radiative transfer codes for atmospheric correction and aerosol retrievals intercomparison study n.
Download
Skip this Video
Loading SlideShow in 5 Seconds..
Radiative Transfer Codes for Atmospheric Correction and Aerosol Retrievals: Intercomparison Study PowerPoint Presentation
Download Presentation
Radiative Transfer Codes for Atmospheric Correction and Aerosol Retrievals: Intercomparison Study

Loading in 2 Seconds...

play fullscreen
1 / 35

Radiative Transfer Codes for Atmospheric Correction and Aerosol Retrievals: Intercomparison Study - PowerPoint PPT Presentation


  • 466 Views
  • Uploaded on

AEROCENTER Fall Seminar Series, October 2 nd , 2007. Radiative Transfer Codes for Atmospheric Correction and Aerosol Retrievals: Intercomparison Study. Svetlana Y. Kotchenova & Eric F. Vermote. The study is being performed in collaboration with: Robert Levy, Alexei Lyapustin, and Omar Torres.

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

Radiative Transfer Codes for Atmospheric Correction and Aerosol Retrievals: Intercomparison Study


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
radiative transfer codes for atmospheric correction and aerosol retrievals intercomparison study

AEROCENTER Fall Seminar Series, October 2nd, 2007

Radiative Transfer Codes for Atmospheric Correction and Aerosol Retrievals:Intercomparison Study

Svetlana Y. Kotchenova & Eric F. Vermote

The study is being performed in collaboration with: Robert Levy, Alexei Lyapustin, and Omar Torres

project description

VPD(vector)

RT3(vector)

6SV1.1(vector)

Svetlana & Eric

SHARM

(scalar)

Monte Carlo

(benchmark)

Robert

Omar

Svetlana & Eric

MODTRAN

(scalar)

Coulson’s tabulated values

(benchmark)

Alexei

Svetlana

Project Description

The project is devoted to the comparison and detailed evaluation of five

atmospheric RT codes incorporated in different satellite data processing algorithms

only molecular atmosphere

only molecular atmosphere

2

applications of the codes
Applications of the codes
  • 6SV1.1 (Second Simulation of a Satellite Signal in the Solar Spectrum, Vector, version 1.1): MODIS atmospheric correction and internal aerosol inversion
  • RT3 (Radiative Transfer 3): MODIS coarse resolution (10-km) aerosol retrieval
  • VPD (Vector Program D): TOMS (Total Ozone Mapping Spectrometer) aerosol inversion
  • SHARM (Spherical Harmonics): MAIAC (Multi-Angle Implementation of Atmospheric Correction for MODIS)
  • MODTRAN (Moderate Resolution Atmospheric Transmittance and Radiance Code): AVIRIS (Airborne Visible/Infrared Imaging Spectrometer) atmospheric correction

3

description of the codes 6sv1 1

Spectrum: 350 to 3750 nm

Molecular atmosphere:

6 code-embedded & 2 user-defined models

Ground surface:

homogeneous &non-homogeneous with & without directional effect (10 BRDF + 1 user-defined models)

Instruments:

  • AATSR, ALI, ASTER, AVHRR, ETM, GLI, GOES, HRV, HYPBLUE, MAS, MERIS, METEO, MSS, TM, MODIS, POLDER, SeaWiFS, VIIRS, & VGT – 19 in total

Aerosol atmosphere:

6 code-embedded & 4 user-defined models & AERONET

Description of the codes: 6SV1.1

Author: E. Vermote (University of Maryland, USA)

Modified: E. Vermote et al.

Language: Fortran 77, 95

Features:

http://6s.ltdri.org

Publications + Interface to create input files

4

description of the codes rt3
Description of the codes: RT3

Author: F. Evan (Colorado State University)

Language: Fortran 77

Input:

Disadvantages:

1) pre-computed sets of output angles (interpolation might be needed)

2) no embedded MIE-code (combination with a MIE-code is needed to simulate aerosols)

5

description of the codes sharm

1.atmMIE

config.par

1L.sfc

Description of the codes: SHARM

Author: T. Muldashev (Space Research Institute, Kazakhstan)

Modified: A. Lyapustin

Language: C/C++

Input:

Advantages:very fast, simultaneous simulations for multiple geometries and wavelengths

6

description of the codes modtran

card 1

card 1a

card 2

tape 5 – molecular atmosphere

Description of the codes: MODTRAN

Author: Berk et al. (Air Force Research Laboratory)

Language: Fortran 77

Modeling Features: molecular atmospheres (a lot of effort is put into gas absorption!), aerosols (with the help of DISORT at 16 Gaussian angles), clouds, surface

Input: in the form of formatted “cards” (quite painful!)

