from Rocket Science to Business Solutions

Adam M. Thodey

Masters of Engineering in Remote Sensing and GIS

University of Michigan

Ann Arbor, Michigan, USA.

Vittala K. Shettigara

Spaced Based Surveillance Group

Wide Area Surveillance Division

Defence Science and Technology Organisation

Salisbury, South Australia, Australia.

INTRODUCTION

ATMOSPHERIC EFFECTS

COMPUTER MODELING AND RADIATIVE TRANSFER CODES

Atmospheric Correction and water retrieval codes

Adapting atrem to other sensors

Conclusion

Appendix A

Appendix B

Appendix C

Appendix D

Appendix E

Appendix F

ABSTRACT

Images taken remotely contain information about both the target, background and atmosphere. In modelling the affects of the atmosphere and writing a computer code to remove these affects, one can then see the true spectra of the target. ATREM is one such code that derives the effects of the atmosphere and removes it from imagery. In adapting this software package for any sensor, the reflectance spectra of the target can then be detected and classified.

INTRODUCTION

Remote Sensing of the Earth is important in both defence and civilian applications. Within defence, remote sensing plays a part in wide area surveillance in detecting targets. Target detection and identification based on remote sensing imagery is one of the goals of multi- and hyperspectral research in the Defence Science and Technology Organisation (DSTO).

To accurately assess remotely sensed imagery, many factors have to be taken into account before target identification can be done. The main factor is to have an image that is free of any interference. The interference is mainly from the attenuation and scattering of the atmosphere.

This paper will concentrate on a few of the available computer programs that models atmospheric radiation (radiative transfer codes) and one program, ATREM, that removes the effects of the atmosphere from images.

ATMOSPHERIC EFFECTS

The atmosphere affects all remotely sensed imagery taken of targets from both aircraft and satellites. The acquired image is a product of several signals, not just the target itself. Several different processes, gaseous absorption and scattering by molecules and aerosols, affect the signal received by the sensor: (Tanre et al., 1990). These molecules and aerosols in the atmosphere attenuate the radiance, change the spatial distribution, and introduce into the field of view scattered radiation (Zagolski et al. 1994; Bornhoeft et al.).

Of the photons in the sun-surface-sensor ray path, only a fraction reaches the sensor. The other photons are lost to absorption processes or scattering processes by molecules and aerosols (Tanre, 1990).

Approximately thirty constituents in the Earth’s atmosphere affect radiative properties of the signal in the Sun-surface-sensor ray path. Within the region from 0.4 to 2.5 mm, only seven gases have a profound effect on the measured radiance. These constituents are water vapour, carbon dioxide, ozone, nitrous oxide, carbon monoxide, methane, and oxygen.

These seven gases have relatively strong absorption bands. Water vapour is the strongest absorber of the radiant energy and most prevalent among the gases. The major water vapour absorption bands are centred at 0.94, 1.14, 1.38, and 1.88 microns (Gao, 1993). Table 1 shows some of the absorption bands for some of the other atmospheric gases.

Table 1: Absorption bands for some of the gases

COMPUTER MODELING AND RADIATIVE TRANSFER CODES

Over the last twenty years, many people have worked to model the atmosphere. It has only been in the last decade that modelling of the atmosphere has been done on the computer. Many different codes have been created. Several of these codes are described below.

The Simulation of the Satellite Signal in the Solar Spectrum, known as the 5S code, was written at the Laboratoire d’Optique Atmospherique at the Universite des Sciences et Techniques de Lille in France. The 5S code allows for the “estimation of the solar radiation backscattered by the Earth-surface-atmosphere system, as it is observed by a satellite sensor”.

The 5S code was written to model the sensor signal based upon

· statistical band models for the accurate estimation of atmospheric absorption

· accurate modellings of various atmospheric functions for a complete treatment of the scattering processes

· an approximate treatment of the interaction between the two effects: the code first estimates the signal at the sensor without gaseous absorption; if the spectral band exhibits some gaseous contamination, the signal is corrected by the gaseous transmission factor.

The input parameters are the geometrical conditions of the site, atmospheric model for gaseous components, aerosol model, spectral band of observation, and ground reflectance type and spectral variation.

