Simulation of Sentinel-2 MSI Multispectral Images Using EO-1 Hyperion Hyperspectral Data.
Louis, J.M.B.
Telespazio, FRANCE

Within the framework of the on-going ESA project Sentinel-2 L2A Prototype Processor (S2L2APP), simulated dataset of Sentinel-2 products are needed to support the performance assessment of the S2L2APP. This poster describes the method used to simulate Sentinel-2 MSI (Multi-Spectral Instrument) images from hyperspectral data of Hyperion instrument flown onboard NASA EO-1 mission.

The MSI instrument onboard Sentinel-2 has three different spatial resolutions (10 m to 60 m) and 13 spectral bands. The three 10 m bands in the visible region enable true colour images with a high spatial resolution. The 10 m wide NIR band also allows a 10 m false colour infrared (CIR) composite. The spectral bands needed to retrieve atmospheric parameters are designed with a coarser resolution of 60 m (channels at 443 nm, 940 nm, 1375 nm). The remaining channels have a spatial resolution of 20 m.

In order to simulate the 13 spectral bands of the MSI instrument, free-of-charge hyperspectral data from the NASA EO-1 mission has been chosen. EO-1 satellite was launched on November 21, 2000 as part of a one-year technology validation/demonstration mission. The Hyperion instrument provides a high resolution hyperspectral imager capable of resolving 220 spectral bands (from 0.4 to 2.5 μm with an average of 10 nm Gaussian FWHM) with a 30 m resolution. The instrument can image a 7.5 km by 100 km land area per image.

To generate Sentinel-2 simulated dataset, the method described here includes the following steps:

1- Hyperion pre-processing

This optional step is proposed to limit the stripping effect on Hyperion hyperspectral data. The processing uses the principle of moment matching and is inspired by the work of Xie et al. (1).

2- Aggregation of Hyperion radiances into MSI radiances

The aggregation methodology is based on the study conducted by Barry et al. (2). The idea is to combine Hyperioni s narrow hyperspectral bands so that they synthesize the Sentinel-2 MSI broader multispectral bands. The general process is to calculate a weighted sum of the Hyperion bands that cover each MSI band. The weights used in the sum are derived by comparing the spectral response of the hyperspectral bands with the multispectral band. Mathematically, the process involves convolving the Hyperion Gaussian spectral response with the Sentinel-2 MSI broad band spectral response and normalizing carefully to ensure consistent units between instruments.

3- Conversion from MSI Radiance to MSI Reflectance

The conversion from MSI radiance to MSI reflectance uses the Thuillier standard solar spectrum, a composite spectrum published by Thuiller et al. (3), to compute the amount of solar radiation for each Sentinel-2 spectral band.

4- Export and format of MSI reflectance into Sentinel-2 tiles

This last step is necessary to evaluate the computational performance of the S2L2APP. It consists in exporting and formatting the simulated MSI reflectance data into Sentinel-2 tiles. The output data is composed by 13 TIF or JPEG2000 tiles of 100 x 100 km in different GSD (Ground Sampling Distance) of 10 m, 20 m and 60 m depending on the native resolution of the different spectral bands. The resampling from 30 m to 20 m and 10 m bands is done using cubic spline interpolation whereas the resampling from 30 m to 60 m is done by linear interpolation.

In conclusion, we proposed a method to generate simulated Sentinel-2 dataset that allows testing the performance of S2L2APP or other Sentinel-2 algorithms on a large variety of acquisition conditions, thanks to the large amount of Hyperion archived data, acquired regularly since 2001 and available free-of-charge from the © USGS Earth Explorer portal.

(1) Y.S. Xie, J.N. Wang, K. Shang (2011): i°An improved approach based on Moment Matching to Destriping for Hyperion datai±. Procedia Environmental Sciences, Vol. 10, 319-324.

(2) P.S. Barry, J. Mendenhall, P. Jarecke, M. Folkman, J. Pearlman, B. Markham (2002): i°EO-1 Hyperion Hyperspectral Aggregation and Comparison With EO-1 Advanced Land Imager and Landsat 7 ETM+i±. :©:γ:γ:γ 2002 International Geoscience and Remote Sensing Symposium, Vol. III, 1648-1651.

(3) G. Thuillier, M. Hers, D. Labs, T. Foujols, W. Peetermans, D.Gillotay, P.C. Simon, H. Mandel (2003): i°The Solar Spectral Irradiance from 200 to 2400 nm as Measured by the SOLSPEC Spectrometer from the Atlas and Eureca Missionsi. Solar Physics, 214(1), 1-22.