SEARCH

SEARCH BY CITATION

Keywords:

  • aerosol dynamics;
  • new particle formation;
  • aerosol/cloud interactions

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods
  5. 3. Performed Simulations
  6. 4. Results and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

[1] The formation of new particles and CCNs under conditions typical for clean continental areas has been investigated by performing simulations with an aerosol dynamical model. The main goal was to quantify the relative importance of the key dynamic parameters to new-particle and CCN formation. To this end, the model output was analyzed using a global sensitivity analysis technique. It was shown that there is no unambiguous answer to the question regarding the most important factor regulating new particle formation (NPF). Instead, the answer depends largely on the relative magnitudes of the pre-existing particle surface area and emission rates of aerosol precursors which vary according to the atmospheric conditions. We also demonstrate that different factors may control the formation of particles with detectable sizes and particles having larger, climatically relevant sizes. These results suggest that varying, problem-specific approaches should be adopted in large scale modeling of NPF.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods
  5. 3. Performed Simulations
  6. 4. Results and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

[2] A large body of experimental evidence supports the notion that new particle formation (NPF) contributes frequently to aerosol populations in the lower troposphere [Kulmala et al., 2004]. Newly formed particles are also observed to grow to sizes where they can affect the radiative transfer through interfering with solar radiation or acting as cloud condensation nuclei (CCN) [Kerminen et al., 2005; Laaksonen et al., 2005]. Given the local nature of such observations, measurements must be complimented with model studies to assess the global significance of the phenomenon [Spracklen et al., 2006]. The task is complicated or even prohibited at the present, however, because there are serious gaps in understanding of the mechanisms underlying NPF. For example, it remains uncertain which compounds are involved in nucleation or through which mechanisms freshly-nucleated particles grow to larger sizes [Kulmala, 2003]. Consequently there is no single theory capable of answering to the question “Does NPF takes place under given conditions or not?” even though parameterizations based on local conditions exist [Hyvönen et al., 2005; McMurry et al., 2005].

[3] One way to approach the above-posed question is to attempt to identify and rank in terms of the importance the factors which limit or promote NPF under various conditions. This would not only serve to increase our mechanical understanding of the process, but also help to develop NPF parameterizations that can be applied in regional/global models. Several attempts towards this direction has been made, and the results highlight the role of pre-existing particles, availability of condensing/nucleating vapours and meteorology in regulating NPF. According to their approach, the studies can be classified broadly into two groups: 1) statistical analysis of measurement data from intensive campaigns or continuous measurements [e.g., Dal Maso et al., 2005; Hyvönen et al., 2005] and 2) modeling studies that utilize measurement data [e.g., Gaydos et al., 2005; Tunved et al., 2006a; Lauros et al., 2007]. While giving important insights into the phenomenon, both of these approaches have their limitations. One problem associated with the statistical analyses is that many quantities involved in NPF, such as the nucleation rate, cannot be directly measured but are inferred indirectly, and hence the variation of such quantities cannot be assessed comprehensively. The modeling studies, in turn, have mainly investigated individual NPF events while less emphasis have been put on statistical features of the phenomenon. Moreover, the sensitivity of the results have been quantified by varying a single parameter at time over a limited region of the parameter space [e.g., Anttila et al., 2004; Gaydos et al., 2005] even though more sophisticated sensitivity analysis methods are available [Tatang et al., 1997; Saltelli, 1999].

[4] The present study attempts to quantify the relative importance of different factors to atmospheric NPF. Here the focus is on conditions typical for continental background areas where NPF takes place regularly and may form significant source of aerosol and CCN populations [Tunved et al., 2006b]. Furthermore, we investigate mainly the effect of the key aerosol dynamical processes. The approach is based on performing a large set of simulations with an aerosol dynamic model and extracting statistical information from the results.

2. Methods

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods
  5. 3. Performed Simulations
  6. 4. Results and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

[5] The aerosol dynamic simulations were conducted using the box model developed by Anttila et al. [2004]. Briefly, the model simulates the time evolution of a particle population undergoing nucleation, condensation and coagulation. In order to facilitate statistical analysis, the following modifications were made to the original model: 1) only one condensing vapour, having the physico-chemical properties of sulfuric acid, is assumed to form, and 2) the nucleation rate (Jnuc) and the production rate of the condensing vapour (Q) do not have any time dependence.