Output: single geometry but for a range of wavelengths

7

project history

6SV

&

&

RT3

&

SHARM

Project History

discussions, calculations,Web site creation ...

6SV

&

VPD

MODTRAN

VPD

&

RT3

2005

2006

2007

2008

&

SHARM

Coulson’s tables

MonteCarlo

Why do you ignore MODTRAN?

8

goals of the project
Goals of the project
  • to evaluate the accuracy of each code based on the comparison with standard benchmark references such as Coulson’s tabulated values and a Monte Carlo approach
  • to illustrate differences between individual simulations of the code
  • to determine how the revealed differences influence on the accuracy of aerosol optical thickness and surface reflectance retrievals
  • to create reference (benchmark) data sets that can be used in future code comparison studies

9

presentation of the results
Presentation of the results

All results will be put on the Internet and summarized in a manuscript titled “Radiative Transfer Codes for Atmospheric Correction and Aerosol Retrievals: Intercomparison Study” which will be submitted to Applied Optics.

http://rtcodes.ltdri.org

10

characterization of a rt code
Characterization of a RT code

In regard to remote sensing applications...

1. Versatility

2. Accuracy

3. User-friendliness

4. Speed

6SV1.1, SHARM, MODTRAN, VPD & RT3 (RT3 needs to be combined with a MIE-code)

to be determined

6SV1.1 & SHARM, RT3, MODTRAN, VPD (VPD is not publicly available)

to be determined

11

code accuracy
Code Accuracy

The general atmospheric RT code accuracy requirement for pure simulation studies is 1%.

Reference: Muldashev et al., Spherical harmonics method in the problem of radiative transfer in the atmosphere-surface system, Journal of Quantitative Spectroscopy and Radiative Transfer, 61(3), 393-404, 1999.

Will violation of this requirement have a significant effect on the resulting satellite product?

Step 1: comparison with benchmarks to see if there is violation

Step 2: evaluation of the impact of violation

12

benchmarks coulson s tables
Benchmarks: Coulson’s tables

Coulson’s tabulated values represent the complete solution of the Rayleigh problem for a molecular atmosphere.

Reference: Coulson et al., Tables related to radiation emerging from a planetary atmosphere with Rayleigh scattering (1960).

13

benchmarks monte carlo
Benchmarks: Monte Carlo

The code is written by F.M. Bréon (le Laboratoire des Sciences du Climat et de l'Environnement, France) based on the Stokes vector approach.

Languages: Fortran, C.

Limitations: large amounts of calculation time and angular space discretization.

14

comparison procedure

direction of incident light

larger particles

direction of incident light

Comparison Procedure

1. Molecular Atmosphere (surf = 0.0; 0.25)

The same procedure was usedin the previous comparison study: A. Lyapustin “Radiative transfer code SHARM-3D for radiance simulations over a non-Lambertian nonhomogeneous surface: intercomparison study”, Applied Optics, 41(27), 5607-5615.

2. Aerosol Atmosphere (surf = 0.0)

3. Mixed Atmosphere (surf = 0.0; 0.25)

surf is the reflectance of a Lambertian surface

15

molecular atmosphere conditions
Molecular Atmosphere: Conditions

All RT codes are compared to the Coulson’s tabulated values.

* mol is the molecular optical thickness

 is the wavelength

surf is the surface reflectance

θs is the sun zenith angle

θv is the view zenith angle

φ is the relative azimuth

** Monte Carlo is used only as an auxiliary means here.

16

molecular atmosphere results
Molecular Atmosphere: Results

We calculate the absolute values of average relative differences:

17

aerosol atmosphere conditions
Aerosol Atmosphere: Conditions

6SV1.1 is compared to Monte Carlo and then the other codes are compared to 6SV1.1...

18

aerosol atmosphere results compared to mc
Aerosol Atmosphere: Results (compared to MC)

... 6SV1.1 can be used as benchmark because it demonstrates good agreement with MC

θs = {0.0°, 23.0°, 50.0°}

black soil

19

aerosol atmosphere results compared to 6sv
Aerosol Atmosphere: Results (compared to 6SV)

We calculate the absolute values of average relative differences:

* maer is the selected aerosol model

20

mixed atmosphere conditions

Ozone,

Stratospheric Aerosols

20 Km

O2, CO2

Trace Gases

8 Km

Molecules (Rayleigh Scattering)

2-3 Km

H2O,

Tropospheric Aerosol

Ground Surface

Mixed Atmosphere: Conditions

We simply added a molecular atmosphere to all considered aerosol models.

