Open DavorJ opened 3 years ago
@DavorJ the drop from mid 2019 is just data that weren't available yet when we made the test series, but I guess you figured that out :) Another item that might influence the multi comparison are the data from the BAOL5xx series. There are some serious doubts about the functioning of these baro's, so I would suggest to drop them for further analysis. Or at least those with major disruptions: BAOL503X_B_00290 BAOL505X_B_B2123 BAOL510X_B_00296 BAOL528X_B_B2128 BAOL529X_B_B2121 BAOL538X_B_002A2 BAOL538X_B_00554 BAOL539X_B_B2122 BAOL541X_B_00640 BAOL547X_B_004AC BAOL547X_B_B2137 BAOL548X_B_B2112 BAOL549X_B_B2138 BAOL550X_B_B2136 BAOL551X_B_B2135
@DavorJ the drop from mid 2019 is just data that weren't available yet when we made the test series, but I guess you figured that out :)
Yes :). Point is that we have to take into consideration potentially missing reference data.
Availability of barometer data for 12h intervals:
The high variance in function of timestamp is because in this period there are lots of barometers that operate on 24h basis instead of 12h basis:
That is why the curve constantly jumps between 60 and 90 barometers. This could have a noisy effect on the multi-comparison (#51) analysis of 12h-interval barometers (e.g. if 24h interval barometers are stationed on higher or lower altitudes).
Plotting in function of 24h gives less variance:![image](https://user-images.githubusercontent.com/4161918/94372930-15aaa780-0102-11eb-8795-32d7a376a32a.png)
And here is how zoomed-in 2019-2020 looks like:![image](https://user-images.githubusercontent.com/4161918/94373012-97023a00-0102-11eb-8f45-541510cdb457.png)
Mid 2019 there is data from some 100 barometers, and then it drops to 50 starting in 2020. This gradual drop (or addition) in barometers might cause inconsistencies in multi comparison (cf. #51).