., , and Kuhn W. R. , 1974: The distribution of tropospheric planetary radiation in the Southern Hemisphere. J. Appl. Meteor., 13 , 417429.

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

    • Search Google Scholar
    • Export Citation
  • Ma, X. L., , Schmit T. J. , , and Smith W. L. , 1999: A nonlinear physical retrieval algorithm—Its application to the GOES-8/9 Sounder. J. Appl. Meteor., 38 , 501513.

    • Search Google Scholar
    • Export Citation
  • Malkmus, R., 1997: Random Lorentz band model with exponential-tailed S−1 line intensity distribution function. J. Opt. Soc. Amer., 57 , 323329.

    • Search Google Scholar
    • Export Citation
  • Menzel, W. P., , Holt F. C. , , Schmit T. J. , , Aune R. M. , , Schreiner A. J. , , Wade G. S. , , and Gray D. G. , 1998: Application of GOES-8/9 soundings to weather forecasting and nowcasting. Bull. Amer. Meteor. Soc., 79 , 20592077.

    • Search Google Scholar
    • Export Citation
  • Morel, P., 1985: Plan for the implementation of the World Climate Research Programme. Adv. Space Res., 5 (6) 1520.

  • NAS/NRC, 1983: Toward an international geosphere–biosphere program. National Research Council Workshop Rep., National Academy Press, Washington, D.C.

    • Search Google Scholar
    • Export Citation
  • Phillips, N., , Susskind J. , , and McMillin L. , 1988: Results of a joint NOAA/NASA sounder simulation study. J. Atmos. Oceanic Technol., 5 , 4456.

    • Search Google Scholar
    • Export Citation
  • Rodgers, C. D., , and Walshaw C. D. , 1966: The computation of infrared cooling rates in planetary atmospheres. Quart. J. Roy. Meteor. Soc., 92 , 6792.

    • Search Google Scholar
    • Export Citation
  • Rothman, L. S., and Coauthors. 1992: The HITRAN molecular database: Editions of 1991 and 1992. J. Quant. Spectrosc. Radiat. Trans., 48 , 496508.

    • Search Google Scholar
    • Export Citation
  • Schmit, T. J., , Feltz W. F. , , Menzel W. P. , , Jung J. , , Noel A. P. , , Heil J. N. , , Nelson J. P. III, , and Wade G. S. , 2002: Validation and use of GOES sounder moisture information. Wea. Forecasting, 17 , 139154.

    • Search Google Scholar
    • Export Citation
  • Schreiner, A. J., , Schmit T. J. , , and Menzel W. P. , 2001: Observations and trends of clouds based on GOES sounder data. J. Geophys. Res., 106 , 2034920363.

    • Search Google Scholar
    • Export Citation
  • SSEC, cited 1997: McIDAS user's guide. Space Science and Engineering Center, University of Wisconsin—Madison. [Available online at www.ssec.wisc.edu/mug.].

    • Search Google Scholar
    • Export Citation
  • Warner, J. X., , and Ellingson R. G. , 2000: A new narrowband radiation model for water vapor absorption. J. Atmos. Sci., 57 , 14811496.

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

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

    • Search Google Scholar
    • Export Citation
  • WMO, 1988: Concept of the Global Energy and Water Cycle Experiment: Report of the JSC Study Group on GEWEX. WMO Tech. Doc. 215, WCRP-5, 70 pp.

    • Search Google Scholar
    • Export Citation
Fig. 1.
Fig. 1.

(a) CERES TRMM orbital passes remapped onto an earth-referenced 10 km × 10 km fixed grid 1600 UTC on 2 Jul 1998. (b) Spatially interpolated GOES observations on the CERES TRMM orbit

Citation: Journal of Atmospheric and Oceanic Technology 20, 1; 10.1175/1520-0426(2003)020<0079:VOATFE>2.0.CO;2

Fig. 2.
Fig. 2.

Scatterplot of GOES and TRMM CERES OLR data from 100 km × 100 km homogeneous scenes for Jul 1998 for all scenes.

Citation: Journal of Atmospheric and Oceanic Technology 20, 1; 10.1175/1520-0426(2003)020<0079:VOATFE>2.0.CO;2

Fig. 3.
Fig. 3.

Scatterplot of GOES and Terra CERES FM1 OLR data from 100 km × 100 km homogeneous scenes for Apr 2000 for all scenes

Citation: Journal of Atmospheric and Oceanic Technology 20, 1; 10.1175/1520-0426(2003)020<0079:VOATFE>2.0.CO;2

Fig. 4.
Fig. 4.

Scatterplot of GOES and Terra CERES FM2 OLR data from 100 km × 100 km homogeneous scenes for Apr 2000 for all scenes

Citation: Journal of Atmospheric and Oceanic Technology 20, 1; 10.1175/1520-0426(2003)020<0079:VOATFE>2.0.CO;2

Fig. 5.
Fig. 5.

Scatterplots of CERES OLR and SW fluxes from 100 km × 100 km homogeneous scenes for Jul 1998 over the Pacific Ocean off the coast of Baja California: (a) CERES version-1 fluxes, and (b) CERES version-2 fluxes

Citation: Journal of Atmospheric and Oceanic Technology 20, 1; 10.1175/1520-0426(2003)020<0079:VOATFE>2.0.CO;2

Fig. 6.
Fig. 6.

Scatterplot of GOES-8 vs TRMM CERES monthly averaged OLR for 50 km × 50 km instantaneous collocated scenes

Citation: Journal of Atmospheric and Oceanic Technology 20, 1; 10.1175/1520-0426(2003)020<0079:VOATFE>2.0.CO;2

Fig. 7.
Fig. 7.

Scatterplot of GOES-8 and -10 vs Terra CERES FM1 monthly averaged OLR for 50 km × 50 km instantaneous collocated scenes

Citation: Journal of Atmospheric and Oceanic Technology 20, 1; 10.1175/1520-0426(2003)020<0079:VOATFE>2.0.CO;2

Fig. 8.
Fig. 8.

Scatterplot of GOES-8 and -10 vs Terra CERES FM2 monthly averaged OLR for 50 km × 50 km instantaneous collocated scenes

Citation: Journal of Atmospheric and Oceanic Technology 20, 1; 10.1175/1520-0426(2003)020<0079:VOATFE>2.0.CO;2

Table 1.

GOES Sounder spectral intervals selected by the OLR regression analysis and their absorber sensitivities. The term IEV denotes the incrementally explained variance as a variable is added

Table 1.
Table 2.

Statistics on the comparisons between CERES TRMM and GOES OLR from 104 km2 homogeneous scenes for Jul 1998

Table 2.
Table 3.

Statistics on the comparisons between CERES Terra FM1 and GOES OLR from 104 km2 homogeneous scenes for Apr 2000

Table 3.
Table 4.

Statistics on the comparisons between CERES Terra FM2 and GOES OLR from 104 km2 homogeneous scenes for Apr 2000

Table 4.
Table 5.

Statistics on the comparison between CERES TRMM and GOES OLR from 104 km2 homogeneous scenes for Jul 1998. The comparisons are made for GOES-8 and -9 separately

Table 5.
Table 6.

Statistics on the comparisons between CERES Terra FM2 and GOES OLR from 104 km2 homogeneous scenes for Apr 2000. The comparisons are made for GOES-8 and -10 separately

Table 6.
Save