Click Here for Full Article GEOPHYSICAL RESEARCH LETTERS, VOL. 34, L11811, doi:10.1029/2006gl029109, 2007 Urban background aerosols: Negative correlations of particle modes and fragmentation mechanism G. Gramotnev, 1 P. Madl, 2 D. K. Gramotnev, 1 and M. J. Burchill 1 Received 16 December 2006; revised 12 February 2007; accepted 10 May 2007; published 13 June 2007. [1] We demonstrate new powerful methods of statistical analysis of atmospheric aerosols on the example of urban background aerosol in the Brisbane area, Australia. It is shown that even in the absence of notable features on the size distribution, it is still possible to identify distinct particles modes and analyze their mutual interactions and transformations. The obtained unique anti-symmetric correlation patterns between particle modes may serve as fingerprints of particular evolutionary processes in the background aerosols. The obtained results suggest that thermal fragmentation of nanoparticles may be one of the major physical mechanisms shaping urban background aerosols. In particular, based on the fragmentation theorem, we demonstrate possible existence of substantially different time scales for fragmentation of atmospheric aerosols. The proposed approaches and obtained results may also be important for the analysis of different types of atmospheric aerosols on local, regional and global scales. Citation: Gramotnev, G., P. Madl, D. K. Gramotnev, and M. J. Burchill (2007), Urban background aerosols: Negative correlations of particle modes and fragmentation mechanism, Geophys. Res. Lett., 34, L11811, doi:10.1029/2006gl029109. 1. Introduction [2] Atmospheric aerosols have a significant impact on the behavior of the atmosphere, climate change, rainfall patterns, urban air quality, etc. [Seinfeld and Pandis, 1998; Jacobson, 1999]. One of the major sources of particulate emissions and increased aerosol load on the atmosphere and urban population is related to anthropogenic human activities, such as transport and industry emissions [Seinfeld and Pandis, 1998; Jacobson, 1999; Ketzel and Berkowicz, 2004; Zhang and Wexler, 2004; Zhang et al., 2004; Pirjola et al., 2006]. However, monitoring and investigation of such aerosols in the real-world environment, detailed analysis of their sources, transformations and interactions between different particle modes are significantly impeded by stochastic fluctuations and variability of the atmospheric and environmental factors. [3] Therefore, new powerful methods of multi-channel statistical analysis have been developed, allowing reliable determination of major evolutionary tendencies and relationships between different types of airborne particles and their sources [Gramotnev and Gramotnev, 2005a, 2007a, 2007c]. In particular, these methods have significantly aided 1 Applied Optics Program, School of Physical and Chemical Sciences, Queensland University of Technology, Brisbane, Queensland, Australia. 2 Division of Physics and Biophysics, University of Salzburg, Salzburg, Austria. Copyright 2007 by the American Geophysical Union. 0094-8276/07/2006GL029109 recent investigation of a major new mechanism of evolution of nano-particle aerosols, based on their thermal fragmentation [Gramotnev and Gramotnev, 2005a, 2005b, 2007a, 2007b, 2007c]. Such fragmentation can occur when binding forces between the particles in a cluster become sufficiently close to the typical thermal energy (of 0.025 ev), and the particles may break away from the cluster due to their thermal motion, similar to evaporation of molecules from a liquid droplet [Gramotnev and Gramotnev, 2005a, 2007b]. It has been shown that this mechanism may have a substantial impact on evolution of combustion aerosols near busy roads, significantly increasing concentration of small nano-particles with increasing distance from the road [Gramotnev and Gramotnev, 2005a, 2005b]. It is also possible that this mechanism may play a role on larger scales, affecting evolution and transformation of atmospheric aerosols far from their source. However, no direct evidence of this has been presented so far. [4] Therefore, the aim of this paper is in monitoring and detailed statistical analysis of background urban aerosols and their particle modes. We will demonstrate that the obtained results are consistent with thermal fragmentation of nano-particles which may thus be one of the major physical mechanisms of evolution of background aerosols. Based on cross-correlations between particle modes and the fragmentation theorem [Gramotnev and Gramotnev, 2007a], we will show that fragmentation may occur on significantly different time scales, i.e., at large distances from the source. 