The main atmospheric effects, gaseous absorption by water vapour, carbon dioxide, oxygen and ozone, and scattering by molecules and aerosols in a cloudless atmosphere are taken in to account to predict the apparent reflectance. The gaseous transmittance, the irradiance at the surface and the different contributions to the satellite signal according to the origin of the measured radiance are also produced by the code.

These calculated parameters allow the code to be used with other atmospheric modelling and atmospheric correction codes.

LOWTRAN

LOWTRAN calculates atmospheric transmittance and/or radiance in the microwave, infrared, visible, and near ultraviolet spectral regions. LOWTRAN 7 has a spectral resolution of 20 cm^-1 full width at half maximum (FWHM, see Appendix E for definition) and its calculations are done in 5 cm^-1 increments (Berk, 1989).

MODTRAN (Berk, 1989)

MODTRAN was created to increase the spectral resolution of LOWTRAN from 20 cm^-1 to 2 cm^-1; thereby being a moderate resolution program.

The technical objects of the program were:

· To provide 2cm^-1 resolution (FWHM) algorithms

· To model atmospheric molecular absorption as a function of both temperature and pressure

· To calculate band model parameters for twelve LOWTRAN molecular species

· To integrate the LOWTRAN 7 capabilities into the new algorithms, maintaining compatibility with the multiple scattering option

MODTRAN has six additional subroutines that provide the increased spectral resolution over LOWTRAN. Although MODTRAN uses the same code as LOWTRAN, small changes in the LOWTRAN code allow for a greater resolution for the calculated spectra. MODTRAN allows for LOWTRAN to be executed with a switch.

Three different band model parameters are used for the band model techniques to determine transmittance over finite frequency intervals:

· Absorption coefficient: measures total strength of lines in an interval

· Line density: line-strength weighted average for the number of lines in the interval

· Line width parameter: line-strength weighted average line width

Due to the amount of data generated for calculations, MODTRAN stores it in a binary data file in look up table form for later use.

MOSART (Cornette, 1994)

The Moderate Spectral Atmospheric Radiance and Transmittance (MOSART) code was produced by Photon Research Associates to predict and evaluate the radiative environment to support a variety of atmospheric applications. MOSART includes features from both LOWTRAN and APART programs. Mainly, MOSART is used to do atmospheric correction of satellite imagery and terrain and cloud scene modelling.

HITRAN

HITRAN is a database containing the line spectra of hundreds of species in the 0 – 17,900 cm^-1 spectral region that has significant absorption for atmospheric paths.

Atmospheric Correction and water retrieval codes

The purpose of atmospheric correction programs is to find the true surface reflectance parameters of a scene (Teillet, 1989). The true surface reflectance can then be checked against spectral libraries of targets for target detection and classification.

F. Fowle demonstrated, in 1912, that it was possible to measure atmospheric water amounts using a differential absorption concept (Carrere, 1993). The differential absorption concept has been used in several studies to retrieve column water vapour values (Carrere, 1993;Barducci, 1995; Gao, 1993; Schmid, 1996). This concept can be used with hyperspectral images without the need other measurements (Gao et al., 1993), see ATREM.

Water vapour is probably the parameter that affects the radiance the most. Water vapour is difficult to model due to its variation with time, season, and altitude. Figure 1 shows how water vapour changes temporally in the atmosphere above Roger Dry Lake, California..

Figure 1: Temporal Change of Water Vapour over Rogers Dry Lake.

ATREM

ATREM, atmospheric removal program(Gao et al., 1996), implemented at the Center for the Study of Earth from Space, models atmospheric effects, and removes them from the image; thus retrieving a scaled surface reflectance image. A by-product of the code is the creation of a water vapour image consisting of the water vapour data derived from the original image.

ATREM was written for the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS). Currently, ATREM can only correct for atmospheric effects on AVIRIS imagery although it can be modified to allow for other sensors (see adapting ATREM to other sensors).

ATREM assumes a flat surface scene which does not allow for the retrieval of actual ground reflectance. Rather, the processed images output is scaled due to the height variations of the terrain. Post processing of the scaled ground reflectance processed image against a digital terrain map could yield better results.

