The primary AOP measurements made by the National Oceanic and Atmospheric Administration Climate Monitoring and Diagnostics Laboratory (NOAA/CMDL) during the INDOEX IFP were (1) the aerosol light scattering coefficient (σsp) at 3 visible wavelengths and at low (typically <45%) and high (usually 80–85%) relative humidity (RH), (2) the aerosol hemispheric backscattering coefficient (σbsp) at the same wavelengths and RHs, and (3) the aerosol light absorption coefficient at a low RH and at a wavelength of 550 nm. The aerosol instruments that comprised the NOAA/CMDL airborne aerosol measurement system to measure these quantities consisted of two integrating nephelometers (TSI Model 3563, St. Paul, MN, USA) connected in series and separated by a humidity control system, and a filter-based light absorption photometer (Radiance Research Model PSAP, Seattle, WA, USA). Figure 1 shows a schematic of the aerosol measurement system deployed on the NCAR C-130 aircraft. Aerosols were sampled from the C-130 community aerosol inlet (CAI) and traveled at a flow rate of ∼30 L min−1 into our main sampling line. It is important to note that in this paper we are reporting the properties of aerosol particles that have passed through an aircraft-mounted, high-speed aerosol inlet, and passage through the inlet may have modified these properties in some way. Any sampling bias or sampling artifact in the CAI could influence our reported aerosol results. The sampled aerosols then passed through a drying tube followed by a line heater that were used to lower the sample line RH to ∼40%. If the ambient RH was below 40%, the drying tube and heater would have little effect, because the drier would not remove much additional moisture and the heater would not be active. Following the sample line heater was a sealed and insulated impactor box, which contained a switched dual-impactor system that permitted us to change the cut size of our sampled aerosols when desired from particles smaller than 10 μm aerodynamic diameter (Dp < 10 μm) to particles smaller than 1 μm aerodynamic diameter (Dp < 1 μm). Most of the airborne measurements (∼90%) during INDOEX were conducted with a Dp < 1 μm size cut, because the upper-limit size estimate of particles passed by the CAI and the uncertainty in this estimate were unknown at the time and not quantified until a later study [Blomquist et al., 2001]. Rather than measuring aerosols with an unknown (and potentially broad and variable) upper size limit, we decided that it would be better to eliminate that with a firm and reproducible 1 μm (aerodynamic) size limit. For this reason, we focused most of our sampling efforts on the submicrometer particle fraction, which was thought to have a high passing efficiency through the inlet. Based on the results of the study by Blomquist et al.  which are discussed below, particles larger than ∼3 μm diameter were not transmitted efficiently through the CAI. Therefore, we have redefined our Dp < 10 μm size fraction to be Dp < 3 μm, with the caveat that any errors in the determination of the CAI cutpoint by Blomquist et al.  would change our upper particle size limit.
Figure 1. Schematic of the NOAA/CMDL aerosol measurement system onboard the NCAR C-130 aircraft. All tubing in the system, with the exception of the low-flow pickoff line to the PSAP instrument, was nominal 1.59-cm (inner diameter) stainless steel or flexible conductive tubing.
Download figure to PowerPoint
 The separate reporting of the properties of the fine aerosol fraction is a common practice and reflects the distinct sources, sinks, and atmospheric lifetimes of these smaller particles [Seinfeld and Pandis, 1998]. One pre-INDOEX study was performed in 1996 that measured atmospheric size distributions during the winter monsoon season over the Indian Ocean and Arabian Sea [Jayaraman et al., 1998]. This study showed average mass-size distributions with minima in the 1–2 μm diameter range, which suggests that the submicrometer aerosol mode of INDOEX aerosols should be reasonably well resolved.
 The aerosols passed from the impactors into the reference nephelometer, which always measured aerosol light scattering at the lower RH. This RH was typically below 40%, but because of the varying cabin temperature in the aircraft, the RH in the reference nephelometer occasionally increased to >45%. Upon exiting the reference nephelometer, the sample passed through a humidifier that was set to maintain a much higher RH, usually in the 80–85% range. Aerosols exiting this humidity control unit then entered the second nephelometer where the humidified aerosol scattering measurements were made. Just upstream from the inlet of the reference nephelometer was a pickoff for the PSAP instrument that sampled aerosols at a flow rate of 1–2 L min−1.
 In Table 1, detailed descriptions of all primary and derived measurements are presented. Formulas used to calculate the hemispheric backscatter fraction (b), Ångström exponent (å), single-scattering albedo (ω0), and aerosol hygroscopic growth factor (f(RH)) are shown. Of these, by far the most complicated calculation is that for f(RH). The calculation of f(RH) requires that the reference and humidified scattering coefficients (which are typically not exactly at 40% and 85% RHs) be adjusted to their respective values at 40% and 85% RH. The adjustment of the scattering coefficient to that at a different RH is done using a multiple step process. First, the hygroscopic fit parameter γ is determined using a nonlinear least squares fit of the scattering coefficients from both the reference and humidified nephelometers (σsp (dry) and σsp (wet), respectively) over the averaging period of interest
where RH(wet) and RH(dry) are the relative humidities measured inside both nephelometers and the exponent γ depends on the hygroscopic nature of the aerosol [Kasten, 1969]. A similar strategy has been used in other recent airborne humidified nephelometry studies [e.g., Gassó et al., 2000]. Humidity-scanning nephelometer measurements conducted by our group during the INDOEX IFP [Clarke et al., 2002] at the Kaashidhoo Climate Observatory (KCO) (J. Lobert et al., Kaashidhoo Climate Observatory (KCO): A new site for observing long-term changes in the tropical Indian Ocean, submitted to Journal of Geophysical Research, June 2000) on the Maldivian island of Kaashidhoo were used to confirm the functional form of this equation as opposed to more complicated multiparameter fits [e.g., Kotchenruther and Hobbs, 1998]. The assumption has been made implicitly here that the hygroscopic nature of the aerosols measured by the humidograph system at KCO [Clarke et al., 2002] was similar to that for aerosols measured on the C-130 aircraft. Once γ has been determined, σsp at any other RH (σsp(RH)) and f(RH) can be determined using the similar equations
Determination of the correct scattering coefficient at an RH other than that measured in the reference or humidified nephelometer using the 2-point statistical fit method of equations (1) and (2) requires that the scattering varies smoothly with RH between RH(wet) and RH(dry). Fortunately, the polluted marine aerosols measured at KCO showed scattering to be a smooth function of RH (<2% of all nephelometer humidity scans displayed potential evidence of deliquescence). For aerosols such as these, σsp can be retrieved via the 2-point fit method with little error. A typical mean value for γ, this for all (N = 74) low altitude (0–1 km) level flight segments over the northern Indian Ocean, is ∼0.33, with a standard deviation of 0.10. This compares well with mean γ value of 0.368 found for our measurements at KCO during the INDOEX IFP [Clarke et al., 2002].
Table 1. In Situ Aerosol Optical Property Measurements and Instruments Onboard the NCAR C-130 During the INDOEX IFP
|Instrument||Primary measurements||Derived measurements|
|TSI Model 3563 three-wavelength, backscatter/total scatter integrating nephelometers, operated at both low (<40%) and high (∼80–85%) relative humidity||Total scattering and hemispheric backscattering coefficients from particles (σsp and σbsp) at 450, 550, and 700 nm wavelength||Hemispheric backscatter fraction, b = σbsp/σsp Ångström exponents, å = −log[σsp(λ1)/σsp(λ2)]/log[λ1/λ2] Single scattering albedo (550 nm), ω0 = σsp/(σsp+σap) Hygroscopic growth factor, f(RH) = σsp (RH=85%)/σsp (RH=40%)|
|Radiance Research Model PSAP particulate light absorption photometer||Light absorption coefficients from particles, σap, at 550 nma||Single scattering albedo, ω0, at 550 nm|
2.2. Data Analysis
 Aerosol optical property data were collected at 1-s resolution and averaged over appropriate time periods. Prior to averaging, the data were quality-checked and edited to remove spikes caused by electronic glitches, PSAP filter changes, system pressure changes, etc. Light absorption data from the PSAP often contain spikes, especially during ascents and descents. These are thought to occur because of flexing and/or settling of the filter substrates due to the internal system pressure changes experienced through changing altitudes. These spikes are often quite large, are relatively easy to identify, and are of short duration (typically <10 s duration), so that few of the vertical profile data are affected.
 Other edits and corrections were applied to the raw data from INDOEX. A sample line heater that malfunctioned at low altitude (<1000 m above sea level, asl) for part of the project caused some particle volatilization in our system. All time periods when this heater was active were flagged and the data rigorously checked. If the volatilization artifact was found, these questionable data were removed from the data set. Dilution corrections were applied when appropriate to correct for the presence of several small leaks in our system that were each present for part of the project. The leaks for the most part showed zero particle penetration efficiency, so a straight dilution correction was deemed appropriate. These heater edits and leak corrections are discussed in more detail in section 2.3, and a complete presentation of them can be found on the NOAA/CMDL website (contact authors for current website address). An empirically derived correction to the humidified nephelometer scattering data was applied to account for particle losses in transit through the humidity control system. The light scattering data from both nephelometers were initially adjusted to conditions of standard temperature and pressure for comparison with the light absorption measurements (which use that reference), and also for comparison with scattering measurements at other locations and altitudes. Both the light scattering and absorption coefficients were also adjusted to conditions of temperature, pressure, and (for scattering) RH that were representative of the ambient atmosphere. Light absorption data from the PSAP instrument were corrected for filter spot size differences from the calibration standard and scattering artifacts using the calibration methods of Bond et al. . Additionally, PSAP data were removed from the data set if the filter transmittance dropped below 0.5, and they were left in but flagged if the transmittance dropped below 0.7. Finally, the TSI nephelometer data were corrected for angular nonidealities including truncation effects as detailed by Anderson and Ogren , who used a procedure similar to that proposed by Rosen et al. .
 For each vertical profile, processed 1-s data were sorted into bins that represented each 100 m of altitude. The vertical profile plots, therefore, show one data point for each parameter every 100 m in the vertical, except where data were missing or edited from the data set. Typical ascent and descent rates for the C-130 during INDOEX were ∼150–300 m min−1, so the 100-m bin averages represent 20–40 s of data.
 A slightly different data processing strategy was employed for the processing of horizontal flight segments. Much of the data presented herein are from level flight segments conducted at various locations and altitudes. Segment averages were generated for aerosol properties that reflect the duration of these flight segments. Level flight segments were not included in the compilations if they lasted <5 min. For higher altitude horizontal flight segments, segment averages were typically computed for segments >15 min in duration. This was not only because most higher-altitude flight segments were >15 min in duration, but also because the very low aerosol concentrations observed on some of these segments required averaging over longer periods to increase the signal-to-noise ratio in both the σsp and σap measurements to acceptable levels.
2.3. Discussion of Measurement Uncertainties
 Uncertainties in the measurements of aerosol optical properties onboard the C-130 aircraft can be broken down into three major categories: those related to getting the ambient aerosols into our instruments, those related to normal instrument operation and those related to instrument sampling or operational problems. For the first category, the major uncertainties are involved with transporting aerosol particles through the Community Aerosol Inlet (CAI) on the C-130 and into our aerosol instruments. The CAI was designed as a shrouded, multiple-diffuser inlet with a boundary layer suction vent located just behind each diffuser. Its internal flow characteristics are not completely laminar, and turbulence intensities at the sampling plane of 7–9% have been noted in the recent CAI evaluation study by Blomquist et al. . This study also concluded that the 50% cut size for particles in the CAI was near 3 μm diameter, but that the size cut was not a sharp one. A small fraction of particles larger than 5 μm in diameter were likely to pass through the inlet, while a few particles smaller than 1 μm diameter probably would be unable to pass. This is why we have estimated the upper size limit of particles reaching our instruments as 3 μm, even though we sampled through a 10-μm impactor.
 Ram heating also occurs to a small extent, although the relatively low turbulence intensities and attention to isokinetic flow considerations minimize this concern. Given these findings, we expect that for submicrometer particles (which were exclusively sampled by our instruments ∼90% of the time through the use of the switched impactor system), the inlet loss effects are thought to be minimal (i.e., no more than a few percent) and the uncertainty in those measurements due to that effect very small. A thorough analysis of the wing probe size distribution data during the INDOEX IFP is necessary before we can determine the CAI effect on the uncertainty associated with our Dp < 3 μm diameter aerosol data. We did not use the wing tip optical particle counter data to attempt to determine the fraction of extinction caused by the unsampled supermicrometer particles. The sizing accuracy of optical particle probes in the supermicrometer range is subject to large uncertainties in the absence of reliable information on particle composition. As the aerosols measured in INDOEX have been shown to be highly variable in composition (and hence their refractive index) and contained significant fractions of light absorbing particles such as carbon soot (J. R. Anderson, P. Crozier, and S. Howell, Electron microscopy of sulfate particles north and south of the ITCZ during INDOEX, submitted to Journal of Geophysical Research, personal communication, 2000), the large uncertainties in particle size derived from these instruments minimizes the usefulness of these data for anything other than qualitative assessments of supermicrometer concentrations (D. Baumgardner, personal communication, 2000). As discussed below, however, several independent measurements suggest that light scattering by supermicrometer particles over the Indian Ocean was typically less than that from submicrometer particles.
 For our nephelometers under normal operating conditions, total measurement uncertainties can be calculated from the individual uncertainty components associated with instrument accuracy, calibrations and corrections, and adjustment of instrument RH to ambient RH. Calculation of the total measurement uncertainty associated with the nephelometers can be calculated from the major sources and is expressed as a combination of the following terms
where δσp designates the uncertainty in σsp associated with the parameter p. For parameters where a percentage number is given for the uncertainty, these represent either 95% confidence intervals of the uncertainty data or commonly accepted uncertainties (e.g., the Rayleigh correction uncertainty). These individual uncertainties are instrument-specific and represent (1) instrument noise (∼10% at σsp = 1 Mm−1 and ∼0.4% at σsp = 50 Mm−1 for 10-min averaging times), (2) drift in the calibrations based on repeated measurements of calibration gases (∼3%), (3) uncertainty in the calibrations due to uncertainties in the measured Rayleigh scattering of air and CO2 (7%), and (4) uncertainty in the truncation or blocking of near-forward scattered light (∼2%), and (5) the uncertainty associated with adjusting σsp to standard temperature and pressure (<1%).Table 2 shows the approximate uncertainties and the calculation method used for each component of the total uncertainty for various magnitudes of σsp. A 10-min averaging time was used in the noise calculation because these uncertainties apply to all level C-130 flight segments in INDOEX, and nearly all of these exceeded 10 min in duration. Therefore, the noise uncertainty reported in Table 2 is actually an upper limit to that experienced during most of the level flight segments. To estimate the noise uncertainty component at a 1-min averaging period (more appropriate for vertical profiles), multiply the noise uncertainty in Table 2 by 3.16.
Table 2. Estimated Uncertainties in σsp at 550 nm for 10-min Averaging Times and Submicrometer Particles (Mm−1)
|σsp (550 nm)||Noisea||Driftb||Calibrationc||Truncationc||STPd||Totale|
 The angular sensitivity uncertainty δσtrunc includes corrections for both the truncation of forward-scattered light and for the slightly non-Lambertian distribution of illumination intensity. The original angular sensitivity measurements made on this nephelometer were made at 550 nm (T. Anderson, personal communication, 2000), and recent angular sensitivity measurements at 700 nm showed excellent agreement with the earlier data (N. Ahlquist, personal communication, 2000). Little or no wavelength dependence for the non-Lambertian effect is expected in a fairly thick, multiple scattering environment like the TSI opal glass diffuser, and assumption is corroborated by the closure tests performed by Anderson et al. , where the difference between measured and modeled scattering was within 7% and showed essentially no wavelength dependence. In any case, the angular sensitivity uncertainty is small for submicrometer particles, which were predominantly sampled by our system during INDOEX.
 An additional uncertainty is present when adjusting values of σsp at a given RH to those appropriate at another RH. This uncertainty depends primarily on the magnitude of the RH to which the measurement is being adjusted. Scattering measurements taken at a typically low instrument RH and adjusted to another low RH (≲50%) show very small additional uncertainties, usually a few percent or less. This is because the nonlinear fits are quite good at lower RH where the scattering coefficient is not changing rapidly. Adjustment of σsp to higher RHs (e.g., 80–90%) shows uncertainties several times larger than for the adjustment to lower RHs. For low RH conditions (<50%) and moderate aerosol levels (σsp = ∼50 Mm−1), total measurement uncertainties in σsp (including the RH adjustment) of 9% were realized for 10-min average data, while for ambient humidities near 90% the uncertainty increased to ∼14%.
 For the light absorption measurement, total uncertainties are calculated by combining the major individual uncertainties associated with instrument accuracy, instrument precision, instrument noise, and adjustment of the detected absorption signal to account for a contribution from light scattering. Uncertainty in the PSAP-derived σap results from uncertainty in the following components of the measurement [Bond et al., 1999]: (1) instrument accuracy (estimated at ∼20%); (2) instrument precision (∼6%); (3) instrument noise (fixed at 0.88 Mm−1 and 0.28 Mm−1 for 1-min and 10-min averaging times, respectively); (4) uncertainty in the calibration that converts the wavelength to 550 nm and corrects for filter-based scattering that is sensed as absorption by the instrument (∼4%). Added in quadrature, these components yield total uncertainties of δσap ∼ 49% for σap = 2 Mm−1, δσap ∼ 28% for σap = 5 Mm−1, and δσap ∼ 23% for σap = 10 Mm−1 for 1-min average data. The propagated uncertainty in σap for 10-min averaging periods was ∼26% for σap = 2 Mm−1, ∼ 22% for σap = 5 Mm−1, and ∼21% for σap = 10 Mm−1. The uncertainties are rather large for our absorption measurements, which unfortunately are typical of the current state of the art in filter-based aerosol light absorption measurements.
 The uncertainties reported below for the derived parameters ω0, b, and å were calculated based on 10-min averaging periods using the methods described by Anderson et al. [1996, 1999] and Anderson and Ogren , from the uncertainties in the appropriate σsp, σbsp, and σap measurements. A typical uncertainty for each parameter is provided below for a high-scattering case (e.g., low altitude segments north of the equator) and a low-scattering case (high altitude segments or those in the Southern Hemisphere). The specific flight segments we used to obtain the scattering and absorption values for the high-scattering uncertainty calculations are the lowest-altitude segments conducted north of 5° north latitude (0–1 km altitude, Dp < 1 μm). The flight segments used for the low-scattering uncertainty determinations were the lowest-altitude segments conducted south of 5° south latitude (0–1 km altitude, Dp < 1 μm). For the high-scattering case (σsp = 58 Mm−1), the absolute uncertainties in ω0, b, and å were 0.036, 0.012, and 0.49, respectively. For the low-scattering case (σsp = ∼6 Mm−1), the uncertainties in ω0, b, and å were 0.049, 0.018, and 0.53, respectively. These values compare with the uncertainties estimated by Anderson et al.  for ω0, b, and å during a polluted marine aerosol episode (with a mean σsp value of 18 Mm−1) of approximately 0.024, 0.018, and 0.33, respectively.
 The uncertainty in f(RH) was calculated by computing the increased uncertainties in the scattering coefficients from the reference and humidified nephelometer after adjustment to 40% and 85% RH, respectively. For the case of low-altitude flight segments conducted north of 5° north latitude (0–1 km altitude, Dp < 1 μm), the relative uncertainty in our f(RH) parameter was ∼21%.
 During the INDOEX IFP, two operational problems contributed at various times to the measurement uncertainty in the light scattering and absorption measurements. One of these problems was that the sample intake line heater was not being controlled properly and stayed on longer than it should have, resulting in higher temperatures than planned in the intake line. The malfunctioning heater caused a periodic volatilization of the aerosol particles and loss of aerosol mass. This malfunction caused pronounced drops in σsp of up to 40% that were not evident when the heater was off. This heater was only activated when sample line RH was >40%. Given that a drying tube upstream of the sample line heater removed some of the moisture from the air stream, the heater only came on when flying through the highest RH air. This only occurred (1) in clouds, and (2) at times in the moist marine boundary layer (MBL) below 1000 m asl. Cloud data have been screened from this data set, so they have not been reported. Based on comparison of the CMDL nephelometer signal with concurrent condensation nucleus and 180°-nephelometer measurements, data periods influenced by the heater malfunction were identified. When the aircraft was flying below 1000 m asl, the condition that allowed the malfunction to occur was found to be present ∼55% of the time. These data are considered unrecoverable, because the particle size distribution, chemistry, and thermal gradient in the sample line are not known accurately enough to correct for the volatility losses of aerosol species. Therefore, all data from the identified malfunction periods were removed. Thus, no additional uncertainty in the aerosol optical property measurements reported in this paper was realized through the heater malfunction.
 Over the course of the INDOEX IFP, several leaks in the CMDL aerosol system were identified and corrected. A detailed discussion of the all leak problems, tests results that quantify the individual leaks, and corrections applied to the data are available on the NOAA/CMDL web site. The air leaks included a leak in the humidified nephelometer, a leak in the sample inlet line, a connector ferrule leak, and a leak in the impactor box. The humidified nephelometer leak was discovered on Flight 12 (13 March 1999) and immediately fixed; the duration of the leak could have been from Flights 1–12 (16 February–13 March 1999). The sample inlet line leak was also discovered on Flight 12. This was in a part of the line that was frequently disassembled and reassembled. We cannot rule out the possibility that this leak was present on Flights 1–12. The connector ferrule leak was only present on Flight 15 (19 March 1999), and the impactor box leak was only present on Flight 16 (21 March 1999).
 Leaks of cabin air into our system were present whenever the cabin pressure exceeded the pressure inside our instrument. The cabin pressure above 1800 m was maintained around 830 mbar, although not consistently. The consequence of a leak on the scattering signal was that the sample air was diluted with cabin air and potentially contaminated with cabin aerosol. A formula that quantifies this correction is:
Here Sa is the ambient signal, Sm is the measured signal, Sc is the cabin air signal, X is the leak rate and E is the aerosol penetration efficiency. Post flight tests of the leak rate and aerosol penetration efficiency as a function of pressure were performed to determine X and E for each of the known leaks. Cabin aerosol scattering signals were serendipitously measured throughout the campaign during in-flight PSAP filter changes, which allowed cabin air to flood the nephelometers.
 Leaks had little impact on the nephelometer signal above ∼3500 m altitude, as both the cabin and ambient aerosol scattering signals were near the instrument detection limit. The altitude range of maximum uncertainty in the contribution of leaks on the nephelometer signals usually fell between 2 and 3 km, when the cabin pressure exceeded the instrument pressure by more than 50 mbar. Fortunately, particles in the size range detected by the nephelometers were only able to penetrate through one of the leaks (the connector ferrule leak), which was present on only one flight (Flight 15), so that the primary effect of the leaks was a 1–10% dilution of the sample with filtered cabin air. All scattering and absorption data were corrected for dilution by filtered air from each leak as a function of the cabin/instrument pressure differential. Uncertainties in σsp and σap associated with errors in the measurement of the leak rate or in the measurement of particle penetration efficiency through the leaks are difficult to estimate, but probably constitute another 5–10% uncertainty for scattering and absorption data for all flights except Flight 15. For Flight 15, where the particle penetration was significant through the ferrule leak, uncertainties are estimated to be in the 10–15% range. Combining these leak uncertainty estimates with the other uncertainty components (added in quadrature) increases the total uncertainty in σsp presented above for 10-min averaging times to ∼10% and ∼14% for low and high RH measurements, respectively. The σap total measurement uncertainty for 10-min average data increased to ∼23% for σap = 5 Mm−1 when accounting for the effect of leaks.