Open nfahlgren opened 3 years ago
In #426 I effectively did this filtering by 1. initializing the output array with 0s and 2. only computing Fv and YII, NPQ when it made sense to i.e. Fo < Fm and Fv > 0, Fm > 0
# To calculate the divisions properly we need to change from unit16 to float64 data types
out_flt = np.zeros_like(mask, dtype='float64')
# Calculate Fvariable, where Fv = Fmax - Fmin (masked)
fv = np.subtract(fmax_mask, fmin_mask, out=out_flt.copy(),
where=fmin_mask < fmax_mask)
# make sure to initialize with out=. using where= provides random values at False pixels. you will get a strange result.
# this was a problem when (because?) mask comes from Fm instead of Fm' so the plant pixels can be different
#mask<0, fmax>0 = FALSE: not part of plant but fluorescence detected.
#mask>0, fmax<=0 = FALSE: part of plant in Fm but no fluorescence detected(!!!, plant movement?)
yii = np.divide(fv, fmax_mask, out=out_flt.copy(),
where=np.logical_and(fv > 0, fmax_mask > 0))
overall it will be (and was for me) more confusing to get negative results in YII and NPQ! If you don't initialize with 0 you get a lot of pixels with values like 1e-16 which is also confusing and non sensical.
Awesome, thanks @dschneiderch!
Is your feature request related to a problem? Please describe. When we calculate a metric like
(Fm - F0) / Fm
, there can be some quality control issues. One that came up is that whenF0 = 0
thenFv/Fm = 1
. I think this is not a productive result both because it lacks biological relevance and because our lack of ability to detect signal in F0 could just be a detection limit issue.Describe the solution you'd like I propose that we only calculate Fv/Fm where F0 > 0 within the
plantcv.photosynthesis.analyze_fvfm
function.Describe alternatives you've considered Alternatively, we do not change anything and leave it to the user to filter the results. I worry a little about creating a non-transparent process (though we would document it) but I think this is a non-controversial filter to add?
cc: @annacasto, @dschneiderch
Additional context If we do add quality filtering, this could also apply to other photosynthesis functions being developed in #426.