[6] The applied statistical technique is termed as probabilistic collocation method (PCM). Detailed descriptions of PCM can be found from the literature [Tatang et al., 1997] and its implementation as well as validation is described by Anttila and Kerminen [2007]. Briefly, PCM is a sensitivity analysis technique for numerical models that is “global” in a sense that the sensitivity of model output with respect to a varied parameter is quantified over the entire parameter space. Also, the net effects of simultaneously varying input parameters are accounted for. PCM is based on approximating the considered model output by polynomials called as polynomial chaos expansions (PCEs), the terms of which are functions of the varied input parameters. Once the PCEs have been generated, the required statistical properties of the model output can be readily calculated. The parameters towards which the model output displays the highest sensitivity are also the parameters whose variation mainly regulates NPF. Thus the method allows for ranking quantitatively the importance of the varied parameters for NPF.

3. Performed Simulations

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods
  5. 3. Performed Simulations
  6. 4. Results and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

[7] The analyzed model outputs are the number of new particles having diameters above 3 and 50 nm, respectively (CN and CCN50, respectively), in the end of a simulation. The former represent the number of particles having observable sizes and the latter particles which have relevance to climate by acting as CCN. Even though the size limit (50 nm) may seem to be quite small considering supersaturations reached in, for example, stratiform clouds, the particles may still persist long enough to experience further growth later, thereby making them relevant to the cloud albedo. The simulation time is fixed to 12 hours which is comparable to a typical time period of a NPF event under conditions considered here [Mäkelä et al., 2000].

[8] The variation of the following input parameters is considered here: the nucleation rate (Jnuc), the production rate of the condensing vapour (Q) and the condensation sink of pre-existing particles (CS). This choice is motivated by the previous studies which suggest that the key aerosol dynamic processes underlying NPF can be quantitatively described using these three parameters [Kulmala et al., 2001; Kerminen and Kulmala, 2002]. The background aerosol is assumed to be log-normally distributed with a mean diameter of 160 nm and a geometric standard deviation of 1.5. Furthermore, the variation of CS is realized by varying the number of pre-existing particles in the range of ∼100 to 1000 cm−3. Sensitivity studies showed that the results discussed in Section 4 are robust to the choice of the mean diameter as long as the value range of CS does not change (not illustrated here).

[9] The varied input parameters are treated as independent random variables with certain probability distributions. It should be kept in mind that the assumption regarding independency may not be realistic under all conditions; Q and Jnuc, for example, are correlated when nucleation and growth are driven by the same compounds. In such case, PCM analysis can be performed only if the functional dependencies (e.g., Jnuc on Q) are known or can be estimated quantitatively. The current uncertainties regarding mechanisms underlying NPF preclude this, however, and the independency assumption is thus a reasonable starting point. This topic is discussed further in Section 4.3.

[10] Choosing the probability distributions describing the variation of the parameter values in the atmosphere is also complicated because 1) no comprehensive model capable of predicting the variation exists at the present and 2) the data on some of the parameters are limited to periods with NPF which makes the data set incomprehensive. To overcome these limitations, we adopt the following approach. First, it is assumed that each varied parameter is distributed uniformly over a certain value range, corresponding to the lack of information on the actual distribution. Second, we make various “scenarios” where the value range of an input parameter is changed. The parameter value ranges for each scenario are shown in Table 1. The base case scenario (termed as “BASE”) is based on direct measurements (in the case of CS) or indirect estimates based on observations (in the case of Jnuc and Q) in a site located in boreal forest [Dal Maso et al., 2005]. In the “NUC” and “PROD” scenarios, we investigate the possibilities that Jnuc and Q, respectively, may vary over a larger value range than the indirect estimates suggest. In both scenarios, the value range is extended by an order of magnitude towards smaller values. Even though the magnitude of the decrease is subjective, the choice is still feasible considering the current uncertainties on atmospheric nucleation and formation of condensable vapors. The scenarios “SINKLO” and “SINKHI” are motivated by the fact that CS seems to control NPF to a large extent [Dal Maso et al., 2005; Hyvönen et al., 2005]. Also, CS is the only variable of which variation can be directly measured. For these reasons we constrain the value of CS to smaller intervals.