2. Monitoring [5] The experimental measurements were conducted in the Brisbane area, Australia, on the upwind side of the Gateway Motorway, so that traffic emissions from the road did not affect the obtained background data (Figure 1). There were no buildings within 150 m upwind from the measurement area that was practically a flat grass field with scattered bushes and trees. Further away from the road, there was a small residential area with a parkland (Figure 1). No particular sources of air pollution were known within at least several kilometers upwind from the monitoring place (marked by a cross on the map Figure 1). The distance from the curb of the road was 60 m, which was sufficient for not registering noticeable particle concentrations coming directly from the Motorway. [6] Particle concentrations in the background aerosol were measured in 113 channels within the range from 13 nm to 763 nm at the height h = 2 m during the afternoon of 30 July 2002 between 1:38 pm and 5:44 pm. 38 scans with 113 equal intervals of Dlog(Dp), where Dp is particle diameter in nanometers, were taken by means of a Scanning L11811 1of5
averaging over the 17 scans in the set and 5 channels in the moving interval. [9] Only one maximum can be seen on the resultant particle size distributions at 100 nm (Figure 2b). Otherwise, the size distributions do not display significant features and/or distinct particle modes. This makes it difficult to identify and analyse possible evolutionary process in the background nano-particle aerosols. Very little can be said about possible relationships between particles of different diameters. Only small variations in the size distribution occur during the sunset (some increase in relative concentration of 30 nm particles and reduction in relative concentration of 110 nm particles compare curves 1 and 2 in Figure 2b). Figure 1. The area of measurements near Gateway Motorway, Brisbane, Australia. The indicated receptor point is at the distance 60 m from the curb of the road. The scale of the map and the direction to the North are as indicated (the distances on the axes are given in meters). The cross indicates the receptor point. The inset presents a section of the map with the area of measurements. Mobility Particle Sizer (SMPS-3071) and condensation particle counter CPC-3010. The time for one scan was 6.5 min (5 min for an up-scan and 1.5 min for a downscan). Simultaneous measurements of temperature, humidity, solar radiation, wind speed and wind direction were conducted by an automatic weather station every 20 seconds at the point of measurements (Figure 1) and the same height of 2 m. [7] 38 scans in total were taken, and two separate sets of 17 scans from 1 to 17 (from 1:38 pm to 3:26 pm) and from 22 to 38 (from 3:54 pm to 5:44 pm) were considered. The choice of these sets was primarily made on the basis of significantly different solar radiation, so that the second set of scans corresponded to nearly zero solar radiation (after sunset), while for the first set of scans it was still significant (Table 1 and Figure 2a). [8] Particle concentration in every channel in each scan was normalized to the total number concentration in this scan. For every set of scans, average (over 17 scans) normalized particle concentrations were found for each of the channels. The average particle size distributions together with their standard errors of the mean for the two sets of scans were then plotted by means of the moving average technique (with 5 channels in the moving interval) Figure 2b. The standard errors of the mean were obtained for the 3. Particle Modes [10] Nevertheless, the recently developed statistical methods of analysis of the size distribution [Gramotnev and Gramotnev, 2005a, 2007a, 2007c] provide a unique tool for the detailed investigation of the evolutionary processes in the background aerosol. These methods are based upon the moving average approach applied to the correlation coefficients between particle concentrations in different channels of the size distribution [Gramotnev and Gramotnev, 2005a, 2007a, 2007c]. For example, considering a 7-channel moving interval and calculating the average correlation coefficient for all pairs of channels within this interval, gives the moving average correlation coefficient and its standard error of the mean as functions of particle diameter for the two