TITLE: Aircraft C-130 MOI sized Anions and Cations (Huebert) AUTHORS: PI: Barry J. Huebert Co-PIs: Byron Blomquist Steve Howell Jackie Heath Timothy Bertram Jena Kline Technician who ran samples: John Zhuang Dept. of Oceanography 1000 Pope Rd., MSB 407 (only needed for courier deliveries) University of Hawaii Honolulu, HI 96822 USA 1-808-956-6896 phone 1-808-956-9225 fax 1.0 DATA SET OVERVIEW: This data was collected from the NCAR/NSF C-130 during the ACE-Asia program. It consists of samples collected using an eight-impactor ?Flying Moudi? (FM) micro-orifice impactor (MOI) built by MSP Corp. This impactor has a higher flow rate (100 lpm) than the typical MOUDI impactor, and distributes particles across 5 stages and a backup filter. The nominal 50% cutpoints of the stages were 5, 1.4, 0.77, 0.44, and 0.26 microns. Analyses of substrates and filters was done by ion chromatography of aqueous extracts. Since there are several ways in which size distributions can be distorted when using a cascade impactor, we present our MOI data in two very different forms here. One is the raw apparent concentrations from each impactor stage: the blank-corrected amount of each analyte per stage, divided by the volume of air sampled. Since this data set does not correct for the ways the distribution has (almost certainly) been modified by the inlet, plumbing losses, and particle bounce, it does not realistically represent the ambient size distribution of these ions. However, since no assumptions nor corrections have been applied, it is useful for looking at ionic ratios in the super- and sub-micron size ranges. [Nitrate is usually on big particles with calcium, for instance, rather than on submicron particles with ammonium.] This uncorrected data set comes with uncertainties that include flow rates, blank corrections, and analytical uncertainties, so you can constrain the range of ion ratios in large and small particles defendably. (In fact, the ratios are even better, since the 5% flow uncertainty drops out.) However, these uncertainties do not include the potential distortions of the distributions during sampling. The other set is corrected to the best of our ability for sampling changes caused by 1) the LTI inlet, 2) some surprisingly large losses in the MOI itself, and 3) the bounce of large particles (especially dry dust) from their intended stage to the backup filter. These corrections change only the backup filter (from which bounced material has been removed) and the supermicron stages (for which losses and LTI effects become important and to which the bounced material has been added). The submicron impactor stages are identical in both data sets. Although this ?corrected? data set is constrained by the total aerosol sampler (TAS) data, the corrections involve a variety of assumptions, some of which are untestable (they are described in detail below). Among them is our ability to represent ambient conditions well enough in the laboratory to derive efficiencies that apply to both sticky and bouncy particles under field conditions. Therefore, the ?corrected? data is not accompanied by uncertainties. Rather, we provide a table that shows how closely the sums of the ionic concentrations on the corrected stages during each sampling interval compares with the TAS concentrations during that same interval. It is somewhat comforting that the over the program, plots of the corrected MOI data vs TAS data have slopes close to one and an R2 > 0.8 for Ca, NSS, and NH4 (0.65 for NO3). We would be remiss if we didn't mention the potential for gas-phase artifacts. For NO3- there is the potential for collection of HNO3 on previously-collected alkaline particles or evaporation of NH4NO3. For NH4+ there is the possibility that NH3 vapor reacted with H2SO4 on the filter or that NH4NO3 evaporated. For SO4= there is the potential that SO2 reacted on the filter with alkaline particles to produce NSS. However, impactors tend to reduce condensation artifacts relative to filters, since the gases don't touch the deposited particles much. Furthermore, our relatively short sampling times reduce the potential for evaporative artifacts, since they happen when vapors in equilibrium with particles disappear during extended sampling. Also, we didn't have much NH4NO3 to evaporate, since most of the NO3- was in a non-volatile form on big particles with Ca++. We don't believe gas phase artifacts are a major source of error in this data set. The data set is in the form of an Excel spreadsheet that includes the sample name, the UTC (Greenwich) date of the sample start, the UTC start and stop times, apparent and corrected concentrations of the ions in each size range (in nanograms of analyte per standard cubic meter of air: 1013 mbar and 298K), and the uncertainty of each apparent value. Each sample includes 6 lines