Table 1. Value Ranges of the Varied Input Parameters for Each Scenario
ParameterAbbreviationBASENUCPRODSINKHISINKLO
Production rate of the condensing vapour, cm−3 s−1Q104 − 2 × 105104 − 2 × 1055 × 103 − 2 × 105104 − 2 × 105104 − 2 × 105
Nucleation rate, cm−3 s−1Jnuc10−1 − 10110−2 − 10110−1 − 10110−1 − 10110−1 − 101
Condensation sink, s−1CS10−3 − 10−210−3 − 10−210−3 − 10−25 × 10−3 − 10−210−3 − 5 × 10−3

[11] For each scenario, around 50 simulations are performed to generate the PCEs and 750 Monte Carlo simulations are performed to validate the method and to obtain more statistical characteristics.

4. Results and Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods
  5. 3. Performed Simulations
  6. 4. Results and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

4.1. General Results

[12] Table 2 shows the results from the sets of simulations performed in a Monte Carlo fashion. Here we do not focus on the results from individual model runs but on statistical features, and thus we note that the aerosol dynamics related to NPF has been analyzed previously through case studies [Anttila et al., 2004; Gaydos et al., 2005; Tunved et al., 2006a] and that the conducted simulations display the same general features than shown in previous works.

Table 2. Statistical Features of the Model Outputa
 BASENUCPRODSINKLOSINKHIMINMAX
  • a

    The average intensity refers to the average number of new particles/CCN50 formed in simulations with new particle/CCN50 formation. The avg. factor of increase is the factor over which the total particle/CCN50 concentration increases in simulations new particle/CCN50 formation, respectively. MIN and MAX refer to the smallest and largest value, respectively, in each data row.

CN
Event frequency0.750.700.630.940.290.290.94
Average intensity41483129416548703803804870
Avg. factor of increase10.17.610.218.60.50.5018.58
CCN50
Event frequency0.200.230.160.310.020.020.31
Average intensity29381859304327562642643043
Avg. factor of increase7.34.67.610.70.40.410.7

[13] Table 2 shows that, depending on the scenario, from 380 to 4870 particles cm−3 are produced on average in one NPF event. Here a NPF event is defined as a model run where the number of new particles, i.e. particles that are formed through nucleation and condensation, having a diameter >3 nm is larger than zero in the end of the simulation. The largest differences in NPF characteristics are found between “SINKLO” and “SINKHI”, highlighting the role of pre-existing particles in regulating NPF. Scenario-wise, ∼3100-4900 particles cm−3 are produced on average in a NPF event excluding “SINKHI”. This corresponds to increases in the particle number concentrations by a factor between 8 and 19 on average, depending on the scenario. Moreover, by taking the maximum of the particle concentrations reached at the end of a simulation for each scenario, a value between 6150 and 19000 particles cm−3 is obtained, depending on the scenario (not shown). These numbers compare well with the observed enhancements in the particle number concentrations due to NPF [Mäkelä et al., 2000; Tunved et al., 2006a]. On the other hand, the simulated NPF frequency is clearly higher in “BASE” (75%) than the observational event frequency (25–45% of classified days). The reason for this lies probably in differing definitions of a NPF event: the observational criterion is stricter than the computational criterion adopted here; for example, according to the former, an appearance of a distinct nucleation mode is required [Dal Maso et al., 2005]. This point is supported by a numerical test which showed that if only those simulations are qualified as NPF events where the particles concentrations are increased by >20%, the calculated event frequency decreases to 57%. Additional factor making the comparison complicated is that a large fraction (around 37%) of observational data could not be classified as an event or non-event, and consequently there might be a large number of days during which observable particles were formed but which were neglected in the classification.

[14] The comparison between “BASE”, “NUC” and “PROD” scenarios shows that lower nucleation rates lead to lower number of particles produced (∼25% decrease), whereas lower values of Q lead to smaller event frequencies (∼16% decrease).

[15] The largest differences in the results for CCN50 are seen between “SINKLO” and “SINKHI” which highlights the role of pre-existing particles in new CCN formation (Table 2). In particular, CCN50 production is almost totally suppressed in “SINKHI”. Regarding the other scenarios, CCN50 varies between 1860 and 3040 cm−3 on average, depending on the scenario, with event frequencies ranging between 16 and 31%. The CCN50 formation events are predicted to increase the pre-existing concentrations by a factor between 5 and 11 on average, implying that NPF may contribute also to the CCN budget significantly. Another interesting feature is that smaller number of CCN50 is formed on average in “SINKLO” than in “BASE”. This is explained by the increased number of events with relatively small CCN50 concentrations which is caused by the fact that lower values of Q and Jnuc are sufficient to produce particles with sizes >50 nm. The main differences between “BASE”, “NUC” and “PROD” are similar to those for CN: lower nucleation rates decrease the average number of new CCN50 (∼37% decrease), whereas lower values of Q decrease the CCN50 event frequency (∼20% decrease).

