INTERMAGNET / wg-definitive-data

Repository to track working group discussion for Definitive Data
1 stars 1 forks source link

IMBOT: report of initial testing phase #5

Open leonro opened 3 years ago

leonro commented 3 years ago

IMBOT: report of initial testing phase

IMBOT, an automatic data checking and reporting routine for INTERMAGNET definitive data products, is currently tested for one-second and one-minute data products. As techniques and reporting are different for both data products, we will deal with both resolution types separately.

IMBOT one-second

Contents and jobs

Checking and levels

Chronology:

IMBOT one-minute

Contents and jobs

Chronology:

Next steps

JanReda commented 3 years ago

I have copied to http://magneto.igf.edu.pl/soft/check1min/ updated version check1min.exe This version has been used by me for more than half year. I think it is worth to implement IMBOT one-minute to productive state.

hiroakitoh commented 3 years ago

Thanks, Roman.

IMBOT is great.

Recently, I began to receive messages from 1-min IMBOT, which I found very useful.

I think IMBOT one-minute can be moved to practical use.

leonro commented 3 years ago

Making IMBOT 1min productive means that I remove the "preliminary test" block from the notification e-mails. I also will grant access to the IMBOT configuration files on GITHUB for all DD members (and upcoming alumnis) if not yet done already. This way all members of the DD committee can review/modify referee and observatory lists. Descriptions on how to perform such changes are found within the IMBOT configuration repository. Some of you have used that already. So if data checker, observatories or responsibilities are changing, then please update the data here accordingly. Jan is listed as fallback, if an observatory or data checker is missing. IMBOT 1min can be extended anytime by adding additional features and reports. Such features can be tested for individual observatories and persons, without affecting the basic functionality of the remaining code. But this is something for the future and some hands-on workshop ... If you have any corrections or modifications for the e-mail notification please post them here.

leonro commented 3 years ago

Just one important thing which I forgot in my report: Here are all the results for one-second data of the years 2018 to 2020. Listed are submitted data sets, time of submission, and the level assigned by IMBOT ( 2-excellent data with all meta information; 1-somthing is missing, usually some meta info, 0-corrupted files or missing data)

2018

IAGA code date level
BDV 2019-07-16 0
EBR 2019-04-22 1
LYC 2019-06-11 1
UPS 2019-06-11 2
WIC 2020-09-17 2
ABK 2019-06-11 2
CMO 2020-08-18 1
ASP 2020-09-13 2
HER 2019-10-20 1
MAW 2020-09-11 2
LRM 2020-09-16 2
CKI 2020-09-16 2
CNB 2020-09-21 2
CSY 2020-09-09 1
CTA 2020-09-09 2
GNG 2020-08-31 2
MCQ 2020-09-22 2
KDU 2020-09-09 2
KAK 2020-09-27 0* (needs to be re-evaluated, probably level 2)
MMB 2020-09-28 0* (needs to be re-evaluated, probably level 1-2)
KNY 2020-09-28 0* (needs to be re-evaluated, probably level 1-2)
CLF 2020-12-13 2
KOU 2020-12-14 2
TAM 2020-12-17 2
BOU 2021-01-03 1
FRD 2021-01-03 1
GUA 2021-01-05 1
HON 2021-01-05 1
FRN 2021-01-06 1
NEW 2021-01-07 1
SHU 2021-01-07 1
SIT 2021-01-06 1
BRW 2021-01-11 1
TUC 2021-01-12 1
SJG 2021-02-11 1
BSL 2021-01-13 1
BOX 2021-03-08 2
PHU 2021-03-14 2
BEL 2021-04-12 2
HLP 2021-04-21 2
HRN 2021-04-14 2

2019

IAGA code date level
EBR 2020-06-25 1
WIC 2020-05-25 2
ABK 2020-09-17 1* (needs to be re-evaluated after bug-fix, probably level 2)
LYC 2020-09-17 1* (needs to be re-evaluated after bug-fix, probably level 2)
UPS 2020-09-17 2
BDV 2020-11-04 0
CLF 2020-12-13 2
TAM 2020-12-17 2
BOX 2021-03-07 2
KOU 2021-03-07 2
BEL 2021-04-12 2
HLP 2021-04-21 2
HRN 2021-04-14 2
LRM 2021-08-10 2
ASP 2021-08-13 2
GNG 2021-08-14 2
KDU 2021-08-24 2
CSY 2021-08-26 2
CNB 2021-09-06 2
HER 2021-09-22 2
CKI 2021-09-16 2
MAW 2021-09-16 2
MCQ 2021-09-24 2
CMO 2021-09-27 1
DED 2021-09-28 1
CTA 2021-09-29 0* (just uploaded today - upload still incomplete)

2020

IAGA code date level
BEL 2021-04-12 2
CLF 2021-02-03 2
WIC 2021-05-12 2
BOX 2021-05-20 2
ABK 2021-03-28 1* (needs to be re-evaluated after bug-fix, probably level 2)
HLP 2021-04-21 2
HRN 2021-04-14 1* (needs to be re-evaluated after bug-fix, probably level 2)
UPS 2021-03-28 2
KOU 2021-05-06 2
LYC 2021-07-07 2
TAM 2021-06-30 2
JanReda commented 3 years ago

This is really very important test, thank you. This test shows that more than 60% provided 1-sec definitive reach the best level (2). So it shows great promise for acceptance without delay great majority of provided 1-sec definitive.

Further around 30% reach level 1. Do you know which usually metadata (for level 1) are not correct? I guess that monthly mean values calculated from 1-sec definitive and 1-min definitive almost completely agree, am I right?

Roman, do you see the possibility checking also all 2014-2017 1sec data? I mean qualify to level 2,1 or 0.

On your list is DED 2018, but directory 2018_step1/DED on Paris is empty. From where do the data come?

leonro commented 3 years ago

The most dominant reason (95%) of level 1 classification is missing meta information, in particular the "StandardLevel" and its description (see IMAGCDF document section 4.7). It is very easy to upload this information by just just filling out the meta_OBSCODE.txt template which is delivered together with the IMBOT report. Hermanus for example just did that last week, moving from level 1 to level 2. In a few cases, some data points especially at the end of the year are missing. (last day has 86399 records instead of of 86400). The agreement between 1sec means and 1min data is good in all cases. Minor differences related most likely to different spike treatments etc are listed in the report, but are not termed critical. I would say that only 5% of the submitted data sets need more attention.

Regarding 2014... this probably the most critical year to be evaluated as data uploads are extremely inhomogenuous to say the least. Anyway, I will give it a try. I guess I can run it over the weekend and post the results here. 2016 has been analyzed for the IMBOT manuscript already, which means we just have 2015 and 2017 left afterwards, at least for automatic analyses.

I don't see DED in 2018, so far only 2019 data has been evaluated.

leonro commented 3 years ago

Analyzed 4. October 2021:

2014

IAGA code date level
BOU 2017-02-21 1
BRW 2017-02-14 1
BSL 2017-02-21 1
CMO 2017-03-09 1
DED 2017-03-09 1
EBR 2016-04-17 0
FRD 2017-02-14 1
FRN 2016-10-20 1
GUA 2017-02-14 1
HON 2017-03-02 1
NEW 2017-02-21 1
SHU 2017-03-09 1
BEL 2021-04-12 2
CLF 2020-12-16 2
HLP 2021-04-21 2
SIT 2017-03-02 1
SJG 2017-03-09 1
TUC 2017-03-02 1
HER 2017-08-01 1
HRN 2021-04-14 2
KAK 2016-10-05 0
KNY 2016-10-05 0
MMB 2016-10-05 0
API memory issue 0