The document presents a methodology for prototyping an albedo algorithm for GOES-R using MODIS data. It uses an optimization approach that incorporates atmospheric radiative transfer modeling and land surface BRDF modeling to estimate surface albedo, spectral reflectance, and aerosol optical depth from MODIS TOA reflectance observations. The estimates were validated against ground measurements and other satellite products, showing good agreement within F&PS requirements for albedo accuracy and reflectance precision. Future work will include additional validations and improving diurnal albedo estimation using geostationary data.

1. Prototyping GOES-R
Albedo Algorithm Based
on MODIS Data
Tao Hea, Shunlin Lianga, Dongdong Wanga
a. Department of Geography, University of Maryland, College Park, USA
Hongyi Wub
b. University of Electronic Science and Technology of China, China
Yunyue Yuc
c. NOAA/NESDIS/STAR, USA
Presented by Tao He
the@umd.edu
Jul 29, 2011

2. Contents
1 Introduction
2 Methodology
3 Data and Results
4 Summary and Conclusions

3. Introduction
§ Surface albedo is defined as the ratio of
outgoing and incoming radiation at Earth
surface.
§ Essential in energy budget
§ Climate change studies
§ Hydrology cycle
§ Weather forcast
§ …
§ Current satellite albedo products
§ MODIS, MISR, MERIS, MSG/SEVIRI, …

4. GOES-R ABI
§ The Advanced Baseline Imager (ABI) is the
primary instrument onboard GOES-R for
weather, climate and environmental studies.
§ Temporal resolution: 15min
§ Spatial resolution: 0.5 – 2km
§ Spectral bands: 6 bands in solar range (0.4 – 2.3
µm)
§ Abundant spectral and angular information
can be available within a small time period
to derive surface spectral BRDF and
broadband albedo.

5. ABI vs MODIS
GOES-R
GOES-R ABI MODIS
Channel Central Wavelength Spatial Channel Central Wavelength Spatial
Number (µm) Resolution Number (µm) Resolution
1 0.47 1 km 3 0.47 0.5 km
2 0.64 0.5 km 1 0.65 0.25 km
3 0.86 1 km 2 0.86 0.25 km
4 1.38 2 km N/A
5 1.61 1 km 6 1.64 0.5 km
6 2.26 2 km 7 2.13 0.5 km
N/A 4 0.56 0.5 km
N/A 5 1..24 0.5 km

6. Existing Methods
1. Atmospheric correction
2. Surface BRDF modeling
A 3. Narrow-2-broadband conversion
(e.g. Schaaf et al. 2002, Geiger et al. 2008)
1. Direct estimation of broadband albedos
(e.g. Liang et al. 2005)
B
1. Atmospheric correction with surface BRDF
modeling
C 2. Narrow-2-broadband conversion
(e.g. Govaerts et al. 2010)

7. Objectives
§ Using MODIS TOA data as proxy to prototype
the future GOES-R albedo algorithm based on
atmospheric correction with BRDF modeling;
§ Estimating instantaneous albedo/reflectance as
well as instantaneous aerosol optical depth;
§ Improving the albedo estimation over rapidly-
changing surfaces;
§ Validating/verifying albedo/reflectance estimates
with multiple datasets.

8. Methodology
Cost Function:
§ x: coefficients of the surface BRDF model and AOD,
§ r(x): calculated surface albedo using the BRDF model
§ rb: background values of albedo from albedo climatology
§ B: uncertainty matrix of the albedo background values
§ ρ: satellite observed TOA reflectance
§ ρ(x): calculated TOA reflectance from the radiative transfer
equation
§ R: error matrix of the calculated TOA reflectance
§ Jc: cost function to account for various constraints (physical
meanings of BRDF parameters, and AODs, etc.).

9. Atmospheric Radiative Transfer Solution
with Land Surface BRDF Modeling
§ Atmospheric Radiative Transfer Formulation for better modeling the
interaction between atmosphere and non-Lambertian surfaces
TOA Path Transmittance Surface Reflectance Matrix
Reflectance Reflectance Matrix
(Qin, et al. 2001)
Spherical Albedo
Transmittance Matrix
Surface Reflectance
Matrix
i, v refer to the incoming and outgoing light directions respectively; all atmospheric
variables in the above model were simulated for each major aerosol type using 6S
software and stored in LUT for computational purpose.

10. Atmospheric Radiative Transfer Solution
with Land Surface BRDF Modeling (cont.)
§ Surface BRDF Modeling
§ Kernel models used with consideration of hot spot
effects (Maignan, et al. 2004)
Where

11. Flowchart
TOA
Reflectances
Radiative
Prior AOD Prior BRDF
Transfer Model
Optimal BRDF
Albedo Spectral
Optimization Parameters
Climatology Reflectance
and AOD
Narrow-2-
Spectral Angular
Broadband
Albedo Integration
Conversion
Broadband
Albedo

12. Data
§ Satellite data & products
§ MODIS L1B data (TOA radiance, geometry)
§ MODIS cloud mask
§ Ancillary data
§ Albedo climatology maps from multi-year MODIS
albedo products (2000 – 2009)
§ NCEP water vapor

13. Albedo Climatology and Uncertainty
§ Ten-year average shortwave albedo (a) and its one-year standard
deviation (b) for Julian Day 121–128 from MODIS albedo product
2000–2009 over North America and Greenland.

