search

BrandonSmithJ / MDN

Mixture Density Network for water constituent estimation
GNU General Public License v3.0
40 stars 34 forks source link

SOLID failed in extremely turbid waters #21

Open SidneyCCCC opened 2 weeks ago

SidneyCCCC commented 2 weeks ago

image image When I used SOLID on an extremely turbid waters (The Yangtze River Estuary), I found that it failed. Is it because there are no such turbid samples in the training set, or did you set a threshold?

BrandonSmithJ commented 2 weeks ago

No, there isn't any threshold like that applied. It's likely something to do with the Rrs being provided - can you share the Rrs values for one of the failed pixels?

SidneyCCCC commented 2 weeks ago

image This spectrum was generated with Acolite-DSF default settings. I don't know where it went wrong, but it turns out to be NAN in TSS.

SidneyCCCC commented 2 weeks ago

I would like to add that I have converted the result to Rrs before input it into SOLID.

SidneyCCCC commented 2 weeks ago

Sorry, I found that I selected the wrong model. SOLID is valid for this area. It is invalid in the model that is exported by tss, chl, and cdom at the same time. Why is this?

BrandonSmithJ commented 1 week ago

Can you show me what your code that generates the estimates looks like? In particular, the code that uses the image_estimates function? It would also help to see the actual Rrs values rather than the rhos values - it's quite possible acolite is generating invalid Rrs spectra, even if the rhos appears valid.