Fast Line-of-sight Atmospheric Analysis of Hypercubes (FLAASH®)

Preprocessing AVIRIS Data Tutorial (demonstrates the use of FLAASH), Example: Multispectral Sensors and FLAASH

FLAASH is a first-principles atmospheric correction tool that corrects wavelengths in the visible through near-infrared and shortwave infrared regions, up to 3 µm. (For thermal regions, use the Toolbox option Radiometric Correction > Thermal Atmospheric Correction.) FLAASH works with most hyperspectral and multispectral sensors. Water vapor and aerosol retrieval are only possible when the image contains bands in appropriate wavelength positions. FLAASH can correct images collected in either vertical (nadir) or slant-viewing geometries. See Background on FLAASH for a scientific discussion of the methods used by FLAASH.

You can also write a script to run FLAASH using the FLAASH task.

See the following sections:

Before You Begin

Radiance and Scale Factors

This section provides guidelines on how to determine the correct scale to apply in FLAASH. Instead of multiplying the scale factor, FLAASH divides the scale factor into the radiance data.

FLAASH requires input data to be floating-point values in units of µW/(cm2 * sr * µm). If the input radiance image does not meet these criteria, you must know the scale factor(s) used to convert radiance data into these units.

Select Input Files

  1. From the Toolbox, select Radiometric Correction > Atmospheric Correction Module > FLAASH Atmospheric Correction. The Data Selection dialog appears.
  2. Select an input file and perform optional spatial and spectral subsetting and/or masking.
  3. Click OK. The FLAASH - Rigorous Atmospheric Correction dialog appears, with the Main tab selected.

Specify Output Files

Use the Main tab of the FLAASH dialog to specify the output files to create.

  1. From the Acquisition Date/Time drop-down lists Day, Month, Year, Hours, Minutes, and Seconds, specify when the scene was collected. Use Greenwich Mean Time. You can also import the settings from the dataset by clicking the Import Dataset button and selecting the dataset to use. These values and the Scene Center values are required to calculate Solar Zenith and Solar Azimuth.

  2. In the Output Cloud Raster field, specify a location and filename for the cloud map classification image. This output is only available with hyperspectral input imagery. The cloud determinations are used during FLAASH processing to refine the estimate of re in equation (1) in Background on FLAASH. This quantity defines the magnitude of the adjacency effect. For details on the cloud classes in the output cloud raster, see Cloud Classes.

  3. In the Output Water Raster field, specify a location and filename for the column water vapor image. This output is only available with hyperspectral input imagery. The image is of column water vapor in units of (atm • cm).

  4. Enter a filename and location for the Output Raster.

Next, set processing parameters in the Sensor tab.

Set Scene and Sensor Options

Scene and sensor details give FLAASH an approximation of the sun's position relative to the surface. To set these options, select the Sensor tab in the FLAASH dialog.

  1. From the Sensor Type drop-down, select the name of the sensor that collected the input radiance image. The options are:

    • AVRIS

    • HYDICS

    • HyMap

    • Hyperion

    • NVIS

    • Compass

    • Hycas

    • Mars

    • Multispectral

    • Unknown: Select this option if you do not know the sensor type, or want to provide custom values.

  2. For sensor types with known values, the Input Scale field will be populated with the default radiometric scale values for the sensor. You can specify single values or an array of custom values. To import a text file that contains the input scale, click the Load Values button and select a file. The text file must be single-column data, with the number or rows matching the number of bands. To save the specified input scale to a text file, click the Save button .

  3. In the Output Reflectance Scale Factor field, enter the value to use to scale the output reflectance image from floating-point into 2-byte integer data space. Signed integers have an absolute data range between -32768 and +32767. The default scale factor is 10,000. This value is stored as the Reflectance Scale Factor Attribute in the header file of the output image.

  4. Select a Calibration File Format from the drop-down list. A calibration file assigns the spectral features that are used to correct the wavelength calibration. The FLAASH calibration file format includes built-in definitions for the AVRIS, HyMap, HYDICE, Hyperion, CASI, AISA, and multispectral sensors. If the input image is from another sensor, or if you want to customize the built-in definitions, select another format from the list and specify your own calibration file.

