Atmospheric correction of the Thorney Island 24 July 2001 data set using ACORN
The program to be used to correct the CASI image for the influence of the atmosphere is called ACORN, and is available commercially from Imspec LLC. This program offers a wide range of options for different types of airborne sensors, as well as those for satellite sensors, but it does not include options to correct for the effects of terrain and brdf, unlike ATCOR4. This is not a problem for the Thorney Island data set as the area is very flat and the swath width of the CASI sensor is relatively narrow. The front page of ACORN allows the user to select from the range of routines provided (Figure 1). In our case, we want Mode 5, 'Atmospheric correction for calibrated multispectral data'. This will provide a basic correction, and if this is successful we can enhance the correction by running ACORN in Mode 6, 'Single spectrum enhancement for atmospherically corrected multispectral data'. It might even be possible to use ACORN Mode 5.5, 'Atmospheric correction for calibrated multispectral data with separate water vapour image', as we have CASI bands located in the well and on the shoulder of the water absorption feature around 940 nm.
Figure 1. The main ACORN menu (version 5.0a)
The original image data provided by the NERC ARSF are coded as unsigned integers. ACORN requires the image data to be band-interleaved by line (BIL) or band-interleaved by pixel (BIP), and coded as integers. If you have access to IDL or ENVI, the conversion to integer is easily done using a program called CIN freely available from Center for Spatial Technologies and Remote Sensing (CSTARS) and the reformatting to BIL can be done in most image processing packages.
Figure 2. Entering parameters into the ACORN control file.
If you have read through the previous examples in this series, most of the parameters in the ACORN control file should be familiar to you. There are the usual entries related to the date, time of acquisition and geographical location of the site. As we are now dealing with airborne data there are also entries for the acquisition altitude and the average surface elevation. The user can specify one of three different atmospheres, and the visibility and water vapour amounts can be entered. The total amount of water vapour in the atmosphere was measured at the time of the flight using a sunphotometer on the ground, so we know that value is 10.0 mm. The horizontal visibility at the time of the flight was estimated as 30km.
ACORN requires three additional files that we have not met before. These describe the shape of the spectral response function of each of the CASI spectral bands and the radiometric calibration of the CASI sensor.
CASI spectral response function
Although the nominal wavelengths of the 'default vegetation bandset' are known
(Table 1), we do not know the precise shape of the bands. In the absence of
better information on this, Dr Liz Rollin used the data values in the HDF file
to create gaussian responses for each of the CASI bands. This file was edited
to be compatible with ACORN and may be downloaded from here.
CASI gain and offset values
These two files contain the scaling necesssary to convert from the radiance
units measured by the CASI to those required by ACORN. In the case of the NERC
ARSF CASI sensor, the data in the HDF file are in units of microwatts per square
centimetre per steradian per nanometre, and ACORN requires data in Watts per
square metre per nanometre per steradian, so each CASI DN value must be divided
by 100. For the NERC ARSF CASI, all the values in the sensor offset file are
zero. The gain and offset files for use with data from the NERC ARSF CASI sensor
in ACORN may be downloaded from here.
Results
The accuracy of the atmospheric correction can be assessed by comparing the reflectance of the asphalt and concrete surfaces measured in the field with those from the corrected image. Figure 4 shows the results for the uniform asphalt surface at the intersection of the two main runways. Figure 5 shows the results for an area of weathered concrete at the southern end of the main runway and Figure 6 shows spectra from a grass area immediately adjacent to the main runway.
Figure 4. Comparison between the reflectance of an asphalt ground calibration target measured in the field shortly after the flight and the same area on the atmospherically corrected image.
Figure 5. Comparison between the reflectance of an concrete ground calibration target measured in the field shortly after the flight and the same area on the atmospherically corrected image.
Figure 5. Comparison between the reflectance of a grass ground calibration target measured in the field shortly after the flight and the same area on the atmospherically corrected image.
This work is licenced under a Creative Commons Licence.
©
NCAVEO, 2005
Network for Calibration and Validation of Earth Observation data
School of Geography, University of Southampton
Southampton SO17 1BJ, UK