Figure 2 show the typical AVIRIS spectrum from 0.4 to 2.5 mm. As you can see in Fig. 2, the curve follows a 6000 Kelvin black body radiation curve with information about the surface and atmosphere within the curve. Figure 3 shows the corrected AVIRIS spectrum processed by ATREM. As you can see, Fig. 3 allows for the spectral signature of the pixel to be clearly seen, as it would be if there was no atmosphere.

Figure 2: AVIRIS spectrum over 0.40 – 2.45 mm.

Figure 3: ATREM corrected image spectrum over 0.40 – 2.45 mm.

Algorithm

The algorithm (see Fig. 4; Appendix B) for ATREM is simple and straightforward:

Figure 4: Flow Diagram of ATREM algorithm.

1. Read in the data from a user file that pertains to the image to be processed.
2. Adjust the bottom boundary of the input model if the surface elevation is greater than 0, and calculate the column water vapour amount in the selected model.
3. Calculate the solar and the observational geometric factors.
4. Predict the signal at the AVIRIS sensor using the 5S code (atmospheric reflectance, downward and upward scattering transmittance, and spherical albedo.
5. Initialise global data for spectrum calculations
6. Obtain solar irradiances above the atmosphere corresponding to the measurement time and geographic location.
7. Generate a table consisting of 60 atmospheric transmittance spectra at the AVIRIS solar and observational geometry and with 60 column water vapour values. This table also include the total amounts of column water vapour used in the calculations, and the 3-channel ratios calculated from the window and absorption channels in and around the 0.94 and 1.14um water vapour bands
8. Process the input cube one pixel at a time to derive the spectral reflectance spectra and to calculate the column water vapour amount. The derived surface reflectance values are written to an output image file with the same dimensions as the input image, and the column water vapour amounts are written to a separate file as a single channel image.

Adapting atrem to other sensors

ATREM only works with AVIRIS data (Gao et al., 1996), and its calculations are based upon this assumption. This lays the ground work for adapting ATREM for other sensors.

A few changes to the ATREM code will allow it to be used with a wide range of sensors. The modification is only a proposal, and the feasibility still has yet to be determined.

Instead of using the supplied solar irradiance and other tables, MODTRAN code be executed with the proper input parameters to provide these values more accurately for the supplied sensor and atmospheric conditions at time of flight.

In the REFLDRV subroutine, the constant to scale the ratio of the observed spectrum against the solar irradiance curve could be written as an input parameter and supplied by the user. As sensors are used, they could be added to the code to change scaling constant for that particular sensor.

Another change would be to allow a different sensor platform height of different sensor platforms at time of flight. The location of this information is currently unknown in the code.

To help reduce the amount of information that needs to be gathered at time of flight for atmospheric correction, the aerosol optical depth can be derived from the data itself, just like with water vapour can be obtained from the data. The aerosol optical depth can be derived from at least 2 spectral measurements in the blue and red portions of the spectrum using the presence of dense dark vegetation (Zagolski et al., 1994).

Recently, the Space Based Surveillance Group acquired data from the HYMAP sensor, from Integrated Spectronics. The task is to process the image and create a method of finding the targets. With a modified version of ATREM, this task could be done such that true spectra are compared with spectral libraries of known targets, as well as subtracting background earth radiance from the image, giving a clear location of targets.

CONCLUSION

Many computer codes have been written to model the Earth’s atmosphere. Theses codes can be used to remove the affects of the atmosphere from images. ATREM is one such computer program to calculate the effects of the atmosphere and then remove them from the image. ATREM can be modified to allow for a more accurate ground reflectance calculations while allowing it be a general purpose atmospheric correction tool for any sensor.

ACKNOWLEDGEMENTS

Many thanks to Kathy Heidebrecht for the help in getting ATREM compiled and running correctly and verifying the outputted data cube was what we suppose to be getting.

REFERENCES

Barducci A., Pippi I. (1995), “Retrieval of Atmospheric Parameters from the Hyperspectral Image Data”, IEEE, pp 138-140.

Berk A., Bernstein L.S., Robertson D.C. (1989), “MODTRAN: A Moderate Resolution Model for LOWTRAN 7”, Spectral Sciences, Inc. Burlington, Massachusetts.

Borel C.C., Clodius W.B., Johnson J., “Water Vapour Retrieval over Many Surface Types”, SPIE, vol 2758, pp 218-228.

