high deep value test line 465

[catsch]

if (np.nanmedian(BBPmf1[iDEEP]) > C_DEEP_BBP700_THRESH) & (len(BBPmf1[iDEEP]) >= C_N_of_ANOM_POINTS):

I think this part is not correct len(BBPmf1[iDEEP]) >= C_N_of_ANOM_POINTS, assuming that you are counting points deeper than the Pressure threshold and not anomalous points deeper than the Pressure threshold

[grgdll]

Thanks for spotting this bug: I (think I) have now fixed it.

[catsch]

iDEEP = np.where(PRES > C_DEPTH_THRESH)[0] # find deep part of the profile
if np.nanmedian(BBPmf1[iDEEP]) > C_DEEP_BBP700_THRESH: # this is the test 
    # find how many points fail the test
    iPointsBelow700dbar = len(iDEEP) # indices of points that fail test
    if iPointsBelow700dbar >= C_N_of_ANOM_POINTS: # check if we have enough points at depth 
        ISBAD = np.where(BBPmf1)[0] # flag entire profile

into (not sure about python syntax and how to deal with nan values)

iDEEP = np.where(PRES > C_DEPTH_THRESH)[0] # find deep part of the profile
iPointsBelow700dbar = np.where(BBPmf1[iDEEP]) > C_DEEP_BBP700_THRESH) # find how many deep points fail the test
if len(iPointsBelow700dbar) >= C_N_of_ANOM_POINTS: # check if we have enough points at depth
ISBAD = np.where(BBPmf1)[0] # flag entire profile

as BBPmf1 is already the median filtered value ?

[grgdll]

I think the issue is in how the test was described in v1 of the manuscript. In fact the test checks if the median value of the median-filteredf BBP values below C_DEPTH_THRESH is > C_DEEP_BBP700_THRESH. But we should also check that we have enough points below C_DEPTH_THRESH.

So I have update the revised manuscript and the flow chart. Please let me know if you see other problems.