MALFUNCTIONS
SEVERITY
DETECTIBILITY
TESTS
 
(1 to 10, good to bad)
(1 to 10, easy to hard)
  
 
General Faults:
Duplicate profile (already in the database)
5  - depending on the number of data in the box (could bias means if there are fewer casts)
1 to 2
 duplicateIO program which also adds new data files to the database
PER (Position error) (Land)
 10
2
 program 2 (to be defined)
PER (Position error) (speed)
1 to 10
1 if have cruise information, 10 if you don't
 program 2
TER (Time error) (primarily wrong year)
6
5
 duplicateIO detects and flags this for hand-checking.
OPA (MBT to XBT)
3
2
 program 2 - change if MBT goes  deeper than 300m - depth correct!
OPA (XBT to MBT)
3
8
 have to ignore this unless it is spotted in the hand checking.
CSA (XBT only...)
2 to 4
1
 program 2 - check t(0)
 
Leakage Faults:      
PFA (Pet Fault) 
(XBT only)
2 to 10 always warm
start of fault - 10
end of severe fault - 1
Easier with system info.  If none, see if dtdz>0 at bottom of the cast over most of the cruise in an area witout inversions.
 t(z)
z(t)
dtdz<0 (t inversion)

unrealistic (anomalous)  dtdz
LER (Leakage - unspecified causes)
2 to 10 - always warm
start of failure - 5
end if severe - 1
 t(z)
z(t)
dtdz<0 (t inversion in area without temperature inversions)

t(z) warmer than neighbours
 IPA/R (insulation penetration, usually with recovery)
4 to 10 
8 
 dtdz extreme & >0
followed by a negative t gradient with a war bias until (if) the cast recovers (t(before spike)=t(z),  If dz <=10m(?) then interpolate, otherwise reject sact from point of initial extreme gradient.
OTHER Erroneous Inversons:      
WSR (wire stretch or other unreasonable inversion)
5+ always warm - warmer than neighbours and area buddies
2
 dtdz<0 and t(z) too high - reject at the start of dtdz<0
BOR (Bowing in the mixed layer - system specific)
2 to 3
6
need system information but can use drop to drop consistency (test failures in areas without known nubs...)if system not available.  Primarily French ships.
 t(mld2) - t(0) >0 (gradient mld)
dtdz<=0 within depth of mld2
CUR (cusping or any form of sinusoidal leakage)
1 to 4
9
 fast fourier transform analysis and check the power spectrum?  some other form of spectral analysis?
       
Gross Gradient problems:      
Spikes / High frequency noise / NGR (no good)
10
1
 dtdz extreme
t(z) out of range but data below is OK.

if only 1 or 2 points, interpolate across spikes.  If more extreme, reject entire cast below first spike.

If many ,small spikes, (HF) reject even if fliterable?  for simplicity?
Wire Break / NTR (no trace)
10
1
 t(z) our of range without recovery and dtdz extreme
Other faults:
DR (Driver error - system specific)
7
10
 cold bias but probably within the sd.  Try to plot mld2 (1?) vs average temp anomal in top 100m? 150m? mld+50m?
CT/HB (constant temperature)
10
3-4
 t(z)>mean
dtdz>0
bathymetry file?
build our own bathymetry form the archive?
If both t(z)>mean and dtdz>=0 then check region neighbours for HB flags and hand check?  remember could be CT and not HB so bottom topography is not the decider in this case.
SAR (surface anomaly - unusual)
2+
3
 unreasonable t(0) (sst).  Easily dealt with as long as it is limited to the surface - chop deeper.  May need to be checked by hand.
       
Faults to be ignored:
Modulo 10 spikes
(MO)
system specific...
4 -cold bias
5
 ignored because require full resolution data with  system type information and this is unlikely to be available.  If it occurs, they will be so rare it isn't worth spending time on detecting this fault unless we come across a specific data set where it is common. And in that case, it will probably be more efficient to just correct them within quest before loading into the database.
Sticking Bit (SB)
system specific...
3 - steppy trace
10
 without system information, this malfunction is indetectable.