EUCAARI AMS Meeting PSI, 28-30 April 2010 Local host: André Prevot Room OLFG 402, PSI, Villigen, Switzerland Short notes, Erik Swietlicki List of participants: André Prevot, PSI, Switzerland Valentin Lanz, PSI, Switzerland Pete de Carlo, PSI, Switzerland André Prevot, PSI, Switzerland Claudia Mohr, PSI, Switzerland Monica Crippa, PSI, Switzerland Erik Swietlicki, Lund University, Sweden Axel Eriksson, Lund University, Sweden Aki Kortelainen, Kuopio, Finland Hao Liqing, Kuopio University, Finland Amewu Mensah, FZ Jülich, Germany Astrid Kiendler-Scharr, FZ Jülich, Germany Laurent Poulain, IFT, Leipzig, Germany Eiko Nemitz, CEH Edinburgh, UK Manuel Dall´Osto, NUIG Galway, Ireland Doug Worsnop, Helsinki University and Aerodyne Samara Carbone, FMI, Helsinki, Finland Sally Ng, Aerodyne, USA Manjula Canagaratna, Aerodyne, USA Donna Sueper, Univ of Colorado, USA Jose Jimenez, Univ of Colorado, USA Will Morgan, UMAN, UK Evelyn Freney, Clermont-Ferrand, France Stefano Decesari, ISAC-CNR (from Thursd 29 15:00) Jay Slowak, Univ Toronto, (from Thursd 29 15:00) Future AMS Meetings AMS workshop: Hyytiälä, 4-6 Sept 2010 (after IAC 2010) AMS Europe community: Barcelona, early March 2011 (Manuel Dall’Osto arranges) HR clinic: Boulder, USA, late March 2011 ACTRIS (EUSAAR follow-up) kick-off meeting, spring 2011, coordinate AMS activities Campaigns suggested to EMEP by European AMS community (via Eiko and André) Jan/Feb 2012 July 2012 Jan/Feb 2013 Deadlines we agreed upon: May 15, 2010: Apply new CE and individual scaling factors. Explain how individual scaling factors were estimated. Submit to PSI server and EBAS. May 15, 2010: Fill in site and instrument information on Google Workbook (link to be circulated). Will include statistics on EC/BC, T, RH. May 15, 2010: Average AMS size distributions (40 bins 20 to 2000 nm). For Q-AMS organics size distributions should be weighted average of m/z measured in pToF. May 15, 2010: Check if m/z 15, 29, 44 vs 43 all go through origin and make necessary corrections. Investigate raw mass spectra peak shapes for m/z 36, 38, 39, 41 to check for surface ionization and report outcome. May 15, 2010: m/z 30, 36, 39 and 46 should be included in the OA matrix in PMF and submitted to PSI server. May 15, 2010: Supply the following to Amewu Mensah, Jülich: - NO2+ t_series from HR (or NO3_ 46 UMR) - NO+ t_series from HR (or NO3_ 30 UMR) - NO3, SO4, NH4, and org t_series (same as to Eiko Nemitz) - m/z 46/30 ratio observed with your instrument during NH4NO3 calibrations Action items and plots we agreed to deliver are also marked in green below. ACTION ITEMS: This is a list of tasks that we agreed that every group should do. CE, Collection Efficiency All groups should use the following CE: CE = 0.5 for nitrate ratios from 0-0.25. CE = 1 when pure ammonium nitrate (nitrate ratio 0.78). CE linear increase in between. Individual scaling factors multiplied with CE If this CE is insufficient to get good agreement with auxiliary data, then apply individual scaling factors (multiply CE with scaling factor). All groups must explain how they estimated their scaling factor, for each individual data set. Use as much auxiliary data possible. Include short description of auxiliary data (e.g. type and size range of SMPS, analytical methods used for filters and on-line samplers, etc.) All groups must send a short description of how scaling factors were estimated to Eiko Nemitz, CEH. Also include inconsistencies with auxiliary data, and explain why they differ, if possible. Resubmit hourly data with updated CE and individual scaling factors. Deadline for resubmission (PSI server and EBAS): 15 May 2010 Update full experiment on PSI server. Organic matrix submission for PMF Checks to be performed on OA data before submission for PMF: Check if m/z 15, 29, 44 vs 43 all go through origin. Plot: m/z 29 vs 43, should go through origin (proper air beam correction?). Plot: m/z 39 vs 43, If this plot is noisy, then the OA data is contaminated by 39K, and not only m/z C3H3+. Plot: m/z 15 vs 43, as check for 15N contamination. Plot: m/z 44 vs 43, should go through origin. m/z 30, 36, 39 and 46 should be included in the OA matrix in PMF. Instructions from Donna Sueper included in minutes below. If surface ionization is suspected: Check raw mass spectra, compare peak shapes m/z 36, 38, 39, 41 (HCl, K). K or HCl? Plot Time series or scatter plots of: 38 vs 36, 41 vs 39. Check these diagnostics before you apply PMF. Scatter plot: m/z 30 vs 46 Deadline 15 May. Upload on PSI server. Organic nitrate Jülich will lead an effort to make every group look at their HR (and unit mass resolution; UMR) data on NO2+ and NO+ to estimate the organic nitrate (OrgNO3) contribution to total nitrate. All groups should do the following: (Data waves only - no full Squirrel experiment needed) To start with: Task 1, high priority: - NO2+ t_series from HR (or NO3_ 46 UMR) - NO+ t_series from HR (or NO3_ 30 UMR) - NO3, SO4, NH4, and org t_series - 46/30 ratio observed with your instrument during NH4NO3 calibrations Nice to have but please do not delay first part with this: Task 2, lower priority: - any data on Na, Ca, Mg that is available from other aerosol measurements - t_series of PMF factors - any gas phase data on NOx, O3, VOCs that is available Deadline, Task 1: 15 May 2010 Please send to: A.Kiendler-Scharr@fz-juelich.de Amines Everyone should look for the amines. Look at m/z 30, 44, 58, 86 for amine peaks in HR. Example given below for SPC 2008. HR analysis Most groups agreed to look at their HR data within next 6 months. 43O, 44O, 44 send to Manuel. Deadline 30 Nov 2010. Manuel Dall’Osto will compile in “Sally triangle plot”. Doug plots: Check SMPS inversion algoritm: Plot SMPS volume distributions to see if there are too high concentrations at high end due to erroneous charge correction in SMPS inversion. Check BFSP IE calibration: Plot BFSP IE calibration histogram for the EUCAARI/EMEP campaigns. Sulfuric acid effect: High concentrations of sulfuric acid will give more spread in CE vs nitrate ratio plot. Check this by plotting CE vs ammonium-to-sulfate ratio. A high ratio means that there is high nitrate. AMS Data server at PSI: ftp://ms_chaos@ftp.psi.ch user: ms_chaos pw: ws**ms (Documents from this meeting: Meeting_PSI) Wednesday 28 April 2010 Start, 13:15 André Prevot Schedule discussions. 14:10 Eiko Nemitz AMS Overview paper, Bulk Aerosol Composition Overview of available data sets. Some data missing still (Hyytiälä, Puy de Dome, Montseny) Concentrations not always reported in STP. At what p/T was your flow calibrated? Average, Feb-March 2009, about 50% ammonium nitrate. Some high ammonium chloride (marine or something else?). Jose: Nature paper 2010 (doi:10.1038/nature08905), Thornton et al, missing source of Cl. Industrial? Wood combustion? Feb-March 2009, large ammonium nitrate peaks across Europe. Spreading manure regulated in Europe, time periods when allowed? Pete de Carlo mz40/28 correction Overview of mz40/28 calculations, instructions on the Wiki. Correct only if ratio is far off. Eiko Nemitz Composition-dependent CE Don´t use CE from Matthew et al (2008) or Crosier et al (2007). Plot that we all made and sent to Eiko: CE vs NO3 fraction (CE = AMS mass vs SMPS volume * estimated density) Plot shown of all Volume AMS/SMPS vs Nitrate ratio (AMS). Large scatter!!! Some outlier instruments. Otherwise increasing trend for CE with nitrate ratio for most AMS. Bin medians look better. Normalized binned plot shows trend clearer. Kink around 0.3 nitrate ratio. Pete: High nitrate goes with large sizes, SMPS misses mass/volume. Reason for disagreement? Internal/external mixture for nitrate may