of concentrations and uncertainties, for the 5 stages and the backup filter. The uncertainties for the uncorrected data were derived by propagating errors from the blanks, flowmeters, and analytical procedures. More detail on the location of each sample and the conditions of each flight can be found in the C-130 Flight Reports and C-130 data in the ACE-Asia catalog (http://www.joss.ucar.edu/ace-asia/dm/). This version is being released without adding the start and stop locations and altitudes to the table. The first reported samples are from the fourth research flight (RF04), flown on 6 April 2001. All research flights have a sample name starting with RF and were flown from Iwakuni MCAS, Japan. The last sample is from a return ferry flight (FF06), on 8/9 May 2001, between Wake Island and Hawaii. There are no samples for some flights because an occasional leak introduced unresolvable uncertainties for those flights. It is important to keep in mind what you can and cannot do with this data. You can use the uncorrected data to look at ratios of various ions in each size range. (Do not use ion ratios from the backup filter, however, as it often includes a mixture of genuinely small particles and large particles that bounced off their intended stage.) You can separately characterize the dust or seasalt mode and the accumulation mode. This can be very powerful when studying ions like nitrate, which are tightly coupled to gas phase photochemistry. It is clear, for instance, that nitrate virtually always has a distribution like calcium in the presence of dust, in stark contrast to the Americas where most of the nitrate is accumulation mode. Mechanistically, where does SO2 go? Does it react on dust or not? SO2 has major impacts in terms of acidity, visibility, hydrological cycles, etc., so we need to know whether models properly describe how it reacts. However, above about 1-3 microns, you cannot equate the uncorrected concentrations to the atmospheric concentration. The apparent concentration of large particles (mass found on a substrate divided by volume of air sampled) does not equal the atmospheric concentration. The relationship between those two involves an LTI, whose enrichment of large particles is a topic of research, and a complex maze of plumbing that includes valves and corners and fittings and many opportunities for large particle loss. Most of the time, the reality for any sample will also involve some bouncing of particles from substrates and some sticking. For many samples these chemical size distributions were collected simultaneously with TAS samples, which give accurate (if noisy) bulk aerosol ion concentrations, free of inlet artifacts for any size. It is comforting to have a dumb (but defendable) bulk concentration measurement at times. One final word of caution: the more we have tested the Flying Moudi, the more problems we have encountered. Since the original MOUDI impactor has such sharp size cuts, we imagined the same would be true for the FM. However, when we measured its kernel functions for small particles in the lab, we were distressed to find that they are very broad (one size of particle might have more than 10% of its mass on each of three stages) and that they depend strongly on how the impactor is mounted! All eight impactor stacks are attached to two plena: an input and an output plenum. Unfortunately, the distance between the plena is about 0.3 cm larger that the distance between the fittings of the properly assembled impactor stacks, so the stages must be pulled apart to be placed in the sampler. Our measurements indicate that this distortion reduces the collection efficiency of some stages by as much as 50%! Thus, the ambient submicron distribution peaks at a larger diameter than the apparent one from the MOI. We feel confident that the ionic ratios on these small stages are useful, but OPC and DMA data should be used for assessing the shape and size of the accumulation mode. 2.0 INSTRUMENT DESCRIPTION: Air was conveyed into the aircraft by a port-side low-turbulence inlet (LTI) that was dedicated to this MOI. The efficiency of this inlet was a function of many factors, including airspeed and particle size. Generally, it enhanced particles in the 0.7 - 1.4 micron range by 3%, in the 1.4 - 5 micron range by roughly 26%, and in the 5 - 10 micron range by about 50%. Size-dependent losses in the MOI plumbing, which offset these to a (variable) degree, are discussed below. Inside the fuselage the flow passed through a cm ID metal Y section and a short section of 1.9 cm ID conductive silicone tubing to the entrance of the MOI. The total length of the interior pl. The other branch of the Y went to an auxiliary 90 mm Teflon filter that was used when there were more sampling legs than we had impactor stacks (7 could be used for samples and