    See Calibration File Formats for examples of the file formats, or click on a link below.

    • ASCII: A two-column text file which specifies channel center and width values. For use with hyperspectral sensors.

    • Discrete: A text file defining channel shapes as paired wavelength and response values. For use with hyperspectral sensors.

    • ENVI: An ENVI-style cube header with wavelength, FWHM, and BBL arrays. For use with hyperspectral sensors.

    • Filter Format: A discrete channel shape data defined within the FLAASH input configuration. MSI sensors often use spectral filters that have complicated shapes. FLAASH uses critical sampling from FWHM to adjust for these shape variations; however, you may get a more accurate result by providing a filter function file containing a discrete array of wavelengths vs response. This file is often provided by the data provider. To use a filter function, select a filter function file in spectral library format (.sli). The file should contain the sensor’s filter functions consecutively, one for each band of the sensor.

    • FLAASH: (default) Sensor calibration data is defined within the FLAASH input configuration, you do not need to specify a calibration file.

    • Spectograph: An ASCII text file containing one line for each spectrometer in the sensor.

  5. For all formats other than FLAASH, specify a filename in the Calibration File field.

    Note: If you find that the corrected wavelengths are an improvement after FLAASH processing is complete, you may want to apply them to the original input radiance image. See Wavelength Calibration for details.

  6. If you specified a calibration file, select Nanometers or Microns for the units from the Calibration Units drop-down list.

  7. The Instantaneous Field of View field defines the angle of individual sensors. The value in this field is derived from the Sensor Altitude setting on the Geometric tab.

Next, set viewing geometry parameters in the Geometric tab.

Set Viewing Geometry Options

To set viewing geometry options, select the Geometric tab in the FLAASH dialog.

For instruments that use a non-nadir viewing geometry, you must specify the zenith and azimuth angles. The zenith angle is defined at the sensor as the angle between the line of sight and the zenith (180 for nadir-viewing sensors). Zenith angles must be positive and between 90 and 180 degrees. The azimuth angle is defined as the azimuth (the angle between the line of sight and due North) of the sensor as viewed from the ground. This angle is arbitrary for nadir-viewing sensors. The azimuth angle is measured as degrees clockwise from North to the line-of-sight vector. Azimuth values must be between -180 and 180 degrees. For example, if your azimuth from the sensor is 90 degrees (due east), then the azimuth from the ground is -90 degrees (due west).

  1. If the input scene has map information, the Scene Center fields Lat and Lon will automatically populate with the coordinates. If no map information is available, enter the latitude and longitude of the scene center, respectively, using negative values for Southern and Western Hemispheres. These values and the Acquisition Date/Time values are required to calculate Solar Zenith and Solar Azimuth.

  2. In the Sensor Altitude field, enter the altitude of the sensor when the image was collected, in kilometers above sea level. For certain sensors, the altitude is automatically set when you select the sensor type.

  3. Ground Elevation calculated by using the default world DEM shipped with ENVI. If the data is not georeferenced, then enter the average scene elevation, in kilometers above sea level.

  4. The Solar Zenith value is calculated when the Scene Center and Acquisition Date/Time values are provided. To use a custom value, specify the relative solar angle as measured from the ground at the imaged site. A zenith angle of 0 degrees is directly overhead while 90 degrees is on the horizon.

  5. The Solar Azimuth value is calculated when the Scene Center and Acquisition Date/Time values are provided. To use a custom value, specify the relative solar angle as measured from the ground at the imaged site. The azimuth angle is measured as degrees clockwise from North to the line-of-sight vector of the sun.

  6. For data with an off-nadir angle of more than 5 degrees, it is recommended that you enter the Line of Sight, in degrees. This is measured as degrees clockwise from North to the line-of-sight vector.

  7. For data with an off-nadir angle of more than 5 degrees, it is recommended that you enter the Line of Sight Zenith, in degrees.

Next, set parameters for the atmospheric model.

