Approximating Empirical Surface Reflectance Data through Emulation: Opportunities for Synthetic Scene Generation.

Area of Science:

• Optical Remote Sensing
• Biophysical Variable Analysis
• Statistical Learning in Geoscience

Background:

• Accurate spectroradiometric measurements are crucial for validating remote sensing vegetation products.
• Data gaps in spectroradiometric collections often necessitate additional fieldwork or advanced processing techniques.

Purpose of the Study:

• To introduce and evaluate an emulation technique for generating empirical-like surface reflectance data of vegetated surfaces.
• To compare the accuracy and speed of emulation against classical interpolation methods for spectral reconstruction.

Main Methods:

• Emulation using statistical learning to reconstruct spectral measurements.
• Validation against an empirical field dataset with CHRIS and HyMap hyperspectral data.
• Comparison with classical interpolation methods.

Main Results:

• Emulation demonstrated higher accuracy in producing surface reflectance data compared to interpolation.
• Emulation significantly outperformed interpolation in speed, generating spectra tens to hundreds of times faster.
• Emulators successfully simulated hyperspectral imagery (CHRIS-like, HyMap-like) and data cubes (S2 spatial texture, HyMap spectral resolution) rapidly.

Conclusions:

• Emulation is a superior method for reconstructing spectral measurements, offering improved accuracy and efficiency over interpolation.
• This technique facilitates rapid generation of large synthetic spectral datasets and accelerates computationally intensive processing, such as synthetic scene generation.
• Emulation enhances the potential for advanced data processing and analysis in optical remote sensing of vegetation.