14. Validation Datasets
§ Ground measurements
§ AmeriFlux
§ Surface Radiation (SURFRAD) Network
§ Greenland Climate Network (GC-Net)
§ Satellite data calibrated with in-situ aerosol data
§ MODASRVN (Wang et al. 2009)
§ Finer resolution satellite data
§ Landsat data from LEDAPS (Vermote et al. 2007)

15. Validation Results: Vegetated Surface
Bondville Lat:40.05 Lon:-88.37
1
Ground measured albedo
Shortwave Albedo
0.8 Estimates
MODIS 16-day albedo
0.6
0.4
0.2
0
50 100 150 200 250 300 350
Julian Day
Example of time series shortwave albedo from MODIS
observations in 2005 over six SURFRAD sites
Mead(Rain fed) Lat:41.1797 Lon:-96.4396
1
Ground measured albedo
0.8 Estimates
Visible Albedo
MODIS 16-day albedo
0.6
0.4
0.2
0
50 100 150 200 250 300 350
Julian Day
Example of time series total visible albedo from MODIS
15
observations in 2005 over four AmeriFlux sites

16. Validation Results: Snow Surface
Saddle
1
Shortwave Albedo
0.8 • Greenland
0.6
sites (GC-Net)
0.4 Ground measured albedo
0.2
Estimates
MODIS 16-day albedo • 2003
50 100 150 200 250 300 350
Julian Day
• Comparison
with MODIS
albedo
NASA-SE
1
Shortwave Albedo
0.8
products and
0.6
Ground measured albedo
ground
measurements
0.4
Estimates
MODIS 16-day albedo
0.2
50 100 150 200 250 300 350
Julian Day

17. .
Validation Results: Surface Reflectance
BrattsLake (50.28,-104.7) Cropland Egbert (44.226,-79.75) Cropland
1 1
Estimated Red Band IBRF 0.9 Estimated Red Band IBRF
0.9
MODASRVN Red Band IBRF MODASRVN Red Band IBRF
0.8 Estimated NIR Band IBRF 0.8 Estimated NIR Band IBRF
Example of time series
MODASRVN NIR Band IBRF MODASRVN NIR Band IBRF
0.7 0.7
Reflectance
Reflectance
0.6
0.5
0.6
0.5
instantaneous reflectance
0.4 0.4 from MODIS observations
in 2005 over AERONET
0.3 0.3
0.2 0.2
0.1 0.1
sites
0 0
0 50 100 150 200 250 300 0 50 100 150 200 250 300 350
Julian Day Julian Day
All Sites Band1 All Sites Band2
Comparison of estimated
1 1
0.9 y=1.0177x+0.0065 0.9 y=0.91535x+0.024
and MODASRVN
R-squared=0.698 R-squared=0.732
0.8 Bias=0.0084 0.8 Bias=0.0025
RMSE=0.0269 RMSE=0.0471
0.7
Estimated IBRF
0.7
Estimated IBRF
instantaneous 0.6 0.6
bidirectional reflectance
0.5 0.5
0.4 0.4
for MODIS band1&2 over 0.3
0.2
0.3
0.2
16 AERONET sites during 0.1 0.1
2005
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
MODASRVN IBRF MODASRVN IBRF

18. Comparison with Landsat Data
Comparison of aggregated
Landsat shortwave albedo
with retrieved 1km albedo
from MODIS observations
over SURFRAD sites (a)3
by 3 pixels; (b)7 by 7
pixels; (c)11 by 11 pixels;
(d)21 by 21 pixels; (e)31
by 31 pixels.

19. Summary of Validation Results
Albedo
Our Retrievals F&PS
Requirement
Accuracy (Bias) 0.0137 0.08
Precision (RMSE) 0.0618 10%
R2 0.8208 N/A
Reflectance
Our Retrievals F&PS
Requirement
Accuracy (Bias) 0.0084 (Red) 0.08
0.0025 (NIR)
Precision (RMSE) 0.0269 (Red) 5%
0.0471 (NIR)
R2 0.698 (Red) N/A
0.732 (NIR)

20. Conclusions
§ Framework of retrieving surface albedo/
BRDF was established using MODIS TOA
observations as proxy;
§ Extensive validation/verification was made
over various land cover types with ground
measurements from multiple network and
good preliminary results were shown;
§ Good agreement was found in comparison
of surface reflectance estimates with
MODASRVN data.

21. Future Work
§ More validations on albedo and reflectance;
§ Validation of aerosol optical depth
estimation;
§ Sensitivity analysis;
§ Improvement of diurnal albedo estimation
based on geostationary satellite data (e.g.
MSG/SEVIRI).

22. Questions?