affect CE. Pete: Barcelona AMS data. High nitrate => higher SMPS volume than AMS mass. AMS nitrate correlates well with PM2.5 nitrate (CIEMAT). Doug: Plot SMPS volume distribution to see if there are too high concentrations at high end due to erroneous charge correction in SMPS inversion. Suggestions discussed: CE linear, 3 sections? CE = constant up to nitrate ratio around 0.25. CE = 1 when pure AN (nitrate ratio around 0.78) Sally: BFSP and DMA+CPC ammonium nitrate IE calibration doesn´t always give same value. Do both! Pete: BFSP better, not so many issues (splitting flows, particle size). Doug: Plot BFSP IE calibration histogram for the EUCAARI/EMEP campaigns. Doug: All, put together one slide for each data set on how they estimated CE. To be discussed on Thursday. Doug: Plot: Your favorite CE (or scaling factor) vs nitrate fraction. Missing components in comparison, e.g. BC. Include whenever possible. Sulfuric acid effect. Will give more spread in CE vs nitrate ratio plot. PDD, strange things happening at low nitrate ratio. Not captured by CE vs nitrate ratio plot. Doug: Check this by plotting CE vs ammonium to sulfate ratio. When A/S is high it means that there is high nitrate. Doug: Nitrate CE correction is for nearly all sites more important than acidity correction. Summary (Doug): CE = 0.5 except at high nitrate or SA effect. Other CE are the result of other issues, most likely IE calibration (also lens alignment, problems with other instruments we compare with). Eiko: Suggested to plot CE vs predicted to measured ammonium ratio (boring plot for most sites). Eiko: Time scale? Issues with some data sets where Eiko got Squirrel experiment. Not able to recreate data on different time scales. CE in Squirrel experiment has to be the right one. No manual post-corrections allowed. Eiko: EMEP data base needs hourly time series. Eiko: Resubmit hourly data once CE is settled. Eiko Nemitz Scope of AMS overview paper Table of concentration statistics Bar graphs of bulk composition Comparison of time series Some discussions in context of synoptic meteorology Some modeled concentration fields? back trajectories? Another paper? Comparison of diurnal cycles Comparison of relative size distributions How much on organic analysis, leave for PMF paper? Statistics on EC whenever available (in table only) Time resolution: 60 minutes How to present comparisons of time-series with different data coverage? Size distribution for Q-AMS organics Meteorological data? Actions All CE, decision Thursday Provide average size distributions (40 bins between 20-2000 nm) EC/BC time series when available T/RH when available Fill in table with further information (circulated) 16:40 Individual work on CE plots. Fading out. Thursday 29 April 2010 Start, 9:15 Valentin Lanz PMF on OA. Overview of available OA data sets, submitted to PSI server. Only OA data for PMF from Melpitz, Puijo, Finokalia, SPC, JFJ. Submit OA time series at unity CE and additionally a time-dependent CE and scaling factor wave. PMF performed for some stations (SPC, Puijo, Finokalia, JFJ). Publications for some sites in Europe. Finokalia, spring 2009: PMF1 (m/z 38, 44, 55, 58, 85) – OOA PMF2 (m/z 44, 58, 72, 86) – amines? Large residual for m/z 41 in PMF. 41K? Amine signal due to surface ionization? 