Measurements of condensation nuclei in the upper troposphere: Sampling issues noted during SUCCESS

This page summarizes part of a research effort, the results of which have been drafted for submission to Journal of Oceanic and Atmospheric Technology. The following individuals contributed to this research:

Paul J. DeMott, David C. Rogers, Sonia M. Kreidenweis, Yalei Chen, D. Eli Sherman

Department of Atmospheric Science, Colorado State University

Cynthia H. Twohy, William A. Cooper

National Center for Atmospheric Research

Charles A. Brock

Department of Engineering, University of Denver

Donald E. Hagen

Cloud and Aerosol Sciences Laboratory, University of Missouri-Rolla

Part of this research was sponsored by the NASA Atmospheric Effects of Aviation program. The interpretations of experiments are those of the authors and do not necessarily reflect the views of the (partial) sponsor.

Abstract

This paper describes condensation nuclei measurements made during the NASA SUCCESS (Subsonic aircraft: Contrail and Cloud Effects Special Study) experiment in 1996 and laboratory investigations of apparent sampling issues. The aircraft measurements were not specially coordinated. Different groups made measurements with different continuous flow condensation nuclei counters, with different goals in mind and from different inlet locations on the NASA DC-8 aircraft. Nevertheless, fair agreement between the measurements was achieved during the balance of the field effort. Reasons for periods of disagreement between different condensation nuclei measurements were investigated. Some differences are attributable to different sensitivities to particle size, while others relate to particle losses due to differences in the configuration of sampling systems. These losses were confirmed for some of the sampling systems in the laboratory. A sensitivity to the rate of aircraft pressure change was also identified in field data and was investigated in the laboratory. This sensitivity seems inherent to CN counters using a critical orifice for flow control and can be exacerbated for certain sampling locations on aircraft bodies.

Aircraft Measurements and Issues

Instruments

The following table and figure summarize the characteristics and sampling locations of the CN counters operated on the NASA DC-8 during SUCCESS.

Figure of locations on DC-8

Each investigator used different sampling procedures that were largely dictated by their investigative goals. These goals were not necessarily focused on CN measurement alone. For example, …[ADD]

Typical Measurements

The following figure shows an example of the type of agreement obtained between the CN counters in background tropospheric conditions and when an ultrafine aerosol mode was not present. This example is from a 2 hour period of sampling on April 21, 1996 over the southern Great Plains of the United States. A time series of CN concentration and aircraft pressure are shown. Open square symbols are the concentrations of particles between 20 and 500 nm measured by the UMR-MASS. Data at 78000 to 79000 s is a cloud penetration based on microphysical sensor data.

April 21, 1996

Sampling aircraft exhausts and contrails

Measurements made in aircraft exhaust trails demonstrated differences in the sensitivities of the different counters and sampling systems to small particles. It was presumed that these differences were related both to the inherent cut sizes of the CN counters and diffusional losses in sample lines. The example that follows shows measurements through T-39 aircraft exhaust trails (1.2 km behind the T-39). The trail locations are clearly indicated by regions of high NO gas concentration that coincide with CN peaks. Measurements from the Cooper/NCAR, CVI/NCAR, and CSU groups are shown. Flight altitude was 9.4 km and outside pressure was 286 mb.

April 18, 1996

May 3, 1996

Effect of aircraft maneuvers

An effect that was clearly noted in the aircraft measurements was an alteration in the efficiency of counting CN during maneuvers that altered aircraft angle of attack. The CSU and Cooper counters, sharing the same external inlet over the back edge of the port wing, were most adversely affected by such things as flight level changes and maneuvers performed for calibrating the DC-8 meteorological measurement system. The UMR CN measurements were also affected, but only small impacts were noted in the CVI CN record. Calibration maneuvers included sideslip (yaw), roll, and pitch. Some of these aircraft movements resulted in strong short-term pressure fluctuations. The figure that follows shows an example of "wiggle" maneuvers and the effect on CN during a flight over the Central and Upper Midwest of the United States on May 8, 1996. Data Acquisition and Display Software (DADS) pressure and CSU and Cooper/NCAR instrument pressure are shown in (a). The Cooper/NCAR pressure has been filtered (see text). Aircraft sideslip, angle of attack and roll are given in (b). CSU, UMR, and Cooper CN concentrations are given in (c). The UMR system was using a metal bellows pump to raise the pressure of sampled air on this date.