selected sets of 17 scans Figure 3 (for more detailed description of this approach see Gramotnev and Gramotnev [2005a]). [11] One of the important features of Figures 3a and 3b is the existence of strong and distinct maximums of the moving average correlation coefficients for both the sets of scans. Concentrations of particles in the channels within every such maximum tend to increase/decrease in maximum correlation with each other. This means that a significant fraction of these particles is likely to come from the same source or originate from the same physical/chemical process. Therefore, we use the term mode for a group of particles corresponding to a maximum of the moving average correlation coefficient [Gramotnev and Gramotnev, 2005a, 2007c]. Any such local maximum can be called mode, if it is statistically significant, i.e., its height is larger than the width of the error band (Figures 3a and 3b), and it Table 1. Average Meteorological Parameters and the Average Total Number Concentrations Together With Their Standard Deviations for the Two Sets of Scans Meteorological Parameters Scans 1 to 17 Scans 22 to 38 Wind direction, degrees 129 ± 55 139 ± 43 to the North Wind speed, m/s 2.0 ± 0.8 1.5 ± 0.8 Temperature, C 21.7 ± 0.4 18.7 ± 0.6 Humidity, % 33 ± 3 46 ± 2 Solar radiation, W/m 2 420 ± 90 33 ± 32 Average total number concentration, cm 3 2360 ± 230 2870 ± 870 2of5
of significant evolutionary processes in the urban background aerosol, which strongly depend on solar radiation. Figure 2. (a) Average solar radiation as a function of scan number. (b) Normalized (to the total number concentration) moving average size distributions in the urban background aerosol in the Brisbane area, Australia for the two sets of scans: (1) scans from 1 to 17 (high solar radiation), and (2) scans from 22 to 38 (low solar radiation). Shaded narrow bands around the curves show the insignificant standard errors of the mean for both the curves. lies above an acceptable level of confidence for the determined correlations (e.g., the dashed lines in Figures 3a and 3b for the 95% level of confidence). [12] Contrary to the original size distributions (Figure 2b), at least four statistically significant modes (corresponding to the particle diameters 27 nm, 55 nm, 110 nm, and 240 nm) can be seen in Figure 3a, and at least four modes in Figure 3b (at 30 nm, 55 nm, 80 nm, and 168 nm). This is a clear demonstration of the effectiveness of this statistical approach. It is likely that the modes with smaller particle diameters (30 nm, 55 nm, 80 nm, and 110 nm) originate from transport on the road network, because they were previously observed in combustion aerosols near a busy road [Gramotnev and Gramotnev, 2005a, 2007c]. [13] It has also been noticed that particle concentrations within the range 120 nm are larger for the second set of scans. For the particles <80 nm this increase can also be seen on the normalized size distributions (curve 2 in Figure 2a). This may be related to increasing traffic on the road network during the peak hours (the second set was taken between 3:54 pm and 5:44 pm). [14] The second important feature of Figures 3a and 3b is that the particle modes substantially change between the two sets of 17 scans (compare Figures 3a and 3b). If the set of 17 scans is moved through the 38 scans (by taking the 17-scan set between 1 and 17, 2 and 18, 3 and 19, etc.), it can be seen that the strong differences between the correlation patterns shown in Figures 3a and 3b develop in a consistent way through several steps. This is an indication 4. Cross-Correlations and Fragmentation [15] To identify and investigate these processes, we consider cross-correlations between different modes of the size distributions [Gramotnev and Gramotnev, 2007a]. For example, take a 7-channel interval of neighbouring channels so that the central channel of this interval corresponds to a mode/channel, for which we wish to determine correlations with other channels. This interval is called primary interval [Gramotnev and Gramotnev, 2007a]. Then we arbitrarily choose a second 7-channel interval in the size distribution (this will be the secondary interval [Gramotnev and Gramotnev, 2007a]). For a set of 17 scans, we will have 17 different particle concentrations in each of the channels. Calculating average concentrations over the 7 channels in the primary and secondary intervals, we obtain two columns with 17 average concentrations. We assume that these two columns correspond to the central channels of