one as a blank on each flight). This MOI impactor was built by MSP, based on a design used for the CIRPAS Twin Otter (it flies in a wing pod on the TO), but with some significant differences. The most obvious of these is that incoming air first encounters a 10 micron scalper before it enters the distribution plenum that is connected to the 8 impactor stacks. The second large difference is that the first stage has a nominal 5 micron cut (the TO MOI starts with 2.5 micron). Both of these are critical, since (as we found in post-field lab calibrations) the plenum has a poor efficiency for transporting particles larger than 2.5 micron and the 5 micron stage does not distribute flow evenly. (We have quantified these effects using monodisperse aerosols in the lab.) Aluminum foil substrates were used in each impactor stage, and 90 mm, 1 micron pore-size Gelman Zefluor Teflon filters were used as the backup stage. The Al substrates were punched from foil at UH. Both filters and substrates were washed to reduce the variability of their blanks. The Teflon filters were first wetted with alcohol and then rinsed 3 times with DI water prior to drying on a laminar-flow clean bench. Still, the uncertainties of the filter concentrations were often an order of magnitude or more larger than those of the impactor stages, because of the highly variable contamination on the Gelman product as determined from the analysis of field blanks. Our group will not use Gelman Zefluor filters in future field programs. The MOI contained a variable-speed DC pump in a feedback loop with a thermal mass flowmeter and P and T sensors, so that it maintained the volumetric flow at 100 lpm. We found that this system and the data-recording processor inside the MOI worked very well, even though the external exposure-control software (for selecting the filter to be exposed, for instance) often had problems. After exposure the stacks were unloaded in a glove box to minimize contamination. Each filter was placed in a microclean polyethylene bag and extracted using 1 ml of ethanol (to wet the filter) and 9 ml of weak acid solution (10-5 M trifluoroacetic acid) in an ultrasonic bath. Each substrate was extracted with 10 ml of the weak acid solution, but with manual agitation rather than ultrasound. The weak acid prevented the loss of ammonium ion, but it probably increased the dissolution of carbonates (and Ca?) relative to DI water. All filter and substrate extracts were analyzed by ion chromatography as soon after a flight as possible. In some cases the analysis began the evening after the flight, but in a few cases analyses were delayed by 2-3 days due to IC repair. The extracts were analyzed by suppressed ion chromatography on two Dionex ICs (one for anions and one for cations), using procedures identical to those described by Huebert et al. (JGR, 103, 16493-16510, 1998). 3.0 DATA COLLECTION AND PROCESSING: 3.1 DESCRIPTION OF DATA COLLECTION: Stacks were loaded with substrates and filters the night before each flight. This loading was done in a glove box to minimize contamination. Just prior to a flight the eight stacks would be loaded into the MOI sampler aboard the C-130. Prior to RF12 we were unaware that o-rings which were intended to seal each stack to the inlet and exhaust plena could sometimes roll out of their slots, causing leaks that invalidated an entire flight's MOI data. After we detected this problem (the results looked strange), we instituted stringent leak-testing after emplacing the stacks and before takeoff, so that data after RF12 is free of this concern. When ready for sampling, the operator (either Tim Bertram or Jackie Heath) would tell the control computer to begin sampling with one of the eight stacks. This would start the MOI pump and open a valve behind the chosen stack, to start flow through that stack. Flow would continue to the end of that sampling leg, which was usually between 20 and 70 minutes, at which time the valve closed. Because we needed to collect material over tens of minutes, most samples corresponded to level legs, with virtually no sampling during ascents and descents. The exception is that on a few legs we had to change altitude (usually to avoid clouds), during which we continued sampling. A few samples had to be shut off briefly due to rain, so those samples have more than one start and stop time. The filters and substrates from each flight were unloaded from their holders as soon as possible after each flight. Substrate and filter holders were all washed with DI prior to loading the media for the next flight. 3.2 DESCRIPTION OF DERIVED PARAMETERS AND PROCESSING TECHNIQUES USED: Flow, temperature, and pressure for the MOI were logged by its dedicated processor and then downloaded to a personal computer. Uncertainties in the flows are estimated at 3-5%. Relative humidity was also measured in the MOI, to estimate growth or shrinkage relative to ambient RH. Concentrations were derived by first subtracting the blank analyte from that on a substrate or filter. The above-blank analyte in nanograms was then divided by the volume of air sampled (in standard cubic meters at 298K and 1 atmosphere) to get ng X/sm3. The uncertainties were derived as: ((X) = SQRT{ (0.05*[X])2 + (2*(X/V)2 + (DLx/V)2 } where ((X) is the uncertainty in the atmospheric concentration, [X], of substance X, (X is the standard deviation of X on blanks in nanograms, DLX is the analytical detection limit of the IC for X expressed as ng X in a 10 ml extract, and V is the sampled air volume in standard cubic meters. To generate ?corrected? data from the apparent data: A. Calcium and sodium were corrected for bounce by assuming that these large-particle ions should only be half as concentrated on the backup filter as on the stage above the filter. We assumed, therefore, that: [Calcium(Backup Filter)] = [Calcium(Stage5)] / 2[Calcium(Backup Filter)] Any excess Ca (or Na) on the backup filter was therefore removed and added to that on Stage 1 to conserve mass. The decision to place it all on Stage 1 (rather than distributing it among the large stages) was arbitrary. Our intent was to make the total amount of each ion similar to that of TAS, so that we could accurately describe sub- and super-micron amounts of each ion. However, the shape of the supermicron distribution will obviously be influenced by our assumption about the source of the bounced material. Users should rely on APS and OPC physical-size data for information about the shape of the supermicron mode, rather than using this corrected chemical data. Sulfate, NSS, and nitrate were corrected along with calcium by assuming that the ratios of these ions on the uncorrected Stage 1 represents their actual ambient ratios. Therefore, sulfate, NSS, and nitrate were moved from the backup filter to the first stage in proportion to the calcium that was moved: [NO3(FilterCorr)] = [NO3(FilterApp)] - {[DeltaCalcium(Filter)]*[NO3(Stage1)]/[Ca(Stage1)]} To avoid generating negative concentrations on the filter, all negative values were replaced with zeros. B. The LTI enhances the concentration of large particles relative to ambient concentrations. This enhancement depends on a number of factors, including the altitude (density) of the ambient air and the airspeed of the plane. DU supplied a table of data from each flight that lists several state and operational parameters vs time. These parameters can be used to reproduce the Fluent-modeled enhancements of the LTI. The MBL and FT enhancement values were roughly as follows: | 0.5 microns | 1 micron | 2.5 microns | 5 microns | 7 microns | 10 microns FT | 1.01 | 1.03 | 1.06 | 1.24 | 1.9 | 2.8 MBL | 1.01 | 1.02 | 1.03 | 1.14 | 1.5 | 2.0 Since the efficiency of the LTI for 5 to 10 micron particles changes by more than a factor of two over the range, we selected an effective value for the purpose of correcting the apparent data (Column I) that assumed most of the mass would be on sizes closer to 5 than 10 microns. This assumption will no doubt be better for some samples than others. C. The correction for losses of particles in the MOI itself is problematic in several ways. The first of these is the fact that particles have very different morphologies, so some (dry dust) may not be lost to the walls as much as stickier (sea salt) particles might. In the PELTI tests of inlets it was clear that mineral particles passed more efficiently through a solid diffuser inlet than sea salt did, for instance. The second problem is that the corrections are so large for supermicron particles. Only 23% of 5 micron particles and 6% of 7.5 micron particles reached their intended substrate (Stage 1). One would therefore need to multiply apparent concentrations by a factor of 4 for 5 micron and of 16 for 7.5 micron particles, to reproduce ambient concentrations. Given that there is uncertainty in the lab efficiencies, one begins to wonder if there is enough information left after these huge losses to even make a correction. Slight differences in the ambient size would lead to large differences in the required correction. We responded to this by making the loss-correction factor for the largest stage into a variable parameter that we used to force the sum of MOI Ca to approximate the TAS Ca for simultaneous samples. Since we wanted all ions to match up, this involved some compromises and did not result in exact agreement for any ion. For most samples we used a Stage 1 efficiency of 0.15, but it was sometimes as small as 0.1 or as large as 0.35. We do not pretend, therefore, that the shape of the ?corrected? supermicron size distribution is the same as the ambient distribution. We do believe, however, that this approach leads to a reasonable approximation of supermicron and submicron mode concentrations for each analyte, which is a valuable thing to know for interpreting physical size distributions and computing indices of refraction in various size ranges. 