[16] The results suggest that NPF can be viewed as a mechanism which maintains aerosol and CCN populations in areas with relative weak particulate emissions and sufficient production of aerosol precursors. Hence the results corroborate the hypothesis that NPF is an important source of particles and CCN in boreal forest environment [Tunved et al., 2006b]. On the other hand, the formation of new particles/CCN may be suppressed in areas with unfavorable emission profiles. Given the large spectrum of atmospheric conditions, the importance of NPF may vary strongly even within a region, as suggested by Pierce and Adams [2007].

4.2. PCM Analysis

[17] The validity of PCM analysis depends on how accurately the generated PCEs approximate model output [Tatang et al., 1997]. Therefore the performance of the PCEs was evaluated against the model simulations (Table 3). Here should be kept in mind that the natural logarithms of the total number of particles/CCN50 are the considered model outputs in the conducted PCM analysis for reasons discussed by Anttila and Kerminen [2007]. Table 3 shows that the degree of determination (R2) for the PCEs is always >0.63 and >0.75 in most cases. A visual investigation of the correlation plots revealed that lower R2 values are caused by a number of outliers (not shown). However, no systematic biases were found. Moreover, the expectation value and variance of the approximated model output can be calculated using the PCEs generated [Anttila and Kerminen, 2007], and Table 3 shows the associated errors. The relative errors in the expectation value are <3%, whereas the errors in the variance are consistently larger, ranging between 3 and 24%. Overall, the generated PCEs are accurate enough to warrant the conclusions based on the results of the PCM analysis.

Table 3. Comparison of the Model Output and Corresponding PCE Calculationsa
 BASENUCPRODSINKLOSINKHIMINMAX
  • a

    Here R2 is the coefficient of determination for the PCE calculations. MIN and MAX are the same as in Table 2.

CN
R20.880.850.830.870.640.640.88
Mean relative error, %5.35.56.35.42.22.26.3
Mean absolute error0.40.40.40.40.10.10.4
Error in the expectation value, %1.91.31.50.60.00.01.9
Error in the variance, %7.418.710.84.221.24.221.2
CCN50
R20.650.630.780.750.860.630.86
Mean relative error, %7.16.65.67.20.60.67.2
Mean absolute error0.50.40.40.40.00.00.5
Error in the expectation value, %2.02.21.90.50.20.22.2
Error in the variance, %13.913.47.124.33.73.724.3

[18] Figure 1 shows the relative contribution of the considered parameters to the total variance of the model output. Regarding CN, CS is seen to cause most of the variance in the base case which corroborates the results of Dal Maso et al. [2005] and Hyvönen et al. [2005]. However, Jnuc and Q dominate the total variation in “NUC” and “PROD” scenarios, respectively, where the value ranges of Jnuc and Q, respectively, are increased (Table 1). Hence CS may have a smaller role in NPF taking place at clean continental boundary layers than the previously thought if Jnuc and/or Q vary more than the indirect estimates suggest. The contribution of CS to the total variance is also smaller in “SINKLO” and “SINKHI” which is partly due to the smaller variation of CS (Table 1). In “SINKLO”, the role of Jnuc is enhanced because small values of CS allow large fractions of freshly-nucleated particles to grow to observable sizes (Table 2). Consequently the nucleation rate regulates the number of new particles. On the other hand, Q dominates the total variation in “SINKHI”. The result can be explained by noting that larger amounts of condensing vapors are needed to overcome the higher scavenging rates of newly-formed particles that are caused by larger values of CS.

image

Figure 1. The relative contributions of the varied parameters to the total variance of the model output. The results are shown for (a) CN and (b) CCN50.

Download figure to PowerPoint

[19] Results for CCN50, shown also in Figure 1, differ clearly from those for CN. First, Q makes the largest contribution to the total variation in all scenarios excluding “SINKHI”. The role of Q is pronounced because sustained and sufficiently rapid growth is needed before newly-formed particles reach diameters >50 nm. This requires, in turn, sufficiently high production rate of condensable vapors. What comes to “SINKHI”, only marginal CCN50 production takes place (Table 2) and thus the CCN50 concentrations are determined by the pre-existing particles of which variation is reflected in CS. As also seen, Jnuc contributes only up to ∼10% to the total variance. This implies that large errors or uncertainties in nucleation rates do not necessarily translate to erroneous predictions regarding CCN production. Rather, more emphasis should be put on determining the production rate of vapours responsible for growth to CCN sizes.

