Example 4. Atmospheric correction of a multispectral image collected by an airborne sensor

Introduction

The data to be used for this example were acquired over the Thorney Island test site on 24th July 2001 by an Itres Instruments Compact Airborne Spectrographic Imager (CASI) operated by the NERC Airborne Research and Survey Facility. Two single flight lines directly along the main runway were acquired at different heights, while a team of six people on the ground made simultaneous measurements of the properties of the atmosphere and the reflectance properties of pre-selected targets. The availability of simultaneous ground data means that it is possible to refine the results from a radiative transfer model to account for the conditions at the time of sensing, and also provides the data needed for simpler empirically-based methods of atmospheric correction.

The data set used in this example is available for use by others interested in testing different approaches or simply learning about the effect of the atmosphere on remotely sensed data. If you would like a copy of the CASI data set, please contact the NCAVEO office at the address below.

The data set

Figure 1 shows simulated true colour composite images of the data from the two flightlines. The higher flight line was acquired first, at approximately 08:47 GMT, and the lower flight line was acquired about ten minutes later. Both were acquired by the aircraft flying north to south following the line of the main runway.

   

Figure 1. The CASI data from Thorney Island, 24 July 2001. The left-hand image was acquired at 3109 metres above sea level and has a nominal pixel size of 4 metres. The right-hand image was acquired at 853 metres above sea level and has a nominal pixel size of 2 metres. © NERC, 2001.

The airborne data were acquired by a CASI imaging spectrometer configured to acquire data in the NERC ARSF 'default vegetation bandset'. HDF Explorer was used to reveal the actual bands sensed by opening the HDF file supplied with the data. The relevant sensor parameters are contained in the variables ?? and ??? within the folder labelled 'CASI' (Figure 2).

 

 

Figure 2. Using HDF Explorer to reveal the mission and sensor parameters from the HDF file.

The CASI was radiometrically calibrated prior to the flight using the methods described by Riedmann (2003).

CASI band	Central wavelength (nm)	Bandwidth (FWHM)(nm)	Comment
1	450.18
2	490.29
3	552.32
4	670.06
5	700.63
6	710.19
7	740.83
8	750.41
9	762.86		Oxygen absorption band
10	781.08
11	820.41		Water absorption band
12	865.49
13	940.2		Water absorption band

Table 1. The default vegetation bandset of the NERC ARSF CASI. FWHM stands for 'Full width, half-maximum', in other words, the width of each band at the points where its relative sensitivity is 50% of its maximum value.

In the case of NERC ARSF data, the HDF file also contains the flying height and the start and end time of each run (GMT), as well as a wealth of other information on the mission, the sensor and its radiometric calibration. Additional information is available on the flight log, a copy of which is available here.

The ground data available from the experiment are listed in Table 2 and may be downloaded from the NCAVEO Resources section of this website.

Measurand	Sensor	Wavelength range	Sampling interval
Total spectral irradiance	ASD FieldSpec Pro spectroradiometer with cosine corrector receptor	300 - 2,400 nm	60 seconds
Total broadband irradiance	Delta-T ES2 pyranometer	400 - 1100 nm	60 seconds
Total and diffuse quantum flux	Delta-T BF2 sunshine sensor	PAR (400 - 700 nm)	60 seconds
Direct solar irradiance	Microtops II sunphotometer	Narrow bands at 440, 675, 870, 936 and 1020 nm	Intermittent
Aerosol optical thickness	Microtops II sunphotometer	Narrow bands at 440, 675, 870, 936 and 1020 nm	Intermittent
Water vapour amount	Microtops II sunphotometer	Narrow bands at 440, 675, 870, 936 and 1020 nm	Intermittent
Asphalt radiance	GER1500 spectroradiometer	300 - 1100 nm	60 seconds
Air pressure	Delta-T barometer		60 seconds
Air temperature	Delta-T thermometer		60 seconds
Relative humidity	Delta-T hygrometer		60 seconds

Table 2. Ground data available to support the CASI data set collected from Thorney Island on 24 July 2001.

Aircraft sensor data present a number of challenges not found with satellite sensor data. First, the position of the sensor within the atmosphere means that we cannot assume that the signal will have been affected by the full thickness of the atmosphere. Second, aircraft sensors generally have a larger field-of-view, so the atmospheric path length from pixels near the edges of the swath is much greater than that from those near nadir. Third, the orientation of the flight line relative to the solar principal plane is not fixed like most satellite sensors, so view and illumination geometry must be taken into account. Finally, aircraft platforms are less stable so the imaging geometry is affected by turbulence and the three-dimensional motion of the platform to a much greater extent.

For these reasons, the challenges involved in accurate atmospheric correction of airborne sensor data are considerable. On the other hand, it is possible to acquire data from airborne sensors at low altitude, such that the influence of the atmosphere on the reflected signal is minimal. This is seen in the data from the low altitude flight from Thorney Island. Figure 3 compares the spectral radiance of an asphalt surface measured on the ground during the flight and data from the same surface imaged near-nadir in the CASI image collected from 853 metres altitude. The difference between the two is less than 10% for most wavelengths, and less than 1% for CASI bands 4, 5, and 6 which cover the chlorophyll absorption region and the start of the red edge. The excellent correspondence between these two data sets is not meant to imply that atmospheric correction is not necessary for airborne sensor data acquired from such altitude, but for some applications, where the sites of interest are close to nadir, it may be less critical than other aspects of pre-processing.

Figure 3. Comparison between the spectral radiance of an asphalt ground calibration target (continuous line) and CASI data collected from an altitude of 853 metres above sea level (dots).

Unlike the CASI data collected from 853 metres, those acquired from 3,109 metres above sea level show a much stronger atmospheric effect (Figure 4). The radiance in CASI bands 4, 5, and 6 is still remarkably close to that measured at ground level, but the values in the shorter wavelength bands are in error by up to 30%, so some form of atmospheric correction is necessary. Two approaches will be tested, the first based on a radiative transfer model, and the second a simple empirical approach.

Figure 4. Comparison between the spectral radiance of an asphalt ground calibration target (continuous line) and CASI data collected from an altitude of 3,109 metres above sea level (dots).

 

Next : Atmospheric correction using a radiative transfer model

Next : Atmospheric correction using an empirical method

---

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