Comparison of atmospheric corrections between MODTRAN and 6S models

 

The purpose of atmospheric correction is to eliminate the influence of atmospheric and light factors on the reflection of ground objects, obtain the real physical model parameters such as ground object reflectivity and emissivity, and surface temperature, and use them to eliminate the influence of water vapor, oxygen, carbon dioxide, methane and ozone in the atmosphere on the reflection of ground objects, and to eliminate the influence of atmospheric molecules and aerosol scattering.

6S vs. MODTRAN: Atmospheric Correction and Radiative Transfer  

6S (Second Simulation of the Satellite Signal in the Solar Reflective Spectrum) and MODTRAN (Moderate Resolution Atmospheric Transmission) are two widely used radiative transfer models for atmospheric correction and spectral simulation of remote sensing imagery. They differ significantly in function, scope of application, and operational procedures. The following comparison focuses on their core functions, band range, and applicable scenarios.

01. Core Functions and Problem Types

  MODTRAN：   
By inputting surface reflectance, atmospheric parameters (aerosol optical thickness, AOT; water vapor content; gas concentration, etc.) and sensor parameters, the radiance or apparent reflectance of the top of the atmosphere (TOA) can be simulated.

Features: Provides high-precision radiative transfer simulation and supports complex atmospheric conditions (such as stratified atmosphere and non-uniform aerosols).

Typical applications: hyperspectral data analysis, sensor design, and spectral simulation in complex environments.

6S：

By inputting the apparent top-of-atmosphere reflectance (TOA Reflectance) and atmospheric parameters (AOT, water vapor, ozone, etc.), the surface reflectance is directly inverted based on a pre-calculated lookup table (LUT).

Features: Easy to operate, designed specifically for solar reflectance spectrum, suitable for standardized atmospheric correction.

Typical application: Atmospheric correction for multispectral satellites (such as Landsat-8 and Sentinel-2).

02. Band Range

 MODTRAN：   
It covers 0.2 μm to 100 μm (ultraviolet, visible, near-infrared, short-wave infrared, mid-infrared to far-infrared) and supports thermal infrared emission and absorption simulation.

Suitable for hyperspectral data (such as PRISMA, AVIRIS) and wideband applications.

6S：

It covers the range from 0.25 μm to 4.0 μm (visible to shortwave infrared, solar reflectance spectrum range), but does not involve thermal infrared.

Designed specifically for multispectral and partially hyperspectral data, it is suitable for the analysis of targets such as vegetation and soil in the VNIR-SWIR bands.

03. Applicable Scenarios and Advantages/Disadvantages

 MODTRAN：

Advantages: Supports wide bands and high precision, suitable for complex atmospheric conditions and hyperspectral data.

Disadvantages: High computational complexity, requires manual inversion, long running time, suitable for research applications.

Scenarios: Hyperspectral image correction (e.g., limestone mineral identification, 2.31–2.33 μm), atmospheric chemistry simulation.

6S：

Advantages: Fast, easy to use, suitable for standardized multispectral data correction, and has extensive community support.

Disadvantages: Limited band range (no thermal infrared) and limited support for complex atmospheric conditions.

Scenario: Agricultural remote sensing (such as red-edge position index calculation), vegetation monitoring.

Operating procedures

1. The FLAASH tool in ENVI uses the MODTRAN model.
   The data used is from the XIGUANG-002 satellite, with 150 bands and a spectral range of 440-840 nm.

The standard data name is: XIGUANG-002_20250301_0000000001_0222_0064_W106.1_N43.7_L1BF_HSI. This data is the DN value data for completing the system geometric correction.

Open the XML file using the XiGuang-XG02 tool in the ENVI China Satellite plugin.

Automatically read metadata to obtain wavelength, FWHM, and radiometric calibration factor.

Radiometric calibration was performed using the Radiometric Calibration tool.

The scaling factor is set to match the FLAASH tool. After radiometric calibration, the data unit changes from DN value to µW/(cm2 * sr * nm).

Next, atmospheric correction is performed using the FLAASH tool, where time-of-flight data is automatically read from metadata. Sensor altitude and average elevation need to be additionally obtained from the XML file. These correspond to the following fields in the XML file:

 485607.870514

 1490.9 Meters

For atmospheric models, refer to the lookup table below; for aerosol models, select according to the type of target area.

2. The 6S model was implemented using the Python library Py6s.

After establishing a 6S model and configuring basic information such as geographic information, solar information, atmospheric model, aerosol model, sensor height, and target height, the model is run band by band and missing bands such as water vapor absorption windows are filled by interpolation. Finally, the output is the same number of surface radiance and reflectance data as the input.

 

Results Display

Below are some atmospheric correction results for XIGUANG-002 data:

Yuncheng City, Shanxi Province, China

Salakhstan, Turkmenistan 

The following is a comparison of results using two atmospheric correction models.

The left side shows the results from the MODTRAN atmospheric correction model (reflectivity magnified 10,000 times), and the right side shows the results from the 6S model.

Because MODTRAN is a more refined molecular absorption spectrum-based model, and ENVI's FLAASH tool integrates post-processing such as spectral smoothing, the oxygen absorption band near 760 is smoothed, while the 6S model is based on physical simulation, so each gas absorption band is preserved.

Overall, the spectral curves show similar trends, with the 6S model slightly higher in absolute reflectance. The MODTRAN model has higher accuracy and reliability, while the 6S model is an open-source model, easy to deploy and further develop, allowing for the selection of different correction methods based on different application scenarios.