search

VincyaneBadouard / TreeData_broken

Harmonization and correction forest data tool.
https://vincyanebadouard.github.io/TreeData/
0 stars 1 forks source link

StatusCorrection is not handling initial diameter error well #86

Open ValentineHerr opened 1 year ago

ValentineHerr commented 1 year ago

If, e.g. the first diameter measured is too big (e.g. typo 91 instead of 19) the function detects the drop but considered the next measurement as too small... I need to see if I translated Vincyane's function wrong or if it is something to work on from scratch.

ValentineHerr commented 1 year ago

Here is an example of output of the Diameter correction function. The measurement I highlighted in yellow is the wrong one, but the function corrects all other measurements.

image

I could try to find a way to anticipate the number of corrections any given error will generate and see if considering the error is coming from an adjacent measurement would generate less corrections. The problem with that is it will likely be extremely slow.

Other option, which I is a lot more feasible, would be to first look at outliers and remove them before detecting growth anomalies. Tricky part is how to define an outlier. That will also require the function to back fill diameters (we are currently only filling them forward). The good thing with that is that it would allow to include the recruitment function in there.

ValentineHerr commented 1 year ago

note that what will fix 100666 will also change 100669...

ValentineHerr commented 1 year ago

issue #22 (which I am closing now) mentioned:

Propose also the other reading direction: consider the last diameter value of a tree as the reference (TrustMeasSet = “last”) (not a good idea).