Set Atmospheric Model Options

To set atmospheric model options, select the Model tab in the FLAASH dialog.

  1. Select one of the standard MODTRAN model atmospheres from the Atmospheric Model drop-down. The options are:

    • Tropical Atmosphere (default (15 degrees North Latitude)

    • Mid-Latitude Summer (45 degrees North Latitude)

    • Mid-Latitude Winter (45 degrees North Latitude)

    • Sub-Arctic Summer (60 degrees North Latitude)

    • Sub-Arctic Winter (60 degrees North Latitude)

    • 1976 US Standard Atmosphere

  2. From the Modtran Resolution drop-down list, select a resolution. The options are:

    • 15.0: 15 wavenumber (cm-1)

    • 5.0: 5 wavenumber (cm-1)

    • 1.0: 1 wavenumber (cm-1)

    • 0.1: 1/10th wavenumber (cm-1)

    The Modtran Resolution setting controls the MODTRAN spectral resolution and the trade-off of speed versus accuracy for the MODTRAN portion of the calculation. Lower resolution yields proportionally better speed but less accuracy. The main differences in accuracy are seen near 2000 nm and beyond. The 5 cm-1 resolution is the default value when you select a hyperspectral sensor as input, but it changes to 15 cm-1 when you select a multispectral sensor. If aerosol is being retrieved, there are two MODTRAN runs performed at 15 cm-1 resolution followed by one MODTRAN run at the resolution you select. If aerosols are not being retrieved, the first two runs are omitted.

  3. In the Modtran Multiscatter Model drop-down list, select a multiple-scattering algorithm to be used by MODTRAN. The multiscatter models options are:

    • ISAACS

    • DISORT

    • Scaled ISAACS (default)

    Multiple scattering becomes more significant in the presence of heavy aerosol loading and/or higher water vapor concentrations. The DISORT model provides the most accurate shortwave (less than ~ 1000 nm) corrections, however it is very computationally intensive. The ISAACS method is fast but oversimplified. The Scaled ISAACS is fast and more accurate than the ISAACS.

  4. In the CO2 Mixing Ratio field, enter the carbon dioxide (CO2) mixing ratio in parts per million by volume. The default is to estimate for the year of the acquisition based on historical trends.

For the best results, select an atmospheric model whose standard column water vapor amount is similar to, or somewhat greater than, that expected for the scene. The standard column water vapor amounts (from sea level to space) for each model atmosphere are in the following table:

Model Atmosphere

Water Vapor
(std atm-cm)

Water Vapor

(g/cm2)*

Surface Air Temperature

Sub-Arctic Winter (SAW)

518

0.42

-16° C (3° F)

Mid-Latitude Winter (MLW)

1060

0.85

-1° C (30° F)

U.S. Standard (US)

1762

1.42

15° C (59° F)

Sub-Arctic Summer (SAS)

2589

2.08

14° C (57° F)

Mid-Latitude Summer (MLS)

3636

2.92

21° C (70° F)

Tropical (T)

5119

4.11

27° C (80° F)

* The “atm-cm” unit is specific to the atmospheric science community, which typically uses one of two units to measure the total amount of a gas in the atmospheric column from the ground to the top of the atmosphere (where 200 to 300 km is generally a good number for the location of the top).

When you use units of “atm-cm,” think of it as bringing all the water molecules down to a thin layer of pure water vapor at the Earth's surface, at 1 atm of pressure and 0° C. That layer has a thickness measured in centimeters, so the water column is described in atmosphere-centimeters. If the pressure were doubled, then the thickness would be halved. Thus, the units of atm-cm (not just cm) are used to describe the amount of gas in the atmospheric column to emphasize that the height and pressure are interdependent.

Obviously it is not physically possible to bring these molecules into such a condensed layer. All of the molecules in the layer would condense into a liquid under these conditions, even if they could be gathered in such a way. The layer is imaginary.

The second set of units, gm/cm2, is more easily understood as the mass of water molecules in the atmospheric column over each cm2 of ground surface. Since liquid water has a 1 gm/cm2 density, this value is numerically equal to the number of centimeters of water on the ground if all the atmospheric water rained out at once.

If no water vapor information is available, select an atmosphere according to the known or expected surface air temperature, which tends to correlate with water vapor. If the temperature is unknown, select an atmosphere from the following table, which is based on a seasonal-latitude surface temperature model.

Latitude (°N)

Jan

March

May

July

Sept

Nov

80

SAW

SAW

SAW

MLW

MLW

SAW

70

SAW

SAW

MLW

MLW

MLW

SAW

60

MLW

MLW

MLW

SAS

SAS

MLW

50

MLW

MLW

SAS

SAS

SAS

SAS

40

SAS

SAS

SAS

MLS

MLS

SAS

30

MLS

MLS

MLS

T

T

MLS

20

T

T

T

T

T

T

10

T

T

T

T

T

T

0

T

T

T

T

T

T

-10

T

T

T

T

T

T

-20

T

T

T

MLS

MLS

T

-30

MLS

MLS

MLS

MLS

MLS

MLS

-40

SAS

SAS

SAS

SAS

SAS

SAS

-50

SAS

SAS

SAS

MLW

MLW

SAS

-60

MLW

MLW

MLW

MLW

MLW

MLW

-70

MLW

MLW

MLW

MLW

MLW

MLW

-80

MLW

MLW

MLW

MLW

MLW

MLW

Next, set parameters for water retrieval.

Set Water Retrieval Options

To set water retrieval options, select the Water tab in the FLAASH dialog.

To solve the radiative transfer equations that allow apparent surface reflectance to be computed, the column water vapor amount for each pixel in the image must be determined. FLAASH includes a method for retrieving the water amount for each pixel. This technique produces a more accurate correction than using a constant water amount for the entire scene. To use this water retrieval method, the image must have bands that span at least one of the following ranges at a spectral resolution of 15 nm or better:

For most of the multispectral sensor types, the Water Retrieval setting is No because these sensors do not have the appropriate bands to perform the retrieval.

  1. Specify the absorption feature to use with the water vapor retrieval in the Water Absorption Wavelength drop-down list, select one of the following:

    • 1130

    • 940

    • 820

    • Automatic Selection (default)

  2. In the Water Column Multiplier fields, specify the Min and Max water column multiplier for the Atmospheric Model that was selected in the Model tab. The defaults are 0.01, and 1.

  3. In the Water Vapor Preset field, optionally enter a value to override the water retrieval with a constant scale factor. The default of 0 indicates to use water retrieval values. Entering a positive value will override water vapor retrieval.

  4. Next, set parameters for aerosols.

Set Aerosol Options

To set aerosol options, select the Aerosol tab in the FLAASH dialog.

  1. Select one of the following models from the Aerosol Model drop-down. The choice of model is not critical if the visibility is high (for example, greater than 40 km):

    • No Aerosol: Do not include aerosol or cloud attenuation in the calculation.
    • High-Visibility Rural: Represents aerosols in areas not strongly affected by urban or industrial sources. The particle sizes are a blend of two distributions, one large and one small, where dust-like aerosol makes up the majority of the course particles. For clear to very clear conditions (visibility) the vertical distribution of the aerosol extinction coefficient is exponential.
    • Low-Visibility Rural: Represents aerosols in areas not strongly affected by urban or industrial sources. The particle sizes are a blend of two distributions, one large and one small, where dust-like aerosol makes up the majority of the course particles. For hazy conditions (visibility) the vertical distribution of the aerosol extinction coefficient is assumed to be independent of height up to 1 km with a pronounced decrease above that height.
    • Maritime: Represents the boundary layer over oceans, or continents under a prevailing wind from the ocean. It is composed of two components, one from sea spray and another from rural continental aerosol (that omits the largest particles).
    • Urban: A mixture of 80% rural aerosol with 20% soot-like aerosols, appropriate for high-density urban/industrial areas.
    • Tropospheric: Applies to calm, clear (visibility greater than 40 km) conditions over land and consists of the small-particle component of the rural model.

