Presentation on theme: "On the Development of GOES-R Longwave Earth Radiation Budget Products Hai-Tien Lee (1) & Istvan Laszlo (2) (1) CICS/ESSIC-NOAA, University of Maryland."— Presentation transcript:
On the Development of GOES-R Longwave Earth Radiation Budget Products Hai-Tien Lee (1) & Istvan Laszlo (2) (1) CICS/ESSIC-NOAA, University of Maryland (2) NOAA/NESDIS/STAR GOES-R 2007 Annual MeetingLansdowne, VA, May 15-18, 2007 Acknowledgments: Nicolas Clerbaux, Steven Dewitte (RMI, Belgium), Jacqueline Russell (Imperial College, UK), GERB and SEVIRI Teams, Fred Rose & CERES Team, and NASA Langley Data Center. GOES-R Risk Reduction and AWG Projects.
4 HIRS OLR Climate Data Record (Lee et al., 2007) HIRS OLR Climate Data Record in excellent agreement with broadband measurements. HIRS OLR Validation (Ellingson et al., 1994) HIRS OLR is Operational since Tropical Mean HIRS Multi-spectral OLR Algorithm N i = channel i radiance = local zenith angle a i = regression coefficient
5 SEVIRI OLR Validation
6 Methods and Issues for Measurements Collocation
8 SEVIRI-CERES OLR Differences June 2004 Negative DifferencesPositive Differences Day Night Highly variable cloud, e.g., ITCZ, is a scene that the measurements were typically “mismatched” - producing both large positive and negative random errors.
9 Space and Time Collocation Observation Time in 15 min. window View zenith angle matched (azimuth is not!) Spatial Homogeneity SEVIRI radiances averaged for a 3x3-pixel target (red, 9km at nadir) collocates with a CERES footprint (yellow, 20km at nadir).
10 Collocated, VZA-matched SEVIRI and CERES OLR June 2004 Night-time How do the observation time differences and the varying spatial homogeneity affect the validation results? Mean OLR DiffStd OLR Diff Homogeneity Index Bins Threshold of Time diff (sec)
11 Threshold and Absolute Homogeneity Statistics for points < thresholdStatistics for points for index bins LimitRange
12 Validation Results
13 SEVIRI Model Comparison June and Dec , 2004 Meteosat8 full disk domain CERES SSF FM1/2 (Ed2b), FM3/4 (Ed1b) View zenith angle matched (<1°), homogeneous scenes (index<0.01), both day & night SEVIRI OLR Model B Channels 5, 6, 7, 11 SEVIRI OLR Model A Channels 6, 9, 11
Y= *X Y= *X Effects of Upper Tropospheric Humidity SEVIRI OLR Limb Darkening Biases Day/Night, Homog<0.01, vza matched, No Desert UTH variation is important in determining LW broadband radiance angular variation. Inclusion of 6.2 m channel (Ch5) significantly reduced the limb darkening biases. Model A: Ch 6, 9, 11 Model B: Ch 5, 6, 7, 11
15 GERB - CERES OLR Dewitte et al., 2006 Empirical correction for limb darkening errors:
16 SEVIRI-CERES June and Dec , 2004 Meteosat8 full disk domain CERES SSF FM1/2 (Ed2b), FM3/4 (Ed1b) View zenith angle matched (<1°), homogeneous scenes (index<0.01), both day & night Mean DiffStd Diff Wm Model B 1°x1°
Summary Preliminary SEVIRI/ABI OLR algorithms were validated against CERES SSF data. The full-disk domain mean OLR difference is within 1%, with rms differences of about 5 Wm -2 and nearly one-to-one relationship. Angular dependence and Regional biases –Empirical correction possible, but not desired! –New OLR model accounted for UTH variation, producing much better limb darkening property, but erred for desert (need to devise different regression technique) GERB OLR validation showed similar limb darkening problems - thin cirrus identification problems. Scene identification may improve regional accuracy - particularly for extreme conditions (desert) and semi- transparent cirrus scenes. Adaptation of CERES LW ADM is under evaluation.
19 Validation Summary June and Dec , 2004 Validated with CERES SSF FM1/2 (Ed2b), FM3/4 (Ed1b) For zenith angle matched, homogeneous scenes (index<0.01), day & night: SEVIRI OLR Model B: Channels 5, 6, 7, 11 SEVIRI OLR Model A: Channels 6, 9, 11
20 Spatial Homogeneity Infrared vs. Visible VIS IR