Airborne Characterization of Smoke Marker Ratios from Prescribed Burning

A Particle-into-Liquid Sampler Total Organic and system was flown 47 aboard a in to obtain smoke marker measurements. The fraction collector provided 2 min 49 time-integrated off-line samples for carbohydrate (i.e., smoke markers levoglucosan, mannosan, 50 galactosan) analysis by high-performance anion-exchange chromatography with pulsed 51 amperometric detection. Each fire location appeared to have a unique  levoglucosan/  water- 52 soluble organic carbon (WSOC) ratio (RF01/RF02/RF03/RF05 = 0.163 ± 0.007  g C/  g C, 53 RF08 = 0.115 ± 0.011  g C/  g C, RF09A = 0.072 ± 0.028  g C/  g C, RF09B = 0.042 ± 0.008 54  g C/  g C where RF means research flight). These ratios were comparable to those obtained 55 from controlled laboratory burns and suggested that the emissions sampled during 56 RF01/RF02/RF03/RF05 were dominated by the burning of grasses, RF08 by leaves, RF09A by 57 needles, and RF09B by marsh grasses. These findings were further supported by the 58  galactosan/  levoglucosan ratios (RF01/RF02/RF03/RF05 = 0.067 ± 0.004  g/  g, RF08 = 59 0.085 ± 0.009  g/  g, RF09A = 0.101 ± 0.029  g/  g) obtained as well as by the ground-based 60 fuel and filter sample analyses during RF01/RF02/RF03/RF05. Differences between 61  potassium/  levoglucosan ratios obtained for these prescribed fires vs. laboratory-scale 62 measurements suggest that some laboratory burns may not accurately represent potassium 63 emissions from prescribed burns. The  levoglucosan/  WSOC ratio had no clear dependence on 64 smoke


MT 59808
3 1. Introduction 91 The smoke marker approach is the most common method used to estimate the 92 contribution of primary biomass burning to the total organic carbon aerosol concentration (e.g.,

5
The TOC Analyzer used was a Sievers Model 800 Turbo TOC Analyzer. This 181 instrument converts the organic carbon in a liquid sample to carbon dioxide through chemical 182 oxidation involving ultraviolet light and ammonium persulfate and quantifies the conductivity of 183 the produced carbon dioxide. The amount of organic carbon in the sample is proportional to this 184 measured increase in conductivity. The analyzer was run in Turbo mode providing a 3 s time-185 integrated measurement of WSOC with a limit of detection (LOD) of 0.1 g C/m 3 . 186 A Foxy 200 Fraction Collector (Teledyne ISCO) was used to collect the samples for off-187 line analysis. It can hold up to two hundred 16 mL uncapped polystyrene test tubes (Becton 188 Dickinson Labware). Test tubes were used as supplied by the manufacturer and required no 189 precleaning before use. The fraction collector program, which was manually started at take-off,

196
Each fraction collector test tube was brought back to room temperature before analysis.

197
Two 600 L aliquots were transferred to separate polypropylene vials for carbohydrate and 198 cation measurements. 199 The carbohydrate analysis was performed on a Dionex DX-500 series ion chromatograph 200 with an ED-50 electrochemical detector operating in integrating amperometric mode using 201 waveform A and a GP-50 gradient pump. The detector contains an ED-50/ED-50A 202 electrochemical cell. This cell includes a pH-Ag/AgCl (silver/silver chloride) reference 203 electrode and "standard" gold working electrode. The separation was performed using a Dionex 204 CarboPac PA-1 column (4 x 250 mm) employing a sodium hydroxide gradient. The complete 205 run time was 59 min and an injection volume of 50 L was used. More details of the method can 206 be found in Sullivan et al. [2011a,b]. Of the carbohydrates that can be detected by this method, 207 levoglucosan, mannosan, and galactosan were found in all samples. Glucose and arabinose were 208 only occasionally detected and will not be discussed further. The limit of detection (LOD) for 209 the various carbohydrates was calculated to be less than approximately 0.10 ng/m 3 . 210 Water-soluble potassium was measured using a second Dionex DX-500 ion 211 chromatograph. This system included an isocratic pump, self-regenerating cation SRS-ULTRA 212 suppressor, and conductivity detector. A Dionex IonPac CS12A analytical column (3 x150 mm) 213 using 20 mM methanesulfonic acid at a flowrate of 0.5 ml/min was used for the separation. The 214 injection volume and analysis time were 50 L and 17 minutes, respectively. Unlike for the 215 carbohydrates, a blank correction was necessary for the water-soluble potassium. Concentrations 216 were corrected by using the average of all particle-free background samples (i.e., with the 217 actuated valve before the PILS in the closed position) collected during a specific flight. The 218 LOD for water-soluble potassium was 0.02 g/m 3 .

221
In the analysis presented in this paper we focus on characterizing source smoke marker 222 ratios from prescribed burning. Other measurements presented here include 3-D location and Fourier transform infrared spectrometer) data analysis products including modified combustion 229 efficiency (MCE) ratios [Yokelson et al., 1999;Burling et al., 2011;Akagi et al., 2013].

230
All aircraft aerosol instruments sampled from a LTI (Low Turbulence Inlet) [Wilson et 231 al., 2004]. Following the LTI was a nonrotating MOUDI impactor with a 50% transmission 232 efficiency at 1 m and 1 atmosphere ambient pressure [Marple et al., 1991]. The combined flow 233 through the inlet and MOUDI was approximately 20 LPM and was then split to the individual 234 instruments. and out of the smoke plume. In order to take a closer look at the levoglucosan data, the WSOC 242 concentrations can be averaged to match the fraction collector times. It can be seen that the 2 243 min resolution of the fraction collector does capture the plume penetrations ( Figure 3a). In 244 addition, the ratio between levoglucosan and WSOC appears to be fairly constant (R 2 = 0.93 for 245 all data), which will be discussed in more detail in the next section. A times series for the 246 absolute concentrations for all three anhydrosugars measured from the fraction collector samples 247 can be seen in Figure 3b for this same flight (RF01). As is typically observed, levoglucosan 248 dominated followed by mannosan then galactosan. All three species concentrations tracked each 249 other and were highly correlated (R 2 > 0.90).

252
In order to investigate smoke marker ratios, we considered only fraction collector 253 samples collected in the smoke plume. Given the longer integration time for the fraction 254 collector system, the fraction collector data set was filtered using the CO data. Only fraction 255 collector samples that directly overlapped with a CO plume penetration are considered. Given  The correlation of levoglucosan with WSOC is shown in Figure 4a. All the burns 266 occurring at Fort Jackson (RF01/RF02/RF03/RF05) appeared to have a similar ratio, based on 267 the slope of the linear correlation, of 0.163 ± 0.007 g C/g C. In addition, there appeared to be 268 no concentration dependence or significant altitude dependence to this ratio ( Figure 4b). The 269 smoke sampled during RF08 had lower levoglucosan/WSOC ratios (approximately 0.115 ± 7 0.011 g C/g C) than the ratios for the Fort Jackson burns. Interestingly, the fire sampled 271 during RF09 appeared to have two distinct groups of levoglucosan/WSOC ratios (denoted 272 RF09A and RF09B). Group B had only a few samples, so linear regression statistics were not 273 reliable. Therefore, throughout for RF09B the average ratio ± standard deviation was calculated.

