February 23, 1996 Read me file for particle size distribution data from the PALMS group. See especially the section on using PCA to define natural size ranges. File is called MSCOUNTR.ASC for ascii counter data from the Mass Spectrometer group. ---------------Purpose ------------------- These aerosol size distributions are intended for 2 purposes: 1) To know what was going into the PALMS instrument 2) To assess the 55 m inlet compared to the inlet on the building Please use the Covert/Heintzenberg/Gras/... aerosol size distributions for accurate ambient data. --------------- Instrument Description ------------------- Two commercial optical particle counters were run as an adjunct to the PALMS aerosol mass spectrometer instrument. They were: 1) A Climet Model 208A (0.33-9 æm, white light) counter with a custom all-digital pulse analyzer board with 10 bit resolution. 7 lpm flow rate. 2) A PM LasAir 1001 (0.1-2 æm, He-Ne laser) counter with 7 size channels. A Hiac/Royko 10::1 diluter was added upstream of the PMS. This has a Data were collected on 1 minute intervals, usually 57 s of data and 3 s of computer overhead. The smallest Climet sizes were only sampled every other minute to allow sampling the larger particles with an assurance that the dead time corrections were small. --------------- Sampling ------------------- Aerosols were collected from the 55 m capped inlet erected by Cape Grim personnel. A flow of at least 22 lpm was maintained at all times with a blower. The particle counters drew an additional 7 lpm almost 24 hours a day. The mass spectrometer drew an additional 1 lpm during most working hours. Filter samples by Cape Grim added another ??? lpm during baseline conditions after ???? date. A 1" inlet sampled from the core of the flow at the bottom of the 4" tower sample, approximately isokinetically. The combined blower , counter, and mass spectrometer flows were combined into the trailer through an insulated line. The blower flow was pulled off just inside the trailer. The PMS and Climet shared about 2 m of 3/8" sample line inside the trailer which then split about 20 cm before the counters into 1/4 and 1/4" lines Calculations based on Agarwal and Liu show that the aerosol transmission through the entire inlet should be excellent for particles < 0.1 æm diameter to somewhere in the 3-10 æm diameter range. Above that, turbulent deposition in the 55 m tube should deplete the particles. --------------- Calibrations ------------------- Very limited calibrations were performed with 0.2 and 1.0 æm latex spheres, and NaCl particles selected by a DMA to be 0.17, 0.25, 0.4, and 0.55 æm. Based on these calibrations (mostly the NaCl), the factory calibration curve for the Climet was shifted by 9% to smaller sizes and the PMS curve shifted 15% to larger sizes. For example, the PMS 0.1-0.2 æm bin was shifted to 0.115-0.23 æm. Note that the calibrations weren't extensive enough to test the calibration bin by bin. The PMS flow was based solely on the factory calibration of its internal flow meter. The Climet flow was set with an Omega rising ball gauge - not highly accurate. In addition, the Climet front panel gauge (which is just a 0-100% scale) was almost a factor of 2 different than a flow calibration 2 years ago. The PMS and Climet overlap in the 0.33- 2 æm size range, offering an opportunity to check them against each other. They tracked short term variations extremely well. However, there were slow drifts of up to a factor of 2 on a day to day basis. The diluter in front of the PMS was a nominal 10::1 diluter with a factory calibration of 11::1. We calibrated it in 1993 as 9.6::1, and used this number for the Cape Grim data. --------------- Data Processing ------------------- Dead time corrections were made to all data. They were very small (generally < 5%) at Cape Grim. The corrections for these counters are highly reliable from previous work we have done in Colorado. The Climet data were reduced from the 10 bit pulse height analysis to 3 size ranges: 0.33-1.5 æm, 1.5-3.85 æm, and > 3.85 æm. These ranges were chosen objectively. First, the Climet data were collected into 63 size bins. Then the time series of these size bins were subjected to principal components analysis, which identifies which groups of data are most highly correlated with each other. There were clear breaks in correlation at about 1.5 and 3.85 æm . For example, the 5 and 6 æm aerosols were highly correlated with each other but barely correlated with the 3 æm aerosols. There was another, much weaker, break somewhere between 0.33 and 0.6 æm, depending on how the analysis was done (which days, etc). This was set at 0.33 æm to maximize the size range covered on the days with no PMS data (see below). These size ranges correspond to modes visible in the size distributions. The PMS data were integrated from their lower limit up to 0.33 æm. That is not an exact channel break for the PMS data, so integral was made by integrating a spline fit to the log of the accumulated (not differential) size distribution. This is a small correction to just summing the bottom two PMS size bins. Surface and volume calculations were made using the midpoint of each original size bin. Since the Climet size bins are so narrow, there is little error introduced for those data. There is some error introduced by using the midpoint of the PMS bins for surface and volume calculations. Surface data in the exchange file are corrected for transport according to Wagner (in Aerosol Microphysics II, ed. Marlow). This is entirely equivalent to Fuchs-Stutigan coefficients. Calculations are for a molecule of molecular weight 108 reacting with a reaction coefficient of 0.1 (like N2O5 on a wet aerosol). The factor of 0.1 is only taken care of in the higher order terms: the aerosol surface area still needs to be multiplied by the reaction probability even if that probability is 0.1. Data were averaged into 5 minute periods. Any step functions in atmospheric concentration will be smeared out during transport down the 55 m inlet, so there is little point in faster data. Note that the 55 m inlet induces a time delay of order 5 minutes. The size range > 3.85 æm has poor statistics at 5 minutes but it also seems to have real short term variability on very short time scales. --------------- Data Continuity ------------------- A plumbing connection to the PMS failed on December 3 and was not discovered until data analysis. Therefore, there are no PMS data after that date. Otherwise, the data are nearly continuous from 951118 to 951214 except for short periods to back up the hard disk and a couple of overnight outages.