Biophysical parameter mapping from optical remote sensing images always require an intermediate modeling step to transform spectral observations into useful estimates. This modeling step can be approached with either statistical, physical or hybrid methods. Here emphasis is put on physical methods. Physical methods refer to inversion of radiative transfer models.
The here presented LUT-based Inversion toolbox provides a suite of lookup-table (LUT)-based radiative transfer model (RTM) inversion routines to enable semiautomatic mapping of surface biophysical variables.
LUT-based inversion is considered as a physically-sound method to retrieve biophysical parameters from remote sensing data, but regularization strategies are required to mitigate the drawback of ill-posedness.
LUT-based inversion in its essential form relies on direct comparison of LUT spectra against a remote sensing observation through a cost function. Various regularization strategies have been proposed to optimize the inversion routine. For instance: (1) the use of prior knowledge about model parameters; (2) the use of multiple best solutions in the inversion (instead of the single best solution); (3) adding noise to account for uncertainties attached to measurements and models.
Conceptual design of the Spectral Indices toolbox to map biophysical variables.
The LUT-based Inversion toolbox makes use of in ARTMO earlier prepared lookup-tables (i.e. simulated spectral databases) and field data to optimize and validate the inversion strategies. The RTM input variables can be retrieved from a remote sensing image through the selected inversion routine on a pixel-by-pixel basis.
ARTMO’s LUT-based Inversion toolbox v.1.15.
In short, the LUT-based Inversion toolbox enables:
- To apply and evaluate multiple LUT-based inversion strategies, e.g. according to different cost functions, and regularization options such as added noise and mean of multiple solutions in the inversion.
- If a land cover map is provided, then for each land cover class a distinct inversion strategy can be optimized.
- When having validation data available, then multiple inversion strategies can be analyzed against the validation dataset by using goodness-of-fit statistical measures. Results are stored in a relational database.
- The best validated strategy can be applied to a remote sensing image, or an inversion strategy can be directly applied to an image without validation, for biophysical parameter mapping.