Bornhoeft K.W., Lucey P.G., Horton K.A., “Atmospheric corrections for mid-IR (3-5 micrometer) spectroscopy”, SPIE, vol 2818, pp 118-127.

Carrere V., Conel J.E. (1993), “Recovery of Atmospheric Water Vapour Total Column Abundance from Imaging Spectrometer Data around 940 nm – Sensitivity Analysis and Application to Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) Data”, Remote Sens Environ, vol 44, pp 179-204.

Clark, R.N. (1997), “Spectroscopy of Rocks and Minerals, and Principles of Spectroscopy”, Manual of Remote Sensing, USGS, http://speclab.cr.usgs.gov/ PAPERS.refl-mrs/refl4.html.

Cornette W.M., Acharya P., Anderson G.P. (1994), “Using the MOSART Code for Atmospheric Correction”, IGARSS.

Gao B., Goetz A.F.H. (1990), “Column Atmospheric Water Vapour and Vegetation Liquid Water Retrievals From Airborne Imaging Spectrometer Data”, Journal of Geophysical Research, vol 95, no D4, 20 March, pp 3549-3564.

Gao B., Heidebrecht K., and Goetz A.F.H. (1993), “Derivation of Scaled Surface Reflectances from AVIRIS Data”, Remote Sens. Environ., no 44, pp 165-178.

Gao B., Heidebrecht K.B., Goetz A.F.H. (1996), “Atmospheric Removal Program (ATREM) version 2.0 User’s Guide”, Center for the Study of Earth from Space/CIRES, University of Colorado.

Godsalve C. (1996), “Simulation of ATSR-2 Optical Data and Estimates of Land Surface Reflectance Using Simple Atmospheric Corrections”, IEEE Transactions on Geoscience and Remote Sensing, vol 34, no 5, September, pp 1204-1212.

Hoffbeck J.P., Landgrebe D.A. (1994), “Effect of Radiance-To-Reflectance Transformation and Atmosphere Removal on Maximum Likelihood Classification Accuracy of High-Dimensional Remote Sensing Data”, IGARSS.

Roberts D.A., et al. (1994), “Temporal and Spatial Relationships between Topography, Atmospheric Water Vapour, Liquid Water and Vegetation Endmember Fractions Determined Using AVIRIS”, IEEE, pp 2366-2368.

Schmid B., et al. (1996), “Comparison of modeled and empirical approaches for retrieving columnar water vapour from solar transmittance measurements in the 0.94-um region”, Journal of Geophysical Research, vol 101, no D5, 27 April, pp 9345-9358.

Shanks J.G., Lynch D.K. (1994), “Remote Sensing through Cirrus Clouds: Visual and Sub-Visual”, IGARSS.

Singh S.M. (1994), “Parametrization of a Single Scattering Atmospheric Correction Algorithm using 5S Code”, International Journal of Remote Sensing, vol 15, no 1, pp 191-196.

Tanre D., et al. (1990), “Description of a Computer Code to Simulate the Satellite Signal in the Solar Spectrum: 5S code”, Int. Journal of Remote Sensing, vol 11, no 4, 1990, pp 659-668.

Teillet P.M. (1989), “Surface Reflectance retrieval using atmospheric correction algorithms”, Proceedings of IGARSS ’89 and the 12th Canadian Symposium on Remote Sensing, Vancouver, Canada, pp 864-867.

Thome K.J. (1994), “Proposed Atmospheric Correction for the Solar-Reflective Bands of the Advanced Spaceborne Thermal Emission and Reflection Radiometer”, IGARSS.

Vane G., et al., “The Airborne Visible/Infrared Imaging Spectrometer (AVIRIS)”, Remote Sensing of the Environment, vol 44, 1993, pp 127-143.

Wang J., et al. (1996), “Validation of FASCOD3 and MODTRAN3: Comparison of model Calculations with Ground-Based and Airborne Interferometer Observations under Clear-Sky Conditions”, Applied Optics, vol 35, no 30, 20 October, pp 6028-6040.

Zagolski F., Gastellu-Etchegorry J.P. (1994), “Atmospheric Correction of AVIRIS Images with a Procedure Based on the Inversion of 5S Model,” IGARSS.