IMPLEMENTATION OF HYPERION SENSOR ROUTINE IN 6SV RADIATIVE TRANSFER CODE Emanuele Mandanici DISTART University of Bologna, V. Risorgimento 2, 40136 Bologna (Italy), Email:
[email protected] ABSTRACT The Second Simulation of a Satellite Signal in the Solar Spectrum Vector code (6SV) is one of the most widely used Radiation Transfer (RT) codes, but its employment in the hyperspectral image pre-processing chain has been limited. Since version 1.1 (released in May 2007) some gaps in respect of Modtran accuracy were bridged, by taking into account the effects of polarization, thus additional validation work is required. This paper presents an implementation of the Hyperion sensor support in the original code and some tests on the use of modified model for the atmospheric correction of Hyperion archive scenes. First results of direct transfer simulation and model inversion to retrieve soil reflectance values are discussed in comparison with those obtained by Modtran-based algorithms. 1.
such as ASTER (Advanced Advanced Spaceborne Thermal Emission and Reflection Radiometer) and Landsat.
BACKGROUND
Atmospheric correction is a critical issue in most remote sensing application, especially when data acquired by different sensors at different time are compared for a multitemporal analysis or a change detection. In such situation a careful radiometric calibration is required to retrieve comparable information, but ground truth spectra to perform the empirical calibration are frequently unavailable, especially for dated archive images. In this case the use of a radiative transfer (RT) code is recommended. Figure 1 shows an example of computation of Normalized Difference Vegetation Index (NDVI) by two image of the same area, acquired in the same day. Histograms demonstrates that radiometric correction is required to reduce artificial differences due to sensor characteristics and atmospheric conditions. The Second Simulation of a Satellite Signal in the Solar Spectrum Vector code (6SV), version 1.1, was released in May 2007 [1, 2]. This code is based on the method of successive orders of scattering approximations and it includes the calculation of four components of the Stokes vector, in order to compute the effect of polarization. Furthermore this release provides the possibility to vary arbitrarily the vertical aerosol profile and a more accurate calculation of aerosol scattering phase functions [1]. The 6S code is used in MODIS (Moderate Resolution Imaging Spectroradiometer) atmospheric correction algorithm, but supports also other multispectral sensors,
_____________________________________________________ Proc. ‘Hyperspectral 2010 Workshop’, Frascati, Italy, 17–19 March 2010 (ESA SP-683, May 2010)
Figure 1 - NDVI histograms over a portion of the Fayyum oasis (Egypt). A Landsat 7 image and an ASTER one taken in the same day (30 April 2002) are compared. NDVI value has been computed from TOA, FLAASH retrieved and 6SV retrieved reflectance. 6SV model seems to provide the best correction.
The present work aims to test the capabilities of the 6SV model to perform atmospheric correction on the hyperspectral data provided by Hyperion sensor. This instrument, on board of the Earth Observing 1 platform,
can image a 7.5 km by 100 km land area per frame, with a ground resolution of about 30 m [3]. Hyperion is a push-broom sensor which provides 242 band images, continuously covering the visible and near infrared portion of the spectrum, between 0.4 and 2.5 µm, with a spectral resolution of approximately 10 nm. 2.
HYPERION IMPLEMENTATION
The implementation of a new sensor in the RT code firstly requires the definition of a relative spectral response function, which describes the sensitivity of the detector. This wavelength dependent function must be determined by laboratory calibration techniques, using well characterized radiation sources. Calibration data for Hyperion sensor are provided by Australian Commonwealth Scientific and Industrial Research Organisation (CSIRO) [4].
Figure 3 - Relative SRF for the first five bands of Hyperion sensor, modeled by eq. (1) with pixel averaged central wavelength and FHWM coefficients.
3.
GRAPHICAL INTERFACE
A Windows based graphical user interface, called Visual SixS, has been developing by Visual Basic 9. The interface (see Figure 4) is primarily designed to manage the atmospheric correction of a given at sensor radiance image. It allows a user friendly parameterization of the model and it returns the surface reflectance image.
Figure 2 - Laboratory measured SRF and interpolated Gaussian curve [6]. Courtesy of CSIRO.
The Relative Spectral Response Function (RSRF) for each band can be well modelled by a Gaussian shape [6], as given by equation (1): 2 x CW RSRF exp 2 2ln 2 (1) FWHM 2
where CW is the centre wavelength of the band and FWHM is the full width at half maximum [7], which is related to the standard deviation. These parameters have been interpolated from laboratory measurement (see Figure 2) and averaged among all detectors. Figure 3 shows the relative SRF for the first five bands of Hyperion sensor, computed by Eq. (1). The RSRF has been implemented in the code trough a specific subroutine.
Figure 4 - Image selection panel of the Windows interface for the RT model. Currently the application is at “alpha” development status.
On the other hand, the original capabilities of the 6SV model are still preserved, so it is possible to simply perform a direct simulation and compute the at sensor radiance, starting from a given surface reflectance. 4.