  2. From the Use Aerosol drop-down list, select the aerosol retrieval algorithm to use to obtain scene visibility. If the selected retrieval fails or if this option is disabled, the default visibility is used. Aerosol retrieval options include:

    • Disabled

    • Automatic Selection

    • Vegetation Based Retrieval (the input image must contain dark vegetation spectra)

    • Water Based Retrieval (the input image must contain bodies of water)

    • Wavelength Dependent Water Based Retrieval

    • Linear Regression Retrieval

  3. In the Aerosol Scale Height field, enter the effective 1/e height of the aerosol vertical profile in km. Typical values are 1 to 2 km. The default value is 2 km. The aerosol scale height is used only for the calculation of the adjacency scattering range.
  4. Specify a target ratio of reflectance values between two bands. Ratio = (upper band) / (lower band) in the Aerosol Band Ratio field.
  5. In the Aerosol Band Wavelength field, specify wavelengths [nanometers] for the band of a custom single-band aerosol retrieval. Specify a single wavelength or an interval.
  6. In the Aerosol Reference Value field, specify the expected reflectance value of a pixel to use in the single band aerosol retrieval. This is primarily useful when a target material is present. (default = 0).
  7. Specify the location of a pixel (sample, line) to use in a single band aerosol retrieval in the Aerosol Reference Pixel field. If undefined, the darkest in-band pixel will be used as a reference.
  8. In the Aerosol Bandlow Wavelength field, specify wavelengths [nanometers] for the lower band of a custom 2-band aerosol retrieval. Specify a single wavelength or an interval.
  9. Specify a maximum reflectance threshold for the lower band of a custom 2-band aerosol retrieval in the Aerosol Bandlow Max Reflectance field. Pixels brighter than this value will be excluded from the aerosol retrieval process.
  10. In the Aerosol Bandhigh Wavelength field, specify wavelengths [nanometers] for the upper band of a custom 2-band aerosol retrieval. Specify a single wavelength or an interval.
  11. Specify a maximum reflectance threshold for the upper band of a custom 2-band aerosol retrieval in the Aerosol Bandhigh Max Reflectance field. Pixels brighter than this value will be excluded from the aerosol retrieval process.

Reference:

Abreu, L. W., and G. P. Anderson, Eds. 1996. The MODTRAN 2/3 report and LOWTRAN 7 model. Phillips Laboratory, Geophysics Directorate, PL/GPOS, Hanscom AFB, MA. Contract F19628-91-C-0132.

Set Miscellaneous Options

  1. From the Adjacency Correction drop-down list, specify how to perform adjacency correction:

    • Disabled

    • Legacy exponential scattering kernel

    • Wavelength-dependent scattering kernel (default)

    Unlike most atmospheric correction models, the FLAASH model accounts for both the radiance that is reflected from the surface that travels directly into the sensor and the radiance from the surface that is scattered by the atmosphere into the sensor. The distinction between the two accounts for the adjacency effect (spatial mixing of radiance among nearby pixels) caused by atmospheric scattering. More accurate reflectance retrievals result when adjacency correction is used; however, there may be occasions when it is desirable to ignore this effect. Disabling adjacency correction sets ρe = ρ in Equation (1) in Background on FLAASH.

  2. In the Initial Visibility field, enter an estimate of the scene visibility in kilometers. The initial visibility value is assumed for atmospheric correction if the aerosol is not being retrieved. FLAASH will use this value if the aerosol cannot be retrieved. The following table gives the approximate scene visibility values based on weather conditions:

    Weather Condition

    Scene Visibility

    Clear

    40 to 100 km

    Moderate Haze

    20 to km

    Thick Haze

    15 km or less

    The visibility, V is defined as the 550 nm meteorological range and is related to extinction coefficient β (base e) by the equation V = 3.912/β. The extinction coefficient β is defined as the horizontal optical depth per km. A related value, the aerosol optical depth (AOD) is measured vertically (from the ground to space). To convert the AOD to β, divide AOD by the effective aerosol thickness layer, which typically has a value of around 2 km, but varies with the visibility, elevation, and other factors.

