Niwot/Fraser girdling study data QC
See also: Activity log
Soil CO2 data
We use two methods of finding and removing bad data in our soil CO~2~
dataset. The first is event-based, where datapoints generated on bad
sampling days, or bad instrument runs are flagged and removed. Usually
this removes a large block of samples from a particular date. These are
explained in the first three subheadings below. The second method is to
remove datapoints that fall far outside of the distribution of a
statistic describing our data (outliers). In this case, we use a
regression line and remove data that falls too far away from this
predicted line. Finding and removing problematic data and outliers, is
done in the irga_gasbench_process.m
script.
Problematic sampling dates
Sometimes sampling dates have lots of outliers. There may be a known reason for this, or not. These are removed by flagging the IRGA date in which they were run. \^ Bad sample dates \^ Description \^ Reason \^ Action \^ IRGA date \^ | 120717 (NWT) | Enriched d13C | Rain storm? | Removed | 121008 | | | | | | |
Problematic IRGA data
IRGA run quality is assessed using the RMS error of the calibration tank measurements. Lower values are better. \^ Bad IRGA dates \^ Description \^ Reason \^ Action \^ | 110928 | ? | Problem with Licor integration | Removed | | 111004 | ? | Problem with Licor integration | Removed | | 111102 | High RMS error | Bad syringe | none | | 121008 | Enriched d13C | Bad sampling date (see above) | Removed | | | | | |
Problematic gasbench data
Gasbench/IRMS run quality is assessed using its distance from a reference line. In 2011, this was the regression line for gasbench date 111209 (see Niwot_gasbench_verX.m comments around line 262). We retained this line for 2012, but it may make sense to make a new one. \^ Bad Gasbench dates \^ Description \^ Reason \^ Action \^ | 111025 | Off the ref. line | | Removed | | 111130 | Off the ref. line | | Removed | | 120831 | Follows last years ref line | New sample loop installed after this date | Still there (for now) | | 121119 | NWT samples have low peak areas | These failed and were re-run by Brad | Removed (NWT only) |
Outlier detection/removal
In 2011, outliers were removed by distance from a regression line of IRMS area on IRGA CO~2~. It looks like this regression line was created before removing all the bad sampling/instrument dates (above). This is documented in the old Niwot_gasbench_verX.m scripts, and in Dave Bowling's lab notebook #5, pp 58-59. For 2012 this was refined by adding two other methods of bad-data removal.
Currently, two regression relationships, IRMS area vs. IRGA CO~2~ and δ13C vs. 1/IRGA CO~2~, are generated for each site in 2011 and 2012. A threshold distance is set for each line and datapoints beyond this threshold are removed. This threshold is the same in both years and for both sites. This technique should remove anomalies for CO~2~ (leaky vials, bad injections, IRMS area problems) and δ13C (diffusional wierdness and IRMS issues) data.
![media/niwot_girdling/nwt_forestd13co2.png?180|Niwot aboveground forest air d13C vs CO2 concentration] There are also deep (>10cm depth) measurements that have near-atmospheric δ13C values (> -12permil). This is most likely a mistake because any atmospheric air (-8 to -12permil - see fig at right) that is advected into the soil mixes with much more depleted CO~2~. These values are also removed (value > -12 permil AND measurement deeper than OA horizon). This was formerly done using depth/concentration thresholds in the soilCO2_tdist.m file.
Final QC'd data
Number of samples removed, 2011
\^ Problem \^ IRGA dates \^ GB dates \^ IRGA/IRMS CO2 \^ d13C/invCO2 \^ Enr. deep d13CO2 \^ Total 2011 \^ | NWT | 70 | 10 | 9 | 1 | 5 | 95 | | FEF | 49 | 56 | 15 | 4 | 0 | 124 | 219 samples removed total
Number of samples removed, 2012
\^ Problem \^ IRGA dates \^ GB dates \^ IRGA/IRMS CO2 \^ d13C/invCO2 \^ Enr. deep d13CO2 \^ Total 2012 \^ | NWT | 70 | 20 | 4 | 4 | 2 | 100 | | FEF | 0 | 0 | 0 | 4 | 7 | 11 | 111 samples removed total
![media/niwot_girdling/girdling_baddata2011.png?400|Bad data removed 2011]![media/niwot_girdling/girdling_baddata2012.png?400 |Bad data removed 2012 - note 2 sample lines from gasbench]