3-factor solution (excluding some outliers): PMF1: m/z 38 PMF2: m/z 44 PMF3: m/z 58 Probably strong signal from inorganics (m/z 41, 38 etc) Amine factor, real or artefact? Rarely observed in ambient data. Agricultural area? Ammonium-deficient? Example of amines from San Pietro Capofiume provided by Samara Carbone. Doug: Amines replace ammonium to neutralize acids if amines are present in ambient air. Doug: m/z 30, 36, 39 and 46 should be included in the OA matrix in PMF. Separate these 4 masses. Donna will send around an example on how this should be done (included below). Doug: If surface ionization is suspected: Check raw mass spectra, compare peak shapes m/z 36, 38, 39, 41. K or HCl? Plot Time series or scatter plots of: 38 vs 36, 41 vs 39. Check these diagnostics before you apply PMF. Puijo, Sept-Oct 2008 San Pietro Capofiume, May 2008 Harwell, Spring 2009 4-factor PMF solution, BBOA-like. HOA-like, OOA: m/z 29 split in separate factor, fairly constant over time. Artifact!! Doug: contaminated with 28. Pete: Frag table not applied properly. Air beam not properly subtracted. Eiko: 29 error due to 40/28 correction wrong. If there are errors like this in the OA data set, Valentin should hand it back to the investigator for correction before PMF analysis. Melpitz fpeak = -0.8, maximizes correlation of solution with auxiliary data available. HOA with NOx, OOA-1 with sulfate. OOA-2 not well correlated with nitrate. If OOA factor is LV, then it can be explained by OOA-1 (SV-OOA) and the inverse of temperature. OOA-2 = f(T-1, OOA-1) 4-factor solution BBOA (m/z 44, 60, 73), HOA, OOA-2, OOA-1 Checks to be performed on OA data before submission for PMF: Doug: Plot: m/z 29 vs 43, should go through origin. Plot: m/z 39 vs 43, also useful diagnostic. If this plot is noisy, then the OA data is contaminated by 39K, and not only m/z C3H3+. Plot: m/z 15 vs 43, as check for 15N contamination. Doug: m/z 39 should be included in the default OA data set at submission for PMF. Lunch 12:30 – 13:30 13:30 Amewu Mensah Organic nitrate discussion. PMF results for Cabauw performed by Jülich. 4-factor solution. PMF1 correlates well with HULIS (HNMR from CNR-ISAC) m/z 29, 40, 44 PMF3 SV-OOA correlation with nitrate not that great. BVOC-nitrate discussed also based on chamber data (Ratio NO2+/NO+). Ammonium nitrate ratio NO2+/NO+ = 0.4, biogenic nitrate ratio NO2+/NO+ = 0.1. Measured NO2+/NO+ ratio can then be used to estimate the fraction of organic nitrate. Estimated organic nitrate then correlates much better with PMF3 (SV-OOA)! Organic nitrates are then estimated to be 30% of total nitrate at Cabauw, which is a nitrate hot-spot in Europe. Org-nitrate/nitrate ratio is higher when nitrate/total-AMS-mass is low. At around nitrate/AMS-mass = 0.5, then Org-nitrate approaches zero. Doug: Include NO2+ and NO+ peaks in the PMF analysis. Doug: Jülich should lead an effort to make every group look at their HR (and unit res) data on NO2+ and NO+ to estimate the organic nitrate contribution to total nitrate. Scatter plot: m/z 30 vs 46 14 :00 Eiko Nemitz Updated AMS overview Cabauw: CE: AMS vs MARGA looks OK. All data for CE, when rescaled, the CE trend looks Conclusions: CE = 0.5 for nitrate ratios from 0-0.25. CE = 1 when pure AN (nitrate ratio around 0.78). CE linear increase in between. Decision taken that all groups use this CE. Apply individual scaling factors on top of that. ACTION: All groups should send a short description how they estimated their scaling factor, for each individual data set. Coffee break 14:10 Pete de Carlo High resolution preparation for PMF Exporting matrices from PIKA Export Diff Matrix as Open – Closed Buttons to push from Donna. Sally Ng, Errors in PIKA 16:00 Samara Carbone HR-PMF results from SPC, Po Valley, Italy. 5-factors solution: HOA, BBOA, OOA-II, OOA-I, Amines (m/z 30 + C2H4N). Diurnal: OOA-I increases during daytime, all others decrease (BL variability) CH3SO2 m/z 79 agrees well with MSA from 24-h filters. Comparison with HNMR PMF analysis (ISAC-CNR, Stefano Decesari). Discussion on interpretation of factors. Van Krevelen Diagram, H/C vs O/C Heald et al. A simplified description of the evolution of organic aerosol composition in the atmosphere, GRL, 37, L08803, 5 PP., 2010, doi:10.1029/2010GL042737 A simplified description of the evolution of organic aerosol composition in the atmosphere Doug: Everyone should make this plot. f44 vs f43 in unit mass space. Compare Ng et al (2009) with HOA removed. The factors span this space. All groups should do this (after PMF analysis and removal of HOA). 