RESULTS AND DISCUSSION
Several Hyperion datasets have been used in order to carry out a first evaluation of the performance. For example, Figure 5 shows the results of a simulation, assuming a standard “Midlatitude Winter” atmospheric model and a “background desert” aerosol profile. The Hyperion at sensor radiance of a target characterized by a constant spectral signature by 30% has been computed.
correction are applied before running the models. The Hyperion data cube has been also reduced from the original 242 bands to 179 only, to avoid uncalibrated and overlapping bands [5]. Figure 6 shows the spectral profile of a pixel covering an arid bare soil in the Libyan Desert. The profiles obtained by FLAASH and 6SV, using the same atmospheric and aerosol models, are compared with the Top of Atmosphere (TOA) reflectance, given by eq. (2):
TOA
Figure 5 - Simulated Hyperion at sensor radiance for a constant signature target. Apparent reflectance and total gaseous transmittance are plotted in the upper part.
In order to evaluate the atmospheric correction performance, several images have been processed with 6SV and the outputs have been compared with the reflectance obtained by FLAASH (Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes), a Modtran based state-of-the-art model, and by QUAC (QUick Atmospheric Correction), an empirical image based algorithm. Radiometric calibration and de-smiling
d2 ESUN cos
L
(2)
where ESUN is the solar irradiance, provided by USGS, and θ the solar zenith angle. TOA reflectance does not consider atmospheric effects at all, so it is useful to evaluate the performance of the models, especially in removing gaseous absorption feature. From Figure 6, it is possible to see that there is no relevant improvements in applying the FLAASH water retrieval algorithm, in this hyper-arid environment. 6SV corrected profile shows a close agreement with FLAASH one in the near infrared portion of the spectrum and in main atmospheric windows at short waves, but seems to fail removing the water vapour absorption bands at 940 nm and 1100 nm.
Figure 6 - Hyperion retrieved spectral profiles for an arid soil. FLAASH correction was performed at 5 cm-1 resolution with (blue) and without (magenta) water retrieval option; 6SV (green) has a 10 cm-1 resolution. Black line represents the TOA reflectance.
Figure 7 - Spectral profiles over a water surface (Lake Qarun, Egypt).
Figure 7 shows a comparison among 6SV, FLAASH and QUAC retrieved reflectances on a lake water surface. In this case FLAASH and QUAC exhibit anomalous behaviour at water absorption band wavelength. This “outlier” values may compromise applications dealing with water quality assessment. 6SV instead seems to provide more reasonable values. 5.
CONCLUSIONS
First results confirm that 6SV represent a challenging option, when atmospheric correction of remotely sensed images is unavoidable, especially in the context of a multitemporal analysis. Further investigation are required in order to carry out a rigorous validation of the model and comparisons with ground truth data will be performed in the near future. Comparisons with Modtran based algorithm over arid and vegetated land cover types highlight deficiencies in the weaker water vapour absorption band modelling. On the other hand 6SV seems providing more accurate results over water surfaces. Implementation work on the Visual SixS interface is still in progress to support all options of the 6SV original code, especially those related with user defined atmospheric and aerosol profiles. REFERENCES 1. Kotchenova, S.V., Vermote, E.F., Matarrese, R. & Klemm, F.J. (2006). Validation of a vector version of the 6S radiative transfer code for atmospheric correction of satellite data. Part I: Path radiance. Applied Optics. 45(26), 6762-6774.
2. Kotchenova, S.V., Vermote, E.F., Matarrese, R. & Klemm, F.J. (2007). Validation of a vector version of the 6S radiative transfer code for atmospheric correction of satellite data. Part II: Homogeneous Lambertian and anisotropic surfaces. Applied Optics. 46(20), 4455-4464. 3. Ungar, S.G., Pearlman, J.S., Mendenhall, J.A. & Reuter, D. (2003). Overview of the Earth Observing One (EO-1) Mission. IEEE - Transactions on Geoscience and Remote Sensing, 41(6), 1149-1159. 4. Pearlman, J.S., Barry, P.S., Segal, C.C., Shepanski, J., Beiso, D. & Carman, S.L. (2003). Hyperion, a Space-Based Imaging Spectrometer. IEEE Transactions on Geoscience and Remote Sensing, 41(6), 1160-1173. 5. Jupp, D.L.B. and Datt, B. (Eds) (2004). Evaluation of Hyperion Performance at Australian Hyperspectral Calibration and Validation Sites. CSIRO Earth Observation Centre Report. 6. Liao, L.B., Jarecke, P.J., Gleichauf, D.A. & Hedman, T.R. (2000). Performance characterization of the Hyperion Imaging Spectrometer instrument. SPIE Proceedings, 4135, 264-275. 7. Liu, B., Zhang, L., Zhang, X., Zhang, B. & Tong, Q. (2009). Simulation of EO-1 Hyperion Data from ALI Multispectral Data Based on the Spectral Reconstruction Approach. Sensors, 9, 3090-3108.