the primary and secondary 7-channel intervals. Calculate the simple correlation coefficient between the two columns with the 17 average concentrations. Considering in the same way 107 different secondary 7-channel intervals (out of the 113 channels in a scan), we obtain the dependence of the moving average correlation coefficient between the selected mode corresponding to the fixed primary 7-channel interval and all other modes/channels in the size distribution. [16] The dependencies of the moving average crosscorrelation coefficient on particle diameter for the 38.5 nm channel and 113 nm channel (the primary 7-channel intervals Figure 3. Moving average correlation coefficients as functions of particle diameter for the two sets of 17 scans: (a) from 1 to 17 (high solar radiation), and (b) from 22 to 38 (low solar radiation). Both the dependencies were plotted using 7-channel moving interval. The shadow bands indicate the corresponding standard errors of the mean. The horizontal dashed lines indicate the 95% level of confidence of the obtained correlations. 3of5
Figure 4. Moving average simple cross-correlation coefficients between the 38.5 nm channel and all other channels (solid curves), and between the 113 nm channel and all other channels (dashed curves). The dependencies (a) for the first set of scans from 1 to 17 (high solar radiation), and (b) for the second set of scans from 22 to 38 (low solar radiation). The dash-and-dot horizontal lines correspond to zero correlation coefficient. being centered at these channels) are shown in Figures 4a and 4b for the two sets of 17 scans. [17] The most interesting aspect of the presented crosscorrelations is nearly perfect symmetry of the curves for the 38.5 nm and 113 nm channels with respect to the zero line (anti-symmetric cross-correlations) after the sunset (Figure 4b). This means that the correlation coefficient between the concentrations of the 38.5 nm particles and, for example, 200 nm particles is the same in magnitude but opposite in sign to the correlation coefficient between the concentrations of the 113 nm particles and the same 200 nm particles. This is correct for all channels/modes (not just for 200 nm particles) Figure 4b. It is also important that the 38.5 nm and 113 nm modes are in almost perfect anti-correlation with each other with the correlation coefficient being close to 1 (Figure 4b). This is a strong indication that these modes are related by some evolutionary process, so that increasing concentration in one of them results in proportionally decreasing concentration in the other. [18] The anti-symmetry of the correlation dependencies in the considered dataset exists only for the pair of 38.5 nm and 113 nm channels. Changing to any other channels quickly breaks the anti-symmetry. It is interesting that though the similar symmetric tendencies can also be seen in Figure 4a (before the sunset), they are far from being as clear as in Figure 4b (after the sunset). [19] Anti-symmetric correlation patterns were previously observed in combustion aerosols within dozens of meters from a busy road [Gramotnev and Gramotnev, 2007a], and explained by thermal fragmentation of aerosol nanoparticles and fragmentation theorem [Gramotnev and Gramotnev, 2007a]. A simplified version of this theorem can be formulated as follows. If particle concentrations in two different modes/channels of the size distribution are characterized by 100% negative correlations, then the dependencies of the moving average cross-correlation coefficients for these two modes must be mirror-symmetric with respect to the zero line Figure 4b. In particular, this may occur when one of the two negatively correlating modes results from fragmentation of the other, and fluctuations of particle concentrations in these modes occur only due to the fragmentation process (for more detail see Gramotnev and Gramotnev [2007a]). [20] Thus the existence of almost perfectly anti-symmetric correlation pattern in the background aerosol Figure 4b is an indication that the fragmentation mechanism may be responsible for the evolution and transformation of background particle modes. However, in the background aerosol, the anti-symmetry of correlations (Figure 4b) occurs for different 38.5 nm and 113 nm modes (compared to the 13 nm and 126 nm modes of Gramotnev and Gramotnev [2007a]), and at significantly larger distances from the source. Therefore, possible fragmentation processes (with the release of 38.5 nm particles) resulting in the antisymmetric pattern