Atmospheric Correction around AERONET sites: Development of Global Land Validation Dataset

PI: A. Lyapustin; co-I: Y. Wang (GEST UMBC/NASA GSFC)Collaborators: B. Holben, J. Privette (GSFC); J. Martonchik, D. Diner (JPL)

GSFC, October 5, 2000 (ETM+)

Challenges of SR Validation

• Heterogeneous surface.

• Spatial resolution (1 km) vs scale of surface variation(20 m) (thousands of measurements).

• Spectral differences between ground-based and space-borne instruments.

• Ground-based measurements require atmosphericcorrection (complex experiments).

Research Objective: Development of Advanced Atmospheric Correction Algorithm for MODISDevelopment of Advanced Atmospheric Correction Algorithm for MODIS

Linked to:Development of Validation Dataset of Surface Reflectance over LandDevelopment of Validation Dataset of Surface Reflectance over Land

Daily Data CollectionM O D IS, M ISR ...

(area ~ 32x32 km 2)

Ancillary DataAER O N ET aeroso l and

W V, N C EP ozone

Automatic AC(s ing le va lida ted R T,

un ified algorithm )

M ain Functions

ASRVN - AERONET-based Surface Reflectance Validation Network

THEORETICAL BASIS

• 3D Radiative Transfer (Lyapustin & Knyazikhin, Appl. Opt., 2001; 2002)• variable anisotropic surface;• arbitrary spatial resolution;• semi-analytical, accurate and fast due to parameterizations.

• Accurate Modeling of Gaseous Absorption • Inversion with MRPVMISR and LSRTMODIS BRF Models

PRODUCTS• BRF, Albedo (spectral & SW broadband)• Surface Radiative Fluxes, PAR

EXPECTED BENEFITS

• Validation of BRF & Albedo over Heterogeneous Surfaces• MODIS-MISR Data Fusion• Calibration Analysis

• Vicarious calibration• Cross-calibration of different sensors• Detection of calibration trend based on a time

series of surface reflectance.

MISR MODIS ETM+ SeaWIFS VIIRS

Study Area:

1. Mongu (Zambia)(15o15’S, 23o09’E), H=1,104 m

2. GSFC (USA)(39o03’N, 76o88’W), H=50 m

Albedo Time Series

(Mongu, 2003) Vis. (RGB) and NIR (grey)

fromMISR Collection 4 DataMISR Collection 4 Data

March 19 May 21 June 22 July 24 August 25 September 26 October 28

March 19 June 22 October 28

Analysis of BRF

Solid lines – MISR radiance/BRFSquares – ASRVN BRF (MRPV algorithm)Triangles – model BRF calculated with the best-fit parameters (MRPV)

1. MISR retrieves correct BRF shape.2. MISR BRF is less anisotropic than the ASRVN BRF in the Vis.

(analysis: in IEEE TGARS Special Issue on Land Product Validation)

Mongu, March 18, 2003Mongu, March 18, 2003Mongu, July 8, 2003Mongu, July 8, 2003

0.1

0.15

0.2

0.25

0.3Urban Pixel

TO

A R

adia

nce

0

0.1

0.2

0.3Green Pixel

0.08

0.12

0.16

0.2

0.24

0.28Urban Pixel

0.06

0.12

0.18

0.24

0.3

0.36Green Pixel

0

0.04

0.08

0.12

0.16

BR

F V

is

0

0.03

0.06

0.09

0

0.05

0.1

0.15

0.2

0.25

0

0.06

0.12

0.18

0.27

0.3

0.33

0.36

-75 -50 -25 0 25 50 75

BR

F N

IR

0.12

0.19

0.26

0.33

-75 -50 -25 0 25 50 750.23

0.29

0.35

0.41

-75 -50 -25 0 25 50 750.32

0.38

0.44

0.5

-75 -50 -25 0 25 50 75

View Zenith Angle

Analysis of Albedo

0

0.2

0.4

0.6

0.8

1

1.2

20

25

30

35

40

45

50Mongu

AO

T SZ

A

SZA

0

0.08

0.16

0.24

0.32

0.4

20

30

40

50

60

70GSFC

AO

T SZ

A

SZA

0.05

0.1

0.15

0.2

0.25

0.3

0.35Urban Pixel

Alb

edo

0.05

0.1

0.15

0.2

0.25

0.3Urban Pixel

Alb

edo

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

Mar-03 Jul Nov Mar-04 Jul Nov

Green Pixel

Alb

edo

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

Jan-03 Apr Jul Oct Jan-04 Apr Jul

Green Pixel

Alb

edo

1. MISR correctly reproduces the time series of albedo.

2. MISR albedo is on average lower by about 0.005 in the green and red bands.

MISR AOT

AERONET

MISR

ASRVN

Analysis of Albedo Spatial Distribution

Mongu, July 24, 2003Mongu, July 24, 2003Blue Green Red NIR

GSFC, August 24, 2003GSFC, August 24, 2003Blue Green Red NIR

MR

PV

-MIS

RL

SR

T-M

RP

V

1. MISR correctly reproduces spatial distribution of albedo.

2. An albedo mosaic may develop due to MISR aerosol retrievals at 17.6 km resolution.

3. ASRVN albedo does not have model-dependent distortions from the variable land cover.

MISRMISR

1. MISR BRF and albedo products are generally accurate in both clear and hazy atmospheric conditions.

2. MISR correctly reproduces the time series and spatial distribution of albedo.

3. MISR BRF on average is less anisotropic than actual BRF in the visible bands.

The difference is greatest in the blue band, decreases with wavelength, and it is negligible in the near-IR band. This discrepancy most likely originates in 1) MISR aerosol retrieval algorithm over heterogeneous land, which tends to select an aerosol model that benefits the spectrally invariant shapes of surface BRF; 2) MISR surface HDRF retrieval algorithm where the iteration loop that removes the diffuse atmospheric transmittance is currently turned off.

4. Our initial results suggest that the MISR surface albedo is on average lower than the ASRVN albedo by about 0.005 in the green and red bands.

BRF Model (LSRT vs MRPV)BRF Model (LSRT vs MRPV)

1. MRPV model fits BRF shapes slightly better than LSRT model, except in the blue band.

2. The AC algorithm based on LSRT model is much faster and more robust (due to linearity).

3. The albedos from the two models are generally similar, with the average difference q0.005. The difference q is spatially homogeneous but site-dependent (function of aerosol absorption).

Summary of Results (local analysis)

What is Next …

MISRMISR

• ASRVN is receiving and processing operational MISR collection 5 data (0.5 Yr in 2000, and since Dec. 2004) for all AERONET stations (166). We plan to carry on MISR validation analysis on the regional and then global scales.

• The web-interface with data analysis tools is under development.

ETM+ETM+

• The ASRVN_ETM+ processing algorithm is being finalized (1 month).

• ASRVN_ ETM+ products will become available this summer.

MODISMODIS

• Cooperating with Goddard Earth Sciences Data & Information Services Center (G. Leptouch, D. Ouzounov) to produce MODIS Collection 5 subsetted and aggregated data for ASRVN.

• Operational ASRVN_ MODIS products will become available this fall.