3.3 DESCRIPTION OF QUALITY CONTROL PROCEDURES: During each flight one stack was exposed as a field blank. This stack was handled just like the others and was mounted in the MOI, but it was exposed for only 10 seconds. Thus, its handling and history were identical to the actual samples. The mass of each analyte from the 5 substrates in a day's field blank stack were averaged and subtracted from the analyte on that days sample substrates. The same was true of the Teflon filter in the blank stack. Twice the standard deviation of a day's substrate blanks was used for the blank uncertainty (see above). For the Teflon filter blank variability, the flights prior to RF12 and those after RF12 were treated as separate populations for computing blank standard deviation: the blanks after RF12 had a much lower variability, so our sensitivity is greater for the later flights. As in the equation above, uncertainties were derived by propagating the uncertainties from the analytical uncertainty and flowmeters (5% of the value), the instrument detection limit, and twice the blank variability, as a sum of squares. Since we knew that some early flights were troubled by leaks between the plena and one or more stacks, we needed (after the fact) to determine which data was valid and which was not. For this we used a comparison between the NSS (which was nearly all submicron and should therefore be minimally impacted by LTI enhancement and tubing losses) and TAS values. If the sum of NSS on the uncorrected MOI stages was not within a factor of two of the NSS found on a simultaneous TAS sample, we took this as an indication that there was a leak in the MOI. On this basis we usually discarded all the MOI data from that flight. If we had MOI/TAS pairs from the same flight where the MOI sum was very close to the TAS concentration, we retained that sample. 3.4 DATA INTERCOMPARISONS: To test our analytical procedure, we participated in an ACE-Asia inter-laboratory comparison organized by P. Quinn of NOAA/PMEL. She prepared two spiked solutions each for anions and cations, with known (to her) amounts of each analyte and sent these to nine ACE-Asia laboratories to test their analytical calibrations. Although we could not keep them refrigerated enroute, we took the intercomparision samples to Japan so we could analyze them while we were analyzing our aircraft samples. Our results were within a 1-3% percent of the calculated concentrations for Cl-, SO4=, Na+, NH4+, and K+, but 7% low for Mg++. Our NO3- values much smaller than what was prepared, which we believe to be due to loss during transit rather than analytical error. Our Ca++ values were 20 - 38% high, which we are unable to explain. We have remade Ca standards from other materials and have been unable to find a reason for this deviation. The stock standards we used in the field seem to be correct. 4.0 DATA FORMAT: The file is an Excel spreadsheet, prepared on a Macintosh with version Excel X. The columns are: A. Sample name, as RF##HHMM IZ-J, where ## is the flight number, HHMM is the UTC start time of the sample, I identifies it as an impactor sample, Z is the stack number, and J is the stage of that stack. B. C-130 research flight number C. Date of the start of the sample, in UTC D. Sample start time, UTC, as HHMMSS. Seconds were not recorded for RF4 and RF6. E. Sample stop time, UTC, as HHMMSS. Seconds were not recorded for RF4 and RF6. F. The nominal 50% cut size of the stage in µm aerodynamic diameter. G. The geometric mean of sizes that should be collected by each stage, which we customarily use as the X-axis for plotting MOI distributions. H. The effective MOI passing efficiency for particles that would be collected on each stage. This laboratory-derived value was weighted to account for the likely variation of mass across the size range of each stage in order to make corrections for particle loss in MOI tubes and bends. In some cases the efficiency of the largest stage was used as a variable parameter to force the corrected ion sums match those of a simultaneous TAS sample. I. The LTI efficiency, derived from field data and Fluent computational fluid dynamics modeling, by Chuck Wilson's group at DU. It also is weighted to account for large changes in the value across the size range of the largest stages. J. Average altitude of the sample, in meters above sea level. K. The apparent chloride concentration on each stage, in ng Cl-/sm3. L. The apparent nitrate concentration on each stage, in ng NO3-/sm3. M. The apparent sulfate concentration on each stage, in ng