atmospheric corrected surface reflectance (rsrf)...all ch except for ch1 can achieve target accuracy...
TRANSCRIPT
1
Atmospheric corrected surface reflectance (RSRF)
Key revisions:
(1) Stable calculation of BRDF
(2) Improvement of cloud screening and aerosol correction of the slant-view telescopes
(3) Increase of validation samples
(4) Addition of shortwave radiation (SWR)
Hiroshi Murakami
JAXA/EORC
June 2020
2
VN11-c1 VN11-c2
Ver. 1 RSRFQ RGB
Ver. 2 RSRFQ RGB
✓ Bug fix about relative zenith angle (should be abs() in BRDF kernel c1
✓ PI02/01 and VN11P/08P samples are used for the BRDF regression of VN11P/08P
1) BRDF calculation
VN11-c1 VN11-c2
Cloud areas where BRDF is available are recovered by the BRDF (flagged by bit-14 of QA_flag)
BRDF kernel 1 BRDF kernel 2
BRDF kernel 1 BRDF kernel 2
3
Ver. 1 2018/08/01 daily data Ver. 2 2018/08/01 daily data
2) Cloud screening and aerosol correction of the slant-view telescopes
X-axis values are calculated by MODIS BRDF product with SGLI geometry
Ver.1 PI02 2019/08/04 T0911 Ver.2 PI02 2019/08/04 T0911
✓ Adding cloud tests by radiance intensity (I) of Pol data
✓ Aerosol is estimated by PI01 for PI01/02 correction (AE is from non-pol data)
✓ Improve agreement with MODIS BRDF(bellows)
Improvement of clouds screening (some are remaining)
4
Validation results Release accuracy (L+1yr) Standard accuracy (L+5yr) Target accuracy (L+5yr)
0.02 (<=443nm)0.07 (>443nm)
0.3 (<=443nm) →150%0.2 (>443nm) →100%
0.1 (<=443nm) →50%0.05 (>443nm) →25%
0.05 (<=443nm) →25%0.025 (>443nm) →13%
The accuracy targets are defined as error ratios of “solar
zenith angle<30 deg, smooth surface of reflectance ~0.2”
O2A absorption
Polarization telescopes
3) Increase of validation samples
✓ About 550 samples for Ver.2 (about120 samples for Ver.1)
✓ Takayama (PEN), Fuji-hokuroku (PEN), RadCalNet (Namibia, France, US), RRV (PI Moriyama), Mongolia (PI Nasahara), tottori(PI Susaki)
Targ
et
accura
cy
Sta
ndard
accura
cy
Rele
ase
accura
cy
Sta
ndard
accura
cy
Validation result at the Ver.1 release
Sample N~120
0.019(<=443nm)0.084 (>443nm)
Worst value (VN09)
Worst value (PI02)
Ver.1 results
3) Increase of validation samples
5
Validation results Release accuracy (L+1yr) Standard accuracy (L+5yr) Target accuracy (L+5yr)
0.02 (<=443nm)0.06 (>443nm)
0.3 (<=443nm) →150%0.2 (>443nm) →100%
0.1 (<=443nm) →50%0.05 (>443nm) →25%
0.05 (<=443nm) →25%0.025 (>443nm) →13%
Error ratio of reflectance~0.2
Polarization telescopes
✓ Channel number achieving target accuracy is increased by improvement of L1B calibration and QA (cloud screening)
Targ
et
accura
cy
Sta
ndard
accura
cy
Rele
ase
accura
cy
Targ
et
accura
cy
Worst value (PI02)
Ver.2 results
O2A absorption
6
Validation results Release accuracy (L+1yr) Standard accuracy (L+5yr) Target accuracy (L+5yr)
0.02 (<=443nm)0.04 (>443nm)
0.3 (<=443nm) →150%0.2 (>443nm) →100%
0.1 (<=443nm) →50%0.05 (>443nm) →25%
0.05 (<=443nm) →25%0.025 (>443nm) →13%
Error ratio of reflectance~0.2
偏光鏡筒
✓ All CH achieve the standard accuracy and CH number achieving target accuracy is increased with conversion to the in-situ observation angles (nadir) by derived BRDF
✓ All CH except for CH1 can achieve target accuracy by following the validation definition, “solar zenith angle<30 deg, smooth surface of reflectance ~0.2”
0.02127%
0.01920%
0.01816%
0.02014%
0.02012%
0.02115%
0.02410%
0.02410%
0.03913%
0.03210%
0.03210%
0.0257%
0.0348%
0.0268%
0.0208%
0.0309%
Targ
et
accura
cy
Sta
ndard
accura
cy
Targ
et
accura
cy
Sta
ndard
accura
cy
3) Increase of validation samplesVer.2 results with BRDF correction
(4) Addition of shortwave radiation (SWR)
7
Monthly RMS/AVG=14%8-day RMS/AVG=18%
Yearly average
Daily RMS/AVG=27%
✓ Daily SWR is estimated by instantaneous transmittance of visible channels and stored by 8 bit (/Image_data/SWR)
✓ Comparison with in-situ measurements by BSRN
✓ 14% does not achieve the SWR target accuracy (monthly 10W/m2); the daily change should be considered in the future version
✓ Anomaly is useful for applications such as agriculture and long-term weather monitoring
8
Summary
✓ Following items are revised for Ver.2
(1) BRDF calculation in the land atmospheric correction and statistics process to derive BRDF)
(2) Cloud screening and aerosol correction of the slant view observations
(3) Increase of validation samples
(4) Addition of an output variable, daily SWR
✓ All CH achieve the standard accuracy with consideration of the derived BRDF
✓ Toward the Ver. 3
• To brush up quality control including the cloud and cloud shadow contamination
• To consider heterogeneousness around the validation sites