16:40 Stefano Decesari HNMR results vs AMS Cabauw: Very nice temporal agreement between HULIS (HNMR) and OOA-1b. Large fraction of OOA must be water-insoluble! In SPC, AMS-HOA can account for all filter-WINSOC. BBOA also in HNMR analysis, correlates with AMS-BBOA. Coffee 17:30 Manuel Dall´Osto Mace Head CE Low CE, coincides with high NaCl (sea spray from strong storms). Marine organic aerosol seen at Mace Head. Doug: Check the CE(SMPS) vs AMS-NHCl to check for marine influence on CE in marine influenced environments. m/z 39 (main aliphatic peak) SV-OOA can be split in two components (biogenic, anthropogenic). Friday 30 April 2010 Start, 9:30 Donna Sueper >>>>>Mail from Donna, 29 April , 23:20>>>>>>>>>>>>>>>< As was discussed at the meeting, it will be helpful for Valentin to have some masses - extra from the masses in the Org species - for his use in PMF. This email briefly discusses what you should do to prepare this 2D spectra for him. The technique is the same for Quad and ToF users, there are minor differences for the quad. (1) Press the AMS panel, batch table button and add the following line (easiest at the bottom of the table): PMFadd PMFadd frag_PMFadd 1 1 1 'blank' The 'blank' is a blank value, and empty string, don't enter 'blank'. If you are a quad user, simply leave a blank in the wave called ToF channels. (2) Add the commands below to the command window and execute. make/t/o/n=(1000) root:frag:frag_PMFadd="" // make the frag wave blank root:frag:frag_PMFadd[30-1]="30" // NO root:frag:frag_PMFadd[46-1]="46" // NO2 root:frag:frag_PMFadd[39-1]="39" // K root:frag:frag_PMFadd[36-1]="36" // Chl (3) create a 2D time series wave and error matrix with the tools you usually use for creating the Org matrix with errors, only with the species PMFadd. If you are using Squirrel, a 2D matrix will be created called for example MatD_allV_PMFadd and MatD_allV_PMFadd_err. You may think it will only have 4 columns, corresponding to the four masses on 30, 46, 36, 39, but you would be incorrect. It will have the same dimensions as the Org 2D matrix, with zeros in all the columns that are not 30, 46, 36, 39. I will leave it up to Valentin as to how he may want to incorporate these values. (Depending on whether you use PET, Ingrid Ulbrich's fabulous tool for looking at multiple PMF results, you may want to keep values in increasing m/z order). Note that an alternative, and perhaps simpler method, is to just generate the time series waves when entering 30, 46,36, 39 as the "species" and press the time series button with the calculate error checkbox checked. Please let me know if anything is not clear, Donna >>>>Mail, Donna, 29 April 23:35>>>>>>>>>>>>>> I wanted to follow up with a small point that was brought up Wednesday morning - species specific time series corrections, such as one might make for sulfate or nitrate. Below are detailed instructions as how to use this feature. It is really very simple and (almost) everyone should be using this feature if you wish to make any time-dependent corrections beyond the AB correction. Despite the length of this email it is VERY simple, and you will be glad you know of this. I also put this info on the wiki. In the batch table there is a text wave on the far right that is by default blank (empty strings) called speccorr_list (species correction factor list). In this list one can enter the name of a wave residing in the root folder that you would use as a multiplicative species correction. Below I outline how one can create and apply such a wave. In this example I create a species specific correction for the organic species - this is for illustrative purposes only - to