Profiles:

Mixture:

Molecular optical thickness:

 = 412 nm - mol = 0.303

 = 440 nm - mol = 0.232

 = 670 nm - mol = 0.042

21

mixed atmosphere results compared to mc
Mixed Atmosphere: Results (compared to MC)

6SV1.1 demonstrates relatively good agreement with MC (within 0.85%)

Molecular + Urban-industrial aerosol

aer = 0.2, θs = 0.0°

θs = {0.0°, 23.0°, 50.0°}

mol = 0.303

black soil

aer = 0.8, θs = 0.0°

22

mixed atmosphere results compared to 6sv
Mixed Atmosphere: Results (compared to 6SV)

Again, we calculate the absolute values of average relative differences:

23

accuracy vs speed
Accuracy vs. Speed

Time for 1 run (the case of a mixed atmosphere (λ = 440 nm, AF, aer = 0.8) + surface):

SHARM: ≈ 5.6 s (7.3 s for a number of angles 6 x 16 x 3)

6SV1.1: ≈ 3 s (this time x number of SZA)

Monte Carlo: ≈ 45 min (for one SZA)

Time is important: code comparison like this one

Time is not that important: calculation of LUTs

Accuracy depends on many factors:

SHARM: the number of harmonics

6SV1: the number of Legendre coefficients, calculation layers and angles

24

vpd for a molecular atmosphere

Aerosol atmosphere by MODTRAN:

VPD for a molecular atmosphere

Molecular atmosphere by VPD:

Good results for a molecular atmosphere do not mean that the accuracy of aerosol simulations will be satisfactory!

25

error on aot retrieval theory

1)

is the TOA reflectance of a vector code,

is the TOA reflectance of a scalar code,

is the error of a scalar code,

2)

From (1) and (2) we can calculate the AOT retrieval error:

, where

Error on AOT Retrieval: Theory

The accuracy of 6SV retrievals ?

1% (compared to MC)

Assumption: TOA reflectance is a linear function of AOT

26

error on aot retrieval results
Error on AOT retrieval: Results

Molecular + Aerosol (African Savanna,aer = 0.2):

SHARM:aer = 0.2 ± 0.14

6SV:aer = 0.2 ± 0.01

27

error on aot retrieval results cont
Error on AOT retrieval: Results (Cont.)

Molecular + Aerosol (African Savanna,aer = 0.8):

SHARM:aer = 0.8 ± 0.15

6SV:aer = 0.8 ± 0.05

28

aot retrieval from modis data

490 nm

470 nm

412 nm

443 nm

470 nm

670 nm

0.5

0.2

0.4

AOT

Ex.: Part of Arabian Peninsula, day 207 of 2005

Ex.: AERONET site Alta Floresta, day 197 of 2003

AOT Retrieval from MODIS data

MODIS Land Surface Reflectance algorithm by Vermote et al.

Multi-Angle Implementation of Atmospheric Correction for MODIS by Lyapustin & Wang

29

error on sr retrieval theory

1)

2)

The SR retrieval error:

, where

Error on SR Retrieval: Theory

The same procedure as for AOT retrievals, but is replaced by surface reflectance L (L = 0.05, dL = 0.01)

30

error on sr retrieval results
Error on SR Retrieval: Results

Molecular + Aerosol (Amazonian Forest, aer = 0.2) + Surface (Lambertian,surf = 0.05):

SHARM:surf = 0.05 ± 0.01

6SV:surf = 0.05 ± 0.002

31

error on sr retrieval results cont
Error on SR Retrieval: Results (Cont.)

Molecular + Aerosol (Amazonian Forest, aer = 0.8) + Surface (Lambertian,surf = 0.05):

SHARM:surf = 0.05 ± 0.01

6SV:surf = 0.05 ± 0.003

32

vector scalar for remote sensing

pre-assigned set of aerosol models:

Smoke LABS

Smoke HABS

Urban POLU

Urban CLEAN

+

Vector ? Scalar for Remote Sensing

Is it important to use a vector code?

1) AOT (+ other aerosol properties) retrievals:

important, from a theoretical point of view

2) surface reflectance retrievals:

important

The accuracy of LUTs directly depends on the RT code simulations.

The best solution is to calculate a product error budget.

33

reference data set
Reference data set

The goal is to use 6SV1 to create a reference data set for further code comparison studies.

34

slide35
...

Thank you for your attention!

Questions?..

skotchen@ltdri.org

35