Followup on Recent Debates with Dan Murphy on AMS Quantification

Transcription

1 Followup on Recent Debates with Dan Murphy on AMS Quantification Jose-Luis Jimenez et al. University of Colorado-Boulder AMS Users Meeting Portland, OR, 21-Oct-2016

2

3 An example of the fallout Paper submitted to PNAS on AMS results for biogenic SOA: Editor: the manuscript is not suitable for publication in PNAS. The manuscript does not sufficiently address the limitations in the methods. Absolute quantification of organic fragments using the AMS is a process prone to uncertainties that are only recently coming to light (see Murphy 2016). Reviewer: [the manuscript] should be very clear about the potential difficulties of absolute quantification using AMS organic signals [given] recent evidence of larger uncertainties in AMS organic signals than are usually reported [Murphy 2016] Clearly many people were taking the Murphy paper as evidence that current AMS ambient data are not quantitative!

4 Key Arguments in Murphy (2016) 1. Proposes simple model for AMS detection and implicitly assumes it is complete! So it can be used to make predictions about the real instrument. 2. AMS sensitivity to OA can vary more than a factor of 3 3. High RIE NH4 (which conflicts w/ model) can be explained by molecular salt evaporation of NH 4 NO 3

5 Group of AMS users involved in early AMS characterization Short version: 1. Simple model: well-known physics, but majorly incomplete 2. AMS sensitivity for ambient: is consistent w/ stated uncertainties Lab studies: always need to calibrate 3. Molecular salt evaporation? Strong evidence against it

6 Signal vs. T v Field: Urban and Forest Lab: Calibration Particles Expected dependence is not observed

7 Evaporation of Molecular Salts? If salts evaporate as molecules, single particle timescales have to be the same for all fragments Completely inconsistent with observations

8 Evaporation of Molecular Salts? If molecular salts evaporate and fragment, RIE NH4 would change with nitrate/sulfate ratio Due to difference in IE of salts, and fragment fractions of each salt No change within the uncertainties Unlikely to observe this if molecular salt evaporation was important

9 Evaporation of Molecular Salts? SV CV Ammonium detection No change in fragments across T v, vaporizers Almost identical to gasphase NH 3 (g) in NIST Sulfate and nitrate have dramatic changes (WW s talk) Very inconsistent with molecular salt evaporation W.W. Hu et al., AMTD, submitted

10 RIE OA : Original data Data taken by Phil Silva ~ 2001 Presented many times (5 conferences + many users meetings) Finally published

11 RIE: all published data MW is not obvious controlling parameter High RIE NH4 and RIE H2O We don t understand why Some high RIE OA for some pure species, but not consistent Lab; always calibrate RIE OA Does this apply to the field?

12 What are our propagated accuracies? Estimated input accuracies (2 ): IE NO3 ~ 10% RIE NH4 ~ 10%, RIE SO4 ~15% RIE OA 20% (means ) CE: 30% (after applying CDCE) Q: ~1% Results NO3 = NH4 ~ 33% ; SO4 ~ 35% OA ~ 37% Note that precisions are often far better Papers often state 15%-20% w/o justification See Supp. Info. of Bahreini et al. JGR (2009):

13 Some ways to check AMS quantification Intercomparisons Well operated SMPS, UHSAS, PILS, MARGA Hygroscopicity closure Light scattering closure

14 Check quantification: intercomp. Compare AMS volume w/ UHSAS Propagate uncertainties Agreement within uncertainties, better at times Field: Urban and Forest Lab: Calibration Particles

15 Check quantif.: hygroscopicity OA ~ 0.1 SO4 ~ 0.6 In situations where OA dominates the mass, OA is very sensitive to RIE OA OA well charact. in lab Does not depend on absolute amounts! Need to calibrate RIE SO4 Field consistent if RIE OA ~1.4 Unphysical results if RIE OA ~ 1.7 Extremely unphysical if OA wrong by x2 or x3

16 Validation of AMS Quant. vs Extinction Extinction / Mass = MSE Use AMS + BC (+metals, dust, salt) mass Use extinction or scattering measurements MSE can be calculated from size distributions Quant. check that relies on simpler instruments, doesn t rely on other chemical instruments First done by DeCarlo et al. (2006) Used last week to disprove a claim that aircraft data was wrong by x2 Washenfelder et al. GRL

17 5 SMPSs in the field (from 2011 AMS Users Mtg) Number Concentration Factor of x5 between instruments None of SMPS agree with CPC! Similar difference in vol. / mass 32

18 SO 4 Intercomparison in Field

19 Lessons from Field Intercomp. Yours truly has spent 2+ years in the field Instruments in prev. 2 slides were all operated by well-known groups, would be trusted by default In reality, little attention to quantification until shown results of intercomparisons (esp. for SMPS) But most campaigns don t do intercomparisons, or there is only 1 instrument to compare to! Determining CE by matching 1 collocated instrument is absurd! Rule of thumb: unless it has a full-time expert operating and calibrating it, and even then there are lots of surprises Campaigns with multiple high-quality instruments, that allow multiple ways of checking, offer the best opportunities to learn

20 An important uncertainty in most situations 2 AMS = 2 CE + 2 RIE + 2 user User-introduced errors Most common: scale to 1 SMPS or PILS etc. Many others occur: - Errors in IE cals - Errors in single SMPS, scaling to it - Errors in software processing - Problems w/ lens or inlet - Problems w/ DAQ threshold - Other problems - This is an important reason why we have Clinics! Ambient: Best IE practices, CDCE, no scaling! Reporting CEs from scaling in literature has created a lot of confusion

21 Check quantification: intercomp. Replotting data in Docherty et al. (2011) Ratio of AMS to Sunset OC vs primary fraction RIE POA may be slightly higher than RIE SOA But all within AMS uncertainties (CE esp.) We hesitated to change RIE POA based only on this dataset (lots of Sunset issues too!) People (with high quality datasets) should do this analysis

22 Dan s Response to our comment Just accepted into AS&T 1. Simple model: more discussion of model details Our take: Implicitly agrees that it is likely incomplete 2. AMS sensitivity: Ambient: There is also agreement that the great majority of AMS data for ambient conditions do not need to be revised by factors of 2 or 3. Lab/source: Importance using a sensitivity appropriate to each type of OA in the AMS rather than a single number Our take: that is exactly what our published response says! But it is not at what people were understanding from his paper. 3. RIE NH4 : I am unable to find any other good explanation for the sensitivity to ammonium. J et al. do not propose an alternative explanation Our take: additional evidence makes his model even less likely to be correct. We don t have a good explanation despite years of discussions. Thus the empirical calibration approach.

23 Outlook & next steps Aerodyne working on updated RIE OA for ambient (next talk) Users should calibrate for lab / source OA This is each user s responsibility. Core users get blamed (e.g. by Murphy) for lack of calibrations in users community, but this is not fair. Users should do these analysis when they have high quality AMS / ACSM data AND high-quality data from other instruments Hygroscopicity closure Scattering closure (we can provide Igor Mie Code)* Vary RIE POA after PMF, see if R 2 and X 2 improve Publish these, bring for discussion at AMS Clinic & Users M Consider capture vaporizer so CE ~1 Removing that uncertainty will bring to light many issues!

24 Plot shown in discussion: RIE cals. in Murphy s instrument Individual cals are x2 away from the line, instead of 20% for AMS cals Extrapolating calibration by a factor of x2 to high-laser power, w/o doing any calibrations (ironic!) They still report concentrations based on these cals. From Liao et al., JGR 2015: