Notes to use the 20200214 updated nc data files:
1) NOAA/INSTAAR 13CH4 data, -999.99 value in 'value_unc' is not replaced by 0.065 per mil in the nc files. Additional code are required to do the replacement. Use 0.065 per mile for data points with -999.99 value, or any other negative value (-0.009 sometimes). When data points already have assigned uncertainties in the database, then we stick with those.  
2) For external 13CH4 data, SUBTRACT the number of ‘ch4c13_scale_offset’ (global attributes) from the value of 13CH4 in the file to bring them to INSTAAR scale, i.e. Offset = External lab - INSTAAR. ‘value_unc’ accounts for measurement uncertainty and scale correction uncertainty.
3) For external CH4 data, multiply external CH4 data and 'ch4_scale_multiplier' (global attributes) to bring them to the NOAA CH4 scale. ‘value_unc’ accounts for measurement uncertainty and scale correction uncertainty.
4) Hourly/high-freq CH4 data use standard deviation as ‘value_unc’, and measurement uncertainty (or include scale correction uncertainty for external sites) as ‘measurement_unc’.  Maybe better to use ‘value_unc’ only instead of adding both in quadrature since the physical meanings are different. ‘measurement_unc’ can be used when ‘value_unc’ is not well defined due to a limited number of measurements within the hour (e.g. N=1).

Notes on the flag: (‘super-safe’ dataset: only data points without the first two columns flag, i.e. flag column should be'..*' while * can be any character)
1) Aircraft and shipboard data are not flagged. Site codes for shipboard data are POC and NGM.
2) For continental sites CH4, mid-day data (mostly for 11-16 local standard time, highest intake is used if available) are selected and a ‘+3SD and -4SD’ filter are used to flag statistical outliers, which has a second column flag of ‘S’. The unselected data (outside 11-16 LST windows) are flagged with a second column flag of ‘U’.
3) For mountain sites CH4, nighttime data (generally 0-5 local standard time) are selected and a ‘+3SD and -4SD’ filter are used to flag statistical outliers, which has a second column flag of ‘S’. The unselected are flagged with a second column flag of ‘U’. Sourish, See the mountain site list I sent earlier for details.
4)  For remote islands CH4, hourly filter is not applied. Only use the ‘+3SD and -4SD’ filter. 
5) All external 13CH4 are event data. Hourly filter is not applied. Only use the ‘+3SD and -4SD’ filter. Very limited data points are flagged as outliers. 
6) For NOAA event data, statistical outliers are already in dataset (from Ed’s QA/QC). For hourly data, we add in statistical outlier flag (except for MLO and BRW who already have flags) for selective height and time: highest intake and 11-16 local standard time, similar as (2). 
7) Aircraft campaign data may be reserved for model evaluation. Sourish will make the call.
8) FYI, I found these sites (eyeballing, without using algorithms) have sparse/short/inconsistent temporal coverage (‘inconsistent’ means up to 1-2 year gaps occurred more than once in between the start and end date):
* *         PDM_event_LSCE
* *         ORL_event_LSCE
* *         TRM_event_LSCE
* *         BIK_event_MPI
* *         BRZ_hourly_NIES
* *         NOY_hourly_NIES
* *         FNE_hourly_EC
* *         HNP_hourly_EC
* *         TAO_hourly_EC
* *         CRI_event_CSIRO
* *         GPA_event_CSIRO
* *         BIK_event_MPI (maybe ok, a 1.5 year gap in 2014 and 2015)
* *         KJO_event_MPI