demonstrate functionality - it does not reflect what one should actually do for any particular species. (1) First I make a wave that consists of all ones. Later I will put different values in this wave, values that starts at 1 and then vary between 0 and 2. I plot this graph to see that it is a constant value. make/o/n=(numpnts(root:index:t_series)) myOrgCorr = 1; display myOrgCorr // 1 + sin(p/100) // this is what I will use later (2) In the batch table, in the row corresponding to organics and the column titled speccorr_list, I enter the text 'myOrgcorr' (without the quotes - this needs to be the name of a wave that lives in the root folder and has the same number of points as the time series wave). (3) Now to demonstrate I create an Org time series wave where this new species correction factor wave has a constant value of 1 (which is the same effect of multiplying by 1, or no effect) and another Org time series wave that changes from the original one where this species specific correction varies between 0 and 2. (3a) I go to the MS tab in squirrel, the time series section, and enter Org as a species, create a new graph, select the all (or allV, etc) todo wave, and press the gold calculate button. I should get an Org time series trace that is familiar. (3b) Now I go to the history window and enter the following make/o/n=(numpnts(root:index:t_series)) myOrgCorr = 1 + sin(p/100) I notice in my plot of myOrgCorr I see something that looks like a sine wave, with values varying between 0 and 2. (3c) In the MS tab in squirrel, the time series section and change the display option to "Append to graph" and check the checkbox "Ask for wave name suffix". Press the gold Calculate button again and give the wave a suffix name such as "spCorr". You should be able to see the two different Org traces, one with the (arbitrary) species correction wave factor multiplied by this sine wave. Donna P.S. For future reference see the text of this email is at http://cires.colorado.edu/jimenez-group/wiki/index.php/ToF-AMS_Analysis_Software#The_Frag_and_Batch_Table Will Morgan BAe-146 2008 flights (EUCAARI LONGREX May 2008 + ADIENT-2 Sept 208) Morgan et al , ACP paper. Atmos. Chem. Phys., 10, 4065–4083, 2010 www.atmos-chem-phys.net/10/4065/2010/ Difficult to resolve HOA, dominated by OOA-1, OOA-2. OOA-2 correlates with nitrate in western Europe. OOA-2 correlates also with sulfate. OOA-1 resembles fulvic acid mass spectrum. “Sally triangle plot” (m/z44/OM vs m/z43/OM) for airborne AMS data. Doug: We need more than 2 factors to span the OA space. Apex (fulvic acid). Near-urban, near-source, to background air mass ageing defined from O3/NOx vs CO/NOx. Vertical profiles over Cabauw Enhancement of nitrate with altitude within the boundary layer. Paper in ACPD Atmos. Chem. Phys. Discuss., 10, 10653–10705, 2010 www.atmos-chem-phys-discuss.net/10/10653/2010/ 10:15 Manuel Dall Osto Mace Head revisited in respones to Will. Doug: Apex corresponds to C:O:H of 1:1:1. One O atom for each C atom. Discussion on evolution of organics. HR data will be useful for this purpose. 10:30 André Prevot Which groups can look at their HR data within next 6 months? Most HR groups responded positively. Coffee Samara Carbone ACTION ITEM: Everyone should look for the amines. Look at m/z 30, 44, 58, 86 for amine peaks in HR. André Prevot Integration with other data sets AMS-HTDMA paper (Erik Swietlicki, lead) Aki Virkkula, Lead (scattering closure involving AMS) Jose: AMS size distributions – Mie model – nephelometer. Erik: Vavihill very good agreement with SMPS data vs neph without any chemistry. Modelers? How to invite modelers to use AMS data? Spyros Pandis: PMCAMx David Simpson, EMEP OA model Dominick Spracklen, GLOMAP. AEROCOM meetings (Michael Schulz) EMEP TFMM, Cyprus, 12-13 May 2010 (Eiko, André attending) Formally ended at 12:40