  3. From the Spectral Polishing drop-down list, select one of the following options. Spectral polishing is a technique that reduces spectral artifacts in hyperspectral data. See Background on FLAASH for details on how spectral polishing is applied.

    • Disabled: (default) No spectral polishing.

    • Polish using reference materials: Remove residual bumps and valleys present in the retrieved reflectance spectra. This method attempts to match reflectance spectra within a scene against a reference library of materials (mostly soils and rocks). A compensation factor is calculated for each band using the deviations between the pairs of matching scene/library spectra.

    • Polish using statistical detection of spectral artifacts: Remove residual bumps and valleys present in the retrieved reflectance spectra. This method calculates the polishing factors using a comparison between a spectrum and a lowpass-filtered version of the spectrum. The method avoids evaluating materials that have inherent structure, such as vegetation and certain minerals, otherwise the polishing vector may unintentionally remove physical features.

  4. In the Polishing Width field, specify the width, in bands, of the smoothing window to be used in the FLAASH spectral polishing algorithm. A larger number generates more smoothing. A value of 9 is recommended for typical 10 nm-resolution hyperspectral sensors (such as AVIRIS). A value of 2 provides minimal smoothing but removes odd-even spectral band imbalances. Odd polishing widths are suggested because they are more computationally efficient.

  5. Select Yes or No to enable or disable Sensor Autocalibration. Sensor autocalibration adjusts the center wavelengths of the sensor bands. It calculates an analog of atmospheric transmission from the image data and compares those values to the known positions of reference absorption features. The adjustment happens independently for each detector if a sensor has more than one detector. Sensor autocalibration will not work on multispectral sensors. If the wavelengths of the original input radiance image were corrected in the ENVI Header file after previous FLAASH processing, set this to No.

  6. The Autocalibration Precision value defines a threshold that determines whether or not an absorption feature is distinct enough to adjust the wavelength calibration. It is a variance is calculated from at least 30 pixels, using the apparent position of a feature and the known position of the feature. Small variances indicate that positions differ by a consistent value and we can therefore adjust the calibration using that feature.

  7. If Sensor Autocalibration is enabled, all of the spectral features listed below are used to by default adjust sensor calibration. To use only a subset of these features, specify the ones to use in Sensor Calibration field. Based on past performance, the first three features in the list are the most reliable.

    Spectral features: O2_763, H2O_940, CO2_2040, Solar_H_434, Solar_H_486, Solar_Mg_517, Solar_H_656, O2_691, H2O_820, H2O_1130, O2_1266, CO2_1614.

References

K. Stamnes, S.-C. Tsay, W. Wiscombe, and K. Jayaweera. Numerically Stable Algorithm for Discrete-Ordinate-Method Radiative Transfer in Multiple Scattering and Emitting Layered Media. Applied Optics. Vol. 27. 1988. pp. 2502-2509.

R. G Isaacs, W. C. Wang, R. D. Worsham, and S. Goldenberg. Multiple Scattering LOWTRAN and FASCODE Models. Applied Optics. Vol. 26. 1987. pp. 1272-1281.

Create the Output

  1. Enable the Preview check box to see a preview of the settings before you click OK to process the data. The preview is calculated only on the area in the view and uses the resolution level at which you are viewing the image. See Preview for details on the results. To preview a different area in your image, pan and zoom to the area of interest and re-enable the Preview option.

  2. To reuse these task settings in future ENVI sessions, save them to a file. Click the down arrow and select Save Parameter Values, then specify the location and filename to save to. Note that some parameter types, such as rasters, vectors, and ROIs, will not be saved with the file. To apply the saved task settings, click the down arrow and select Restore Parameter Values, then select the file where you previously stored your settings.

  3. To run the process in the background, click the down arrow and select Run Task in the Background. If an ENVI Server has been set up on the network, the Run Task on remote ENVI Server name is also available. The ENVI Server Job Console will show the progress of the job and will provide a link to display the result when processing is complete. See ENVI Servers for more information.

  4. Click OK.