support RasterTimeseriesManager format in SpectralView

Benjamin Jakimow

I turned out that, agains the documentation https://numpy.org/doc/stable/reference/generated/numpy.fromstring.html

numpy.fromstring('1,2,3,4', count=10, sep=',')

does not raise a ValueError if there are less elements than told with the count keyword. I’ll change this so that the x values of pixel profiles needs to be defined in an array of length = number of bands. In case of your dataset this will leads to “stacked” y values for all bands with the same observation date:

‌

The metadata in the testdata.zip shows some inconsistencies, e.g to the ENVI standard described [here](https://www.l3harrisgeospatial.com/docs/ENVIHeaderFiles.html):

“wavelength” has as many entries as bands in the dataset (in this case 60). This is what I expect in an ENVI header. Otherwise, “center_wavelength” contains 6 bands only. Why, in the first case, are redundant entries used (each decimal date is repeated 6 times), but not in the second?
“center wavelength” and “center wavelength units” are not part of the ENVI standard. Are they evaluated by ENVI at all? If not, why to write them into the ENVI domain at all?

If the aim is have a nested band order that combines wavelenght and temporal order, than I think it woul be better to stick to the ENVI standards as close as possible, e.g.

use “wavelength” for wavelenght (e.g. band nanometers), not decimal years, and repeat each wavelength value all 6 bands (as it is the truth)
use a none-ENVI standard tag to describe the temporal dimension, e.g. “decimalyears”
allow more precission, e.g. “datetime” or “acquisitiondatetime”(https://gdal.org/user/raster_data_model.html?highlight=acquisition date) instead of “date” only, as this allows to have multiple observations on the same day.

‌