274
The levoglucosan/WSOC ratio was 0.042 ± 0.008 g C/g C for RF09B, which was lower 275 than the ratio of 0.072 ± 0.028 g C/g C for RF09A, suggesting that the vegetation may have  The importance of fuel type combusted can be further illustrated by comparing the 282 airborne smoke marker ratios to those from typical biomass burning source samples collected 283 from controlled laboratory burns (0.149 ± 0.012 g C/g C for grasses, 0.095 ± 0.006 g C/g C 284 for leaves, 0.064 ± 0.008 g C/g C for needles, and 0.017 ± 0.014 g C/g C for marsh grasses, 285 Table 2). Similarities in smoke marker ratio values suggest that the Fort Jackson burns 286 (RF01/RF02/RF03/RF05) were dominated by the burning of grasses, RF08 by leaves, RF09A by 287 needles, and RF09B by marsh grasses. Ground-based sampling of the Fort Jackson burns 288 included fuel characterization [Yokelson et al., 2013], which indicated the interior environment 289 was a longleaf pine/wiregrass system. One Hi-Volume quartz filter sample was collected across 290 each burn at the Fort Jackson ground-based sampling site. Analysis of the filters provided an 291 average levoglucosan/WSOC ratio of 0.198 ± 0.001 g C/g C, which is on the higher end of the 292 range of ratios observed for RF01/RF02/RF03/RF05 (Figure 4a).

302
However, the mannosan tolevoglucosan ratios observed in the research flights do not 303 compare as well to ratios measured in controlled laboratory burns. The controlled laboratory 304 burn ratio for grasses of 0.051 ± 0.005 g/g is much lower than the ratio for 305 RF01/RF02/RF03/RF05 of 0.207 ± 0.004 g/g. The controlled laboratory burn ratio for leaves 306 of 0.027± 0.008 g/g is also much lower than the ratio for RF08 (0.174± 0.008 g/g). By 307 contrast, the mannosan/levoglucosan ratio for RF09A (0.169 ± 0.102 g/g) is less than the 308 controlled laboratory burn ratio for needles (0.249 ± 0.016 g/g).

309
Water-soluble potassium has long been used as an inorganic marker for biomass burning.

310
As can be seen in Figure 5c, the prescribed burn observations contain quite a bit of scatter in the 311 potassium to levoglucosan ratio, even for a particular burn location. For example, attempting 312 a linear fit to the data from the burns at Fort Jackson (RF01/RF02/RF03/RF05) yields a very low 313 R 2 value of 0.13. Although, there is somewhat of a better correlation for RF08 (R 2 = 0.41). Poor 314 8 correlation between potassium and levoglucosan concentrations in biomass burning smoke is not 315 surprising. The presence of a small amount of inorganic substances, such as potassium, in a fuel 316 can cause changes in the product yields of levoglucosan during cellulose pyrolysis, with 317 potassium suppressing the formation of levoglucosan [Radlein et al., 1991;Richards et al., 1991;318 Patwardhan et al., 2010;Eom et al., 2012]. In addition, potassium is predominately emitted 319 from the flaming phase of a fire, whereas levoglucosan is emitted across both smoldering and 320 flaming fire phases [Ward et al., 1991;Echalar et al., 1995;Lee et al., 2010]. Changes in the 321 mix of flaming and smoldering combustion in a laboratory or prescribed burn, therefore, can 322 readily yield large differences in the emitted abundances of potassium. plume dilution (which can influence gas-to-particle partitioning) and photochemical reactions, 332 but very little data quantitatively examines the impact (if any) of these processes on smoke 333 marker ratios. Since a smoke marker ratio is needed to apportion the contribution of biomass 334 burning this is important to investigate. But it is also important to note this impact would depend 335 on the rates of reaction of levoglucosan and WSOC, which are unknown and could be similar. 336 Figure 6a shows the levoglucosan/WSOC ratio as a function of time since emission.

337
Over the range of smoke plume ages (up to approximately 1.5 h), the observations give no clear 338 indication that the ratio changes across a fuel type or fire location in a consistent manner as the 339 plume ages. Low ratios in RF09B, for example, remain low, while higher initial ratios in RF03 340 remain high. These observations suggest that the levoglucosan/WSOC source smoke marker 341 ratio is stable for at least 1-1.5 h as the plume dilutes and ages. 342 We can also make use of a subset of observations from the AMS to look at plume