[20] The results displayed in Figure 1 show that even for a specific atmospheric environment, such as the clean continental boundary layer considered here, the relative importance of different factors to NPF may largely depend on the relative magnitudes of the particle surface area, photochemical activity and precursor emissions. Furthermore, the comparison of results for CN and CCN50 show that different factors may control the formation of particles with detectable sizes and particles having larger, climatically relevant sizes. These points suggest that different approaches, which put emphasis on different factors according to the problem, should be adopted in large scale modeling of NPF to reach optimal compromise between accuracy and computational costs.

4.3. Discussion

[21] An important assumption made is the independency of the analyzed input parameters (Section 3). In order to explore if possible interdependencies could affect the results, we considered a case where Jnuc is coupled to Q and CS as follows. Based on Kulmala et al. [2006], Jnuc was allowed to depend linearly on the concentration of the nucleating vapour which we assumed, for simplicity, to be the same which grows newly-formed particles (vapour X). Consequently, Jnuc = k[X]. The parameter k was treated as an independent random variable which is varied over one order of magnitude (chosen according to Sihto et al. [2006]), but the value interval of k was shifted so that the resulting Jnuc values are consistent with the observational estimates.

[22] Using the above-described model setup, “BASE”, “NUC” and “PROD” scenarios were repeated. Here “NUC” was realized by extending the value range of k by one order of magnitude towards smaller sizes. The results show that the event frequencies differ only <5% from the corresponding values displayed in Table 2, but the factor of increases are higher in the coupled simulations: for CN and CCN50, they are around 20–22 and 18–19, respectively. This can be explained by the positive coupling of Jnuc with Q. Regarding the relative contributions to the variation in the model output, the contribution of k is <7% in all cases. Also, Q dominates the total variation of CCN50 in all scenarios, CS makes dominant contribution to the variation of CN in “BASE” and “NUC”, and Q is largest source of the variation of CN in “PROD”.

[23] The investigated kind of coupling does not thus affect our main results. This seems to suggest that the results hold regardless of possible covariances, but further work is needed to assess this. Here we would like to note that extracting covariances from the existing data sets could be hampered by the fact that the current estimates regarding Q and Jnuc are not entirely independent [e.g., Kulmala et al., 2001]. Therefore approaches which would allow independent measurements or estimates regarding nucleation and particle growth are desirable.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods
  5. 3. Performed Simulations
  6. 4. Results and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

[24] The work has been supported by the Academy of Finland (project 208208). T. A. acknowledges financial support from the Emil Aaltonen Foundation.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods
  5. 3. Performed Simulations
  6. 4. Results and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information
  • Anttila, T., and V.-M. Kerminen (2007), On the contribution of Aitken mode particles to cloud droplet populations at continental background areas: A parametric sensitivity study, Atmos. Chem. Phys., 7, 46254637.
  • Anttila, T., V.-M. Kerminen, M. Kulmala, A. Laaksonen, and C. D. O'Dowd (2004), Modelling the formation of organic particles in the atmosphere, Atmos. Chem. Phys., 4, 10711083.
  • Dal Maso, M., M. Kulmala, I. Riipinen, R. Wagner, T. Hussein, P. P. Aalto, and K. E. J. Lehtinen (2005), Formation and growth of fresh atmospheric aerosols: Eight years of aerosol size distribution data from SMEAR II, Hyytiälä, Finland, Boreal Environ. Res., 10, 323336.
  • Gaydos, T. M., C. O. Stanier, and S. N. Pandis (2005), Modeling of in situ ultrafine atmospheric particle formation in the eastern United States, J. Geophys. Res., 110, D07S12, doi:10.1029/2004JD004683.
  • Hyvönen, S., et al. (2005), A look at aerosol formation using data mining techniques, Atmos. Chem. Phys., 5, 33453356.
  • Kerminen, V.-M., and M. Kulmala (2002), Analytical formulae connecting the “real” and “apparent” nucleation rate and the nuclei number concentration for atmospheric nucleation events, J. Aerosol Sci., