Comparing BOM data with Jim Cumming's Argo data
Time profiles for 2010
Possible juld_qc values for 2010
| bom_juld_qc | -999 (32,045), 0 (3,982), 1 (69,617), 4 (1,277) |
|---|---|
| bom_jim_juld_qc | -999 (5,566), 0 (26,832), 1 (68,587) and 4 (1,274) |
Relationships
- bom_juld_qc = -999 => bom_jim_juld_qc = 0 (22,895 times),
-999 (4,747 times) or 1 (8 times)
- bom_jim_juld_qc = -999 => bom_juld_qc = -999 (4,756 times), 1 (745 times), 0 (62 times) or 4 (twice)
- bom_juld_qc = 0 => bom_jim_juld_qc = 0 (except for 62 values)
- bom_jim_juld_qc = 0 => bom_juld_qc = 0 (3,924 times) or -999 (22,847 times) (or 1 (199 times) and 4 (once for argo 1900848 at juld=22024).
- When bom_jim_juld_qc = 0 and bom_juld_qc != 0 (24,341 times), bom_juld_diff = -999 (24,088 times - and bom_pos_qc = -999) and bom_juld_diff != -999 (253 times).
- bom_juld_qc = 1 <=> bom_jim_juld_qc = 1 (except for 952 values, 944 when bom_juld_qc = 4 and bom_jim_juld_qc != 4 and 8 for the other way round)
- bom_juld_qc = 4 <=> bom_jim_juld_qc = 4 (except for 3 values where bom_juld_qc = 4 and bom_jim_juld_qc = 0 or -999, which are argo 1900848 at juld = 22,024.2 and argo 2900890 at 22,105.6 and 22,185.8)
- bom_juld_qc = -999 <=> bom_juld_diff = -999 (without exception)
- bom_jim_juld_qc = -999 <=> bom_jim_juld_diff = -999 (without exception)
Summary
Looks like generally:
- For QC values of 1 and 4, bom_juld_qc = bom_jim_juld_qc
- When bom_jim_juld_qc = -999 => bom_juld_qc = -999.
- When bom_juld_qc = 0 => bom_jim_juld_qc = 0
- However, a lot of bom_juld_qc = -999 seems to have interpretted by bom_jim_juld_qc at 0. So undefined BoM QC (=-999) is taken in Jim's data as not done (=0).
- bom_juld_qc = -999 <=> bom_juld_diff = -999 and bom_jim_juld_qc = -999 <=> bom_jim_juld_diff = -999
Follow-up, involving position information. It's always true that BoM set JULD_QC to fill value (-999, here) when POS_QC is fill value and vice versa - this is probably when we don't have any BoM data. Maybe BoM only keep the data when both the JULD_QC and POS_QC tests are passed.
Whereas it's possible for Jim's BoM data to have the JULD_QC not done and for the POS_QC to be set to the fill value.
Position profiles for 2010
Possible pos_qc values for 2010
| bom_pos_qc | -999 (33,927), 0 (37), 1 (82,232), 3 (138), 4 (60), 5 (173), 8 (6) |
|---|---|
| bom_jim_pos_qc | -999 (35,229), 1 (37), 2 (80,953), 4 (138), 5 (44), 6 (172) |
Relationships
- bom_pos_qc = -999 <=> bom_jim_pos_qc = -999 (33,927 times)
- when bom_jim_pos_qc = -999 and bom_pos_qc != -999 (1,302 times), bom_pos_qc = 1 (1,281 times), 4 (16) and 8 (5).
- When bom_jim_pos_qc = -999 and bom_pos_qc = 1 (1,281 times), bom_jim_juld_qc = -999 (1,014 times) or 0 (267 times).
- bom_pos_qc = 0 <=> bom_jim_pos_qc = 1 (without exception)
- bom_pos_qc = 1 <=> bom_jim_pos_qc = 2 (80,951 times)
- bom_pos_qc = 3 <=> bom_jim_pos_qc = 4 (without exception)
- bom_pos_qc = 4 <=> bom_jim_pos_qc = 5 (44 times)
- bom_pos_qc = 5 <=> bom_jim_pos_qc = 6 (172 times)
- bom_pos_qc = 8 => bom_jim_pos_qc = -999 (5 times) or 2 (once)
- bom_juld_qc = -999 <=> bom_pos_qc = -999
- bom_jim_juld_qc = -999 => bom_jim_pos_qc = -999, but 24,361 times when the opposite is not true (on 24,088 times this happens when bom_juld_qc/bom_pos_qc = -999.
Summary
Looks like generally, bom_jim_pos_qc =
- -999 if bom_pos_qc = -999 or 8,
- = (bom_pos_qc + 1) otherwise
- I have more positional data than Jim (less -999s and more of the other values). Possibily because more data has been processed in the proceeding two weeks from when it was created.
Temperature profiles for 2010
Possible tmp_qc values.
| bom_tmp_qc value | -999 (33,971), 0/A (74,355), 1/B (5,242), 2/C (354), 3/D (138), 4 (80) and 5 (2,433) |
|---|---|
| bom_jim_tmp_qc | -999 (35,225), 7 (73,583), 8 (5,178), 9 (344), 10 (131), 11 (77) and 12 (2,035) |
Relationships
- bom_tmp_qc = 0/A <=> bom_jim_tmp_qc = 7 (73,581 times)
- bom_tmp_qc = 1/B <=> bom_jim_tmp_qc = 8 (5,172 times)
- bom_tmp_qc = 2/C <=> bom_jim_tmp_qc = 9 (344 times)
- bom_tmp_qc = 3/D <=> bom_jim_tmp_qc = 10 (131 times)
- bom_tmp_qc = 4/E <=> bom_jim_tmp_qc = 11 (77 times)
- bom_tmp_qc = 5/F <=> bom_jim_tmp_qc = 12 (2,034 times)
- bom_tmp_qc = -999 <=> bom_jim_tmp_qc = -999 (33,971 times) except
1,254 times when bom_jim_tmp_qc = -999 and bom_tmp_qc equals
- 0/A (767 times)
- 1/B (69 times)
- 2/C (10 times)
- 3/D (7 times)
- 4/E (3 times)
- 5/F (398 times)
Temperature levels for 2010 for 4900* argo data
Possible tmp_qc_lvl values for 4900* argo data.
| bom_tmp_qc_lvl | -999 (1,170,930), 0 (2,127,725), 1 (422,751), 3 (3,204), 4 (2,204), 7 (1), 9 (137) |
|---|---|
| bom_jim_tmp_qc_lvl | -999 (3,115,532), 2 (421,944), 4 (3,179), 5 (1,993) and 9 (3) |
I need to check but looks like for bom_tmp_qc_lvl => bom_jim_tmp_qc_lvl that -999, 0 => -999, 1 => 2, 3 => 4, 4 => 5.
ETS score for temperature profiles for 2010
Taking 0 to be good and 5 to be bad
| bom | ||||
| Reject | Accept | Total | ||
| d_mode | Reject | 2004 (2.682%) | 319 (0.427%) | 2323 (3.108%) |
| Accept | 335 (0.448%) | 72073 (96.443%) | 72408 (96.892%) | |
| Total | 2339 (3.130%) | 72392 (96.870%) | 74731 (100%) | |
bias: 0.993159469858914