in Figure 4b should be significantly slower than those considered by Gramotnev and Gramotnev [2007a]. They seem to occur on a substantially larger time scale, though the fragmenting particles are still likely to come from transport emissions on the urban road network (see above). As suggested by Gramotnev and Gramotnev [2005a, 2005b], fragmentation occurs due to stochastic evaporation of bonding volatile molecules (bonds) between coagulated primary particles. Evaporation of different types of bonding (volatile or semi-volatile) molecules may occur at significantly different rates. In addition, bonding of larger (38.5 nm) particles is expected to be stronger, because of larger bonding area. Therefore, different time scales of particle fragmentation can be expected. [21] The breach of the anti-symmetric correlation pattern before the sunset (Figure 4a) does not necessarily mean that fragmentation does not occur before the sunset. On the contrary, it should be faster due to higher temperature and solar radiation. However, the conditions of the fragmentation theorem [Gramotnev and Gramotnev, 2007a] seem to be satisfied only after the sunset. 5. Conclusions [22] In conclusion, the discussed new statistical approaches open unique opportunities for the detailed analysis of different types of atmospheric aerosols, their possible sources and evolutionary processes. It has been shown that even in the absence of any significant features on the size distribution, it is still possible to identify distinct particles modes and analyze their mutual interactions and transformations. This provides new deep physical insights into the evolutionary processes in atmospheric aerosols. In particular, it has been demonstrated that thermal fragmentation of nano-particles may be one of the major physical mechanisms shaping urban background aerosols. It has been shown that thermal fragmentation of airborne 4of5
nano-particles is likely to occur on significantly different time scales, leading to a possibility that this mechanism plays a major role in the formation and evolution of atmospheric aerosols not only on the regional, but also on the global scale (though further research in this direction is required). References Gramotnev, D. K., and G. Gramotnev (2005a), A new mechanism of aerosol evolution near a busy road: Fragmentation of nano-particles, J. Aerosol Sci., 36, 323 340. Gramotnev, D. K., and G. Gramotnev (2005b), Modelling of aerosol dispersion from a busy road in the presence of nanoparticle fragmentation, J. Appl. Meteorol., 44, 888 899. Gramotnev, D. K., and G. Gramotnev (2007a), Multi-channel statistical analysis of combustion aerosols. part II: Anti-correlations of particle modes and fragmentation theorem, Atmos. Environ., 41, 3535 3545. Gramotnev, D. K., and G. Gramotnev (2007b), Kinetics of stochastic degradation/evaporation processes in nanoparticle aggregates and polymerlike systems with multiple bonds, J. Appl. Phys., 101, 084902. Gramotnev, G., and D. K. Gramotnev (2007c), Multi-channel statistical analysis of combustion aerosols. part I: Canonical correlations and sources of particle modes, Atmos. Environ., 41, 3521 3534. Jacobson, M. Z. (1999), Fundamentals of Atmospheric Modelling, Cambridge Univ. Press, Cambridge, U. K. Ketzel, M., and R. Berkowicz (2004), Modelling the fate of ultrafine particles from exhaust pipe to rural background: An analysis of time scales for dilution, coagulation and deposition, Atmos. Environ., 38, 2639 2652. Pirjola, L., et al. (2006), Dispersion of particles and trace gases nearby a city highway: Mobile laboratory measurements in Finland, Atmos. Environ., 40, 867 879. Seinfeld, J. H., and S. N. Pandis (1998), Atmospheric Chemistry and Physics: From Air Pollution to Climate Change, John Wiley, New York. Zhang, K. M., and A. S. Wexler (2004), Evolution of particle number distribution near roadways Part I: Analysis of aerosol dynamics and its implications for engine emission measurement, Atmos. Environ., 38, 6643 6653. Zhang, K. M., A. S. Wexler, Y. F. Zhu, W. C. Hinds, and C. Sioutas (2004), Evolution of particle number distribution near roadways Part II: The road-to-ambient process, Atmos. Environ., 38, 6655 6665. M. J. Burchill, D. K. Gramotnev, and G. Gramotnev, Applied Optics Program, School of Physical and Chemical Sciences, Queensland University of Technology, GPO Box 2434, Brisbane, Qld 4001, Australia. (d.gramotnev@qut.edu.au) P. Madl, Division of Physics and Biophysics, University of Salzburg, A-5020 Salzburg, Austria. 5of5