Plots

We provide several plots for the validation of retrieval quality.

Reflectance

Modelled and measured reflectance curves show if the modelled curve is close enough to the measured curve.

../_images/modelled_vs_measured.png

As there might be many spectra we do not show all the figures immediately after the run finishes but we save them all in the matlab graphical array named figures with plot.reflectance_hidden() To actually see them set visibility property of the j-th figure to ‘on’:

set(figures(1), 'Visible', 'on')    % show the plot for the 1st spectrum
set(figures(5:7), 'Visible', 'on')  % show the plots for the 5, 6, 7 spectra
set(figures, 'Visible', 'on')       % show the plots for all spectra

If you close the figure (by pressing the upper right red cross) it will also disappear from figures array; in fact there will be “handle to the deleted figure”.

Note

You can draw all these plots again from the “%Y-%m-%d_%H%M%S.xlsx” file with plot.replot_all()

Validation (1 : 1)

To see how well the retrieval worked for your spectra you can provide validation on Filenames tab.

Validation file requirements:

  • first columns - parameter names as they are on Input tab

  • other columns - measured values of this parameter for each spectra.

Overall the number of columns in validation file equals to the number of columns in the reflectance file +1 for parameter names.

Example:

names

spec_1

spec_2

spec_3

LAI

1.0

NA

1.5

Cab

40

30

50
../_images/quality_of_synthetic.png

Jacobian

The algorithm of local minimum search lsqnonlin() uses the gradient descent method where the next step is driven by the cost-function partial derivative matrix - Jacobian matrix.

For each spectra we calculate the Jacobian matrix at the last step of optimization to propagate standard deviation of reflectance to that of the retrieved parameters.

Jacobian can be plot with plot.jacobian(J_all, tab, measured.wl, j)() and its SVD (singular value decomposition) can be plot with plot.jacobian_svd(J_all, tab, measured.wl, j)() for j-th spectra.

Note

We do not write the Jacobian matrix to file. If you want to analyze it plot it immediately after the model run or save as the .mat file yourself.

../_images/Jac.png