SO4-/sm3. N. The apparent oxalate concentration on each stage, in ng Ox-/sm3. O. The apparent sodium concentration on each stage, in ng Na+/sm3. P. The apparent ammonium concentration on each stage, in ng NH4+/sm3. Q. The apparent potassium concentration on each stage, in ng K+/sm3. R. The apparent magnesium concentration on each stage, in ng Mg++/sm3. S. The apparent soluble calcium concentration on each stage, in ng Ca++/sm3. T. The apparent non-sea salt sulfate (NSS) concentration on each stage, in ng SO4=/sm3. NSS was derived by subtracting the product of the seawater ratio of sulfate mass to sodium mass (0.251) and the observed sodium concentration from the observed sulfate concentration. If there was extractable sodium in the dust we sampled, this would cause our reported NSS to be an underestimate of true ambient NSS. U. The uncertainty in the apparent chloride concentration on each stage, in ng Cl-/sm3. V. The uncertainty in the apparent nitrate concentration on each stage, in ng NO3-/sm3. W. The uncertainty in the apparent sulfate concentration on each stage, in ng SO4-/sm3. X. The uncertainty in the apparent oxalate concentration on each stage, in ng Ox-/sm3. Y. The uncertainty in the apparent sodium concentration on each stage, in ng Na+/sm3. Z. The uncertainty in the apparent ammonium concentration on each stage, in ng NH4+/sm3. AA. The uncertainty in the apparent potassium concentration on each stage, in ng K+/sm3. AB. The uncertainty in the apparent magnesium concentration on each stage, in ng Mg++/sm3. AC. The uncertainty in the apparent soluble calcium concentration on each stage, in ng Ca++/sm3. AD. The uncertainty in the apparent non-sea salt sulfate (NSS) concentration on each stage, in ng SO4=/sm3. AE. The ?corrected? chloride concentration on each stage, in ng Cl-/sm3. A description of the corrections is AF. The ?corrected? nitrate concentration on each stage, in ng NO3-/sm3. AG. The ?corrected? sulfate concentration on each stage, in ng SO4-/sm3. AH. The ?corrected? oxalate concentration on each stage, in ng Ox-/sm3. AI. The ?corrected? sodium concentration on each stage, in ng Na+/sm3. AJ. The ?corrected? ammonium concentration on each stage, in ng NH4+/sm3. AK. The ?corrected? potassium concentration on each stage, in ng K+/sm3. AL. The ?corrected? magnesium concentration on each stage, in ng Mg++/sm3. AM. The ?corrected? soluble calcium concentration on each stage, in ng Ca++/sm3. AN. The ?corrected? non-sea salt sulfate (NSS) concentration on each stage, in ng SO4=/sm3. This is Version 2.0 of the Huebert group MOI data. 5.0 DATA REMARKS: Cascade impactors suffer from several possible artifacts. We have submitted two data sets, one which is not corrected for any sampling effects and one which is the best we can do at making those corrections. Each has serious limitations, but we believe each also can be very useful. The ?apparent? or uncorrected data contains easily-defended information about what ions are associated with what other ions. It was a surprise, for instance, that nitrate was so tightly associated with calcium, rather than ammonium. The ratios on the supermicron stages and submicron stages provide a defendable picture of the composition of particles in each size range, even though they do not necessarily represent ambient concentrations faithfully. The ?corrected? data should give a reasonable picture of the ambient concentrations of each ion in submicron and supermicron modes, even though it may misrepresent the shape and peak size of those modes. Our assumptions and our fitting to reproduce TAS concentrations involved some arbitrary choices about what sizes to allocate bounced material to or the proper size (in a range) that best represents the impact on any one stage. Yet the constraint of TAS and the clear separation of large and small particles (on all but the backup filter) means that the concentration of each ion in each mode is realistic, something that has not been possible in most previous aircraft sampling. Combining the corrected data with physical size distributions from the APS, OPC, and SMPS will provide an excellent starting point for estimating optical properties from chemical data and studies of the uptake of acidic gases by dust particles. Since the LTI enhancement is well-constrained, in future experiments one needs to minimize the amount of tubing between the LTI and the impactor stack to reliably represent all sizes of particles with an airborne cascade impactor. We will mount each impactor stack directly at the point where air enters the fuselage from the LTI. This seems to be the only way to avoid questions about the stickiness of particles and its impact on plumbing losses. 5.1 SOFTWARE COMPATABILITY: The file MOIHuebertSubmitOct02.xls is an Excel file which should open with most recent versions of Excel.