May 8, 1996 "wiggle" maneuvers

Sampling volatile particles

The CSU sampling system appeared to have difficulty transferring certain types of volatile particles for counting during SUCCESS. Clear occurrences of apparent losses of particles at sizes above the cutoff size for the TSI 3010 counter that were also volatile at temperatures above 175° C were noted for extended periods on two of nineteen missions. The following figure shows data from May 2, 1996, one of two days on which the failure of the CSU sampling system to detect small volatile particles was inferred. The DC-8 was flying near the tropopause, over the top of orographic wave clouds (vicinity of Boulder, CO) at the time of sampling. The CSU signal corresponds more closely to the UMR nonvolatile (heated) CN signal than to the UMR ultrafine signal. The correspondence of CSU CN values with UMR-MASS CN > 20 nm is also indicated. The CVI CN sampled only over short intervals, but agreed at all times with the UMR CN signal (e.g., around 76200 UT s).

May 2, 1996

Based on laboratory studies, also described here, the CSU system was not very efficient in transferring particles below 25 nm. The small size of ambient particles, their shrinkage within the CSU sampling lines due to heating of air brought into the DC-8, and the sensitivity of penetration losses to size and pressure in the CSU system (see next section) probably combined to create the peculiar sampling problem noted in the figure. The drying and shrinking of volatile particles may have been exacerbated in the CSU system because of the placement of an in-line dryer (Permapure) in the sample line upstream of CN measurements. The CSU CN concentrations exceeded the concentrations of particles above 20 nm on the large percentage of other occasions during SUCCESS.

Other sampling issues

The CVI CN counter (TSI 3760) measured the highest average concentrations of any of the CN counters over the balance of the time that it sampled in SUCCESS. This was not expected based on the data in Table 1. The CN concentrations detected by the TSI 3760 should have approached, but not much exceeded the concentrations detected by the ultrafine (TSI 3025) counters. However, the CVI CN counter had the shortest run of tubing following a special isokinetic sampling inlet. In addition, some other factors affected the efficiency of the ultrafine CN counters during parts of SUCCESS. In the case of Cooper’s counter, excessive losses were noted at lower pressures that may have resulted from degraded flow or degraded (electronic) detection of particle pulses (noted in laboratory). In the case of the UMR ultrafine counter, sensitivity to small particles was apparently degraded later in the field program. The onset of this loss of sensitivity appeared to coincide with the beginning of operation of a metal bellows pump used to pressurize the ambient air sample at the entry to the UMR-MASS system. There are ample examples of both of these sampling inefficiencies in the SUCCESS data archive after the date of May 7, 1996. An example is shown in the figure below. This figure shows data from the May 15, 1996 flight around cirrus located off the coast of California. The UMR ultrafine counter (with metal bellows pump) does not pick up the onset of high concentrations of small CN during the DC-8 ascent at 84500 s, following instead the decaying CSU CN signal. The CVI CN counter appears most efficient for sampling small CN on this day. The Cooper ultrafine instrument initially responded to the small CN burst but its efficiency quickly degraded as pressure fell below 400 mb. The UMR and CSU signals did not come back into agreement with the CVI CN concentrations until more than 1000 s later, while the flight pressure was still 200 mb. The Cooper/NCAR CN counter did not "recover" until the aircraft descended to pressure levels higher than 300 mb.

May 15, 1996

Laboratory Studies

During August of 1997, three of the four groups that measured CN on the DC-8 convened at the Cloud Simulation and Aerosol Laboratory at Colorado State University to conduct a series of experiments on particle losses and pressure effects in CN measurements. The groups involved were the CVI/NCAR group, Cooper /NCAR group and the CSU group. The counters involved were listed in Table 1. In addition, CSU provided a second TSI 3010 instrument for referencing CN concentrations without added tubing used during airborne sampling.

Experimental Setup

The experimental configuration is shown below. All of the counters sampled aerosol particles that had been collected in the CSU dynamic cloud chamber (DCC) (see, DeMott and Rogers, 1990, J. Atmos. Sci., 47, 1056. Extra sampling tubing that more or less reproduced some (CVI and CSU groups) of the configurations that existed inside the aircraft during SUCCESS were used in some experiments. Flow was provided by the same pumps used on the aircraft and was returned to the chamber. In this way, the counters could sample at the DCC pressure. The return flow to the DCC was passed through a charcoal fiter and did not mix directly with the sample aerosols.

Size-selected ammonium sulfate aerosol particles were generated by passing a polydisperse aerosol distribution, formed by atomization and drying, through a DMA (TSI 3071). When a concentration of about 200 cm³ was present in the chamber, the chamber was locked against room pressure and an expansion was begun. The expansion rate was approximately equivalent to a 3 m s^-1 ascent of air from a starting pressure of about 840 mb. In normal operation (see, for eample, DeMott and Rogers, 1990), the cloud chamber inner walls are force-cooled to match the adiabatic cooling caused by evacuation. The walls were not cooled in these experiments, so the expansion remained approximately isothermal (~20° C). Expansion was halted at regular intervals to make measurements at a single pressure and then conduct a small, but fast, negative pressure deflection of maximum change between 3 and 5 mb s^-1. Data were recorded at 10 Hz and analyzed at 1 Hz.

Experimental Results

The CSU sampling system on the DC-8 suffered from rather large diffusional losses at certain particle sizes. Theoretical losses were calculated, as will be described in the full article. The theoretical efficiency of particles penetrating through the full SUCCESS tubing to the CSU CN counter as a function of pressure and particle size is presented in part a) of the following figure. The penetration efficiency measured during the laboratory workshop is shown in part b). The form of the experimental penetration follows the theoretical calculations, but experimental losses are about 10% greater in most cases. Thus, the CSU system could have underestimated concentrations of 20 nm particles by up to 2 times in the pressure range of SUCCESS measurements. Losses of 15 nm could have easily exceeded a factor of 3. The assumption of a penetration factor of 0.72 used for archiving the SUCCESS data was a good approximation for losses when the ambient CN aerosol mode was between about 25 to 100 nm and the ambient pressure was between about 300 to 500 mb. Corrections for losses were probably overestimated or underestimated by a maximum of 10 to 15 percent at any pressure if aerosol particle sizes were predominately larger than 25 nm. The problematic size regime for the CSU system was between 12 (cutoff size) and 25 nm.

CSU system: particle penetration efficiency

A critical finding in the laboratory was of CN measurement errors during pressure excursions. The following figure shows a segment of an experiment in which a pressure was first sharply reduced in the DCC, then the expansion was halted, and then the expansion was continued at the standard ascent rate. The experiment used 35 nm ammonium sulfate particles. Data from all of the counters are plotted. These results show that a continuous pressure reduction of any magnitude leads to a steady-state reduction (error) in the measured CN concentration. The magnitude of the error depends on the CN counter. The TSI 3025A was most sensitive to pressure changes in our study, followed by the TSI 3010 counters. The TSI 3760 counter was least sensitive to large dp/dt.

CN sensitivity to dp/dt

The magnitude of the CN error during pressure reduction was also a function of the magnitude of dp/dt and the starting pressure. This is shown below for the TSI 3010 counters belonging to the CSU group. The ratio of the change in CN concentration (initial value minus the "well" value) over the initial CN concentration is plotted versus the maximum 1 Hz dp/dt. The legend indicates the approximate pressure (± 30 mb) at which the change in pressure and CN response was measured. The open data points were from experiments using a stronger airflow pump. The pump capacity was not a factor.

