Posez vos questions :
PANDORA
Notes regarding oxygen calibration, Pandora Cruise A. Ganachaud, 2013. Non-public.
Sensors
Calibration sheets: main primary: 1506 (pre calibration 5-may-2012; same with Tcor); main secondary: 0514 (pre calibration 28-oct-2011; post calibration 15-dec-2012; same with Tcor)
Primary and secondary sensors: " Généralement le circuit primaire est meilleur. Sur le secondaire il y a souvent du bruit en conductivité. Donc je prendrais systématiquement le primaire. cas particulier, le cast 6801, que Jacques a récupéré après inversion des circuits, à vérifier même si les capteurs sont maintenant dans le bon ordre. Et toujours les problèmes en surface pour 1301,1302, 3903,3904.
Station characteristics:
Parameters and remarks from cruise (J. Grelet)
Summary table, *** Primary circuit ***
| Station |
Gaps in the vertical | sigma-match > 50m (except MLD) | Upcast problem | Downcast problem | Spikes | Remarks |
|---|---|---|---|---|---|---|
| 1-8 | X | Unit deck setting problem | ||||
| 2 | X |
Data gap near 1705m? STEP at 1646m; maybe real |
No filtered DVDT: 1705-1740m; 1765-1785; 1795-1905; | |||
| 3 | X |
No filtered DVDT below 1885 bump at 1275m |
||||
| 4 | X | No filtered DVDT below 1680 | ||||
| 5 | X | No filtered DVDT below 1835 | ||||
| 6 | X | No filtered DVDT below 1810 | ||||
| 10 | 2500-2700m | No spike in calibrated o2 | ||||
| 12 | 735m; in Smax that was almost constant | No incidence on bottle | ||||
| 13 | SSS | (GE) | ||||
| 16 | spike-looking | cast 1: structure in both sensors | ||||
| 21 |
1865m, constant salt |
? proche de Guadalcanal No incidence on bottles 2102-2111: flagged 3 below 1000m (jellyfish problem) |
||||
| 25 | 100-700m |
S spike at 592m and some in 200-300m flag 3 salt 57-799m because bottles confirm values below flag 3 Temp all profiles because many glitches |
||||
| 28 | 1869m | 3155-bottom |
No incidence on bottles flagged to 3 below 3155m; also in S |
|||
| 31 | 1650, 1850 | |||||
| 3601 | 0m |
Left pipe: removed data ? nothing in salinity |
||||
| 3904 | 0-25m | Left pipe: removed data | ||||
| 40 | No sigma match 900m | Problem during upcast | ||||
| 42 | No sigma-match for 1000, 700, 500 |
800-970m; Almost all casts then pump was changed |
Problem during upcast; 400m is also off due to spurious sigma match Arrival in Solomon Strait Systematic for all casts (GE). Secondary better. |
|||
| 4201 |
834-1001m |
Removed data |
||||
| 4202 |
813-941m |
Removed data |
||||
| 4205 |
818-963m |
Removed data |
||||
| 4207 |
791-1210m |
Removed data |
||||
| 4209 |
798-1310m |
Removed data |
||||
| 48 | 1336m; Salinity const | |||||
| 53 | Samples are in the gradients | |||||
| 54 | 2650 in raw o2 | Nothing in calibrated o2 or salinity | ||||
| 6001 |
3543-bot |
All profile flagged to 3 (deviation from samples and next casts) Removed bad data below |
||||
| 6002 (5506) | at 2542, 2750, 3000, 4000, 4250, 4500, 5001, 5250m | 3525m |
Constant density so limited match to 3500m; Bottles 14,15,17 are higher Removed data below 3525m Used circuit 2 for T/S bottle (upcast) |
|||
| 6003-6008 |
Removed data below 3525m |
|||||
| 6009 | 9m |
Removed surface data (pipe); Sali |
||||
| 6013 (1300) | 715, 304, 191 | Used circuit 2 for T/S bottle (upcast) | ||||
| 61 |
1650 1750 |
dVdt problem 1705m; O2 Despiked @ 1699-1710m Goes below 3800m Despiked S: 1646-1653; 1726-1754; 1839-1842 |
||||
| 62 | at 70m | Mixed layer very homogeneous | ||||
| 63 | at 55m | Mixed layer very homogeneous | ||||
| 66 | X | Missing 234m and 235m | ||||
| 67 |
1250 1500 Also in Sali |
Looks noisy below 1500m Pump power supply problem fixed after 67 Unspiked: 1492-1539m; 1546-1565m; 1670-1684m; 1748-1769m; 1823-1832m; 1894-1960m; 1965-2000m Unspiked salinity also Not temperatures |
||||
| 6801 | Initially, CTD1/2 inversion and bad pump: re-done (6802). But CTD1/2 were put back in order in the dataset (email JG 19/11/21) | |||||
| 6802 | at 76m | |||||
| 72 | X | Missing 103m and 104m | ||||
| 82 | X | Many places | Unit deck setting problem. No filtered dvdt 1800-2000m |
Unit deck setting: "le scan average dans le deck unit était réglé à 24, j'ai corrigé ensuite (mis à 1, cad tout les cycles)" Conséquence: trous dans les données de quelques mètres
Bottle depths and leak detection (all-stat-depth.xlsx)
Residuals before interpolation:
Calibration
| Experiment | Option | Residuals (STD) | Remark |
|---|---|---|---|
| 1: all stats 1-82 | regular LS | 0.03 (1.64) (fig) | See linear structure with station after 12 |
| 2: all stats 1-82 | linear stat | 0.02 (1.13) (fig) | Much better but remaining struct stat 1-10; removed more outliers |
| 3: stats 12-82 | linear stat | 0.02 (1.05) (fig) | Very nice fit. Deviation at stat 60 below 4000m |
| 4: stats 1-10 | linear stat | -0.095 (1.011) (fig) | Ok but bias looks large |
| 5: stats 1-10 | regular LS | 0.003 (1.17) (fig) | Much better without slope |
| 6: only deep reaching (>3000m) stats | regular LS | 0.02 (1.20) (fig) |
Still 3.5 umol/kg at 5500m/ stat 60 Not sufficient. G. Johnson recommends using fit 3; then fit on station 60, only one coefficient (poor): experience 9 |
| 7: try wswitch to 1 (weight on deep measurements) |
linear ws=1 |
0.06 (1.25) (fig) |
Keeps many more outliers but still residual slope with depth |
| 8: try wswitch to 3 (L4 norm) |
linear ws=3 |
0.08 (1.165) (fig) | Keeps many more outliers but still residual slope with depthSupposed to better fit low and high values |
| 9: Station 60 only; fit all coefs except Pcor | |||
| 10: stats 1-10; wswitch to 1; slope=0 |
regular ws=1 |
Better than exp 5 but still large at station 3 | |
| 11: stats 1-10; wswitch=1; slope=1 |
linear ws=1 |
-0.009 (1.179) (fig) (bydepth) |
** DOES NOT CONVERGE ** linear station dependent bias and weight on deep measurements give a good residual |
|
12: stats 12-82; remove calibration below 3510m |
linear ws=0 |
suffix_3510m. Very nice fit | |
| 13: stats12-82; based on exp 3 but using in situ oxygen solubility |
linear ws=0 |
0.023 (1.056) (fig) | little difference with experience 3 (see in situ T-S-O2) |
| 14: stats 12-82; same as exp 12 but ws=1 because saw about 1umol/kg deep bias |
linear ws=1 |
0.03 (1.08) (rms 1.08) (fig) (bydepth) | At depths, difference with Exp 12 is very small. |
| FROM HERE WITH NEW SOFTWARE |
|
||
| 15: stats 1-10 |
iswitch=2 ws=0 |
0.01 (1.007) (fig) (bydepth) |
Theta-S-O2 comparison |
*** FINAL CHOICE FOR CALIBRATION ***
Bias analysis
- T-S-O2 for experiments 11 and 12 (fig) (but since 11 does not converge…)
- T-S-O2 for experiments 15 and 12 (fig)
Secondary circuit
Used oxygen from secondary sensor
T/S from primary
Summary table
Spikes spotted: stations 2,6,9,10,14,15,16,17,18,21,22,23,29,31,32,40,59,60,61,62,63,7201
| Station |
Gaps in the vertical | sigma-match > 50m (except MLD) | Upcast problem | Downcast problem | Spike | Remarks |
|---|---|---|---|---|---|---|
| All stations | Many spikes in general | |||||
| 4209 |
| remove, 1271-1301m | ||||
| All casts 42 | Noise on small scales; all flagged to 3 | |||||
| 6002-6301 | All oxygen wrong/noisy |
Residuals before interpolation:
Calibration, secondary O2 sensor
| Experiment | Option | Residuals (STD) | Remark |
|---|---|---|---|
| 1: all stats 1-82 | regular LS | 0.05 (1.96) (fig) | Same linear structure with station after 12; but also in stations 1-10 |
|
11: stats 1-10; wswitch=1; slope=1 |
linear ws=1 |
0.02 (1.3)(1.3)(fig) (bydepth) |
Good residual; vertical structure is unbiased: need to check if it converged |
|
New scripts (March 2014) from now on |
|
|
|
|
12: stats 12-82; remove calibration below 3510m |
linear ws=0 |
0.001 (1.23) (fig) (bydepth) |
Optimal as for circuit 1 |
*** FINAL CHOICE FOR CALIBRATION ***
Choice between Primary and Secondary
- Use secondary O2 only for profiles 4201 4202 4204 4205 4207 4208 4209 6001. **WATCH OUT** for calibration purpose we used TS from primary, but we use T and S from secondary: (they remain flagged at 3)
- Comparison primary/secondary for 6001 (compare_primary_secondary_station60.m)
- All profiles with only those from secondary: Theta-O2 for all; isolating secondary
More info
- CCHDO (WOCE/WHP) format: see here
Actions sur le document