Scaling of CN error with pressure and dp/dt (CSU system)

The CVI group (TSI 3760) CN counter was much less sensitive to rates of pressure change (open symbols), while the Cooper ultrafine (TSI 3025A) counter (filled symbols) was even more sensitive than the 3010 units.

Scaling of CN error with pressure and dp/dt (CVI and Cooper CN systems)

The following figure compares the effect of observed pressure changes on CSU CN signal predicted based on laboratory studies versus observed changes for a segment of the period shown on May 8. The Cooper pressure signal was used as the basis for the predicted signal. Predicted positive concentration excursions above the ambient values do not occur in the CSU CN signal, suggesting that this feature of the Cooper pressure signal was not realized in the CSU system.

Predicted versus measured effect of "wiggle" maneuvers on CN (CSU) 

Summary

Reasonable explanations for CN measurement discrepancies observed during the NASA SUCCESS mission were obtained by considering a combination of factors. These factors included different counter threshold size sensitivity, sampling loss differences versus particle size and composition, and newfound counter pressure change sensitivities that were exacerbated at certain sampling location on the aircraft.

1. All of the counters were capable of equivalent measurement of CN concentrations at sizes above the largest "cut-size." The aircraft data record substantiates that such agreement did occur over large segments of time and space during middle and upper tropospheric sampling.
2. Sensitivity of pulse heights to pressure was noted for the Cooper/NCAR (TSI 3025A) ultrafine CN counter, but this was mitigated in the laboratory by adjusting a threshold potentiometer. This adjustment was not made during the aircraft campaign and probably explains the falloff of the Cooper/NCAR ultrafine CN concentrations at pressures below about 300 mb.
3. The CSU sampling system had significant sampling efficiency losses due to diffusion for particles below about 50 nm. Fractional penetration losses typically exceeded theoretical values by about 0.1. The correction factor for diffusional losses applied to the CSU data in the SUCCESS archive was most correct for an ambient particle mode centered near 35 nm. Correction factors of up to 5 appear possible if the ambient particle mode was centered below 20 nm at upper tropospheric pressures. These particle losses can explain the occasional undercounting of the CSU CN counter (TSI 3010) by a factor of 3 to 4 compared to the CVI CN counter (TSI 3760).
4. The failure of the CSU counter to detect high concentrations of small volatile particles was indicated during extensive flight periods on two days during the field program. This could explain even larger discrepancies fferences compared to the CVI CN. This factor was not investigated in the laboratory. The UMR ultrafine CN system probably also suffered losses of small volatile particles when a metal bellows pump was used to pressurize ambient samples during the latter stages of SUCCESS.
5. Large, false CN concentration decreases can occur during certain aircraft maneuvers were related to instrumental error during pressure decreases. Every CN counter tested showed some sensitivity to the rate of pressure decrease of the sample air in laboratory tests. The CN measurement error was found to scale with the rate of negative pressure change and inversely with sample pressure. It is surmised that the pressure changes either alter counter supersaturation profiles, lead to a non-steady critical flow phenomenon, or simply induce a true stagnation of the flow within the measurement cavity. The CN pressure induced errors in the TSI 3010 counters were sensitive to the level of the saturator working fluid. This observation appears to support an effect on the supersaturation profiles as the source of CN errors.
6. Aircraft maneuvers that led to erroneous CN readings were sharp or extended changes in aircraft angle of attack. Thus, CN errors can be expected during altitude changes, turbulent conditions (e.g., during aircraft exhaust sampling) and during special procedures for calibrating meteorological sensors. Over-the-wing sample locations susceptible to stronger pressure perturbations showed the strongest and most consistent response to maneuvers that created rates of pressure change. These sampling locations should be avoided for CN sampling unless special sampling procedures are developed (e.g., Cofer et al., 1998).
7. Errors due to positive pressure deviations were apparent in varying degrees for some of the counters during aircraft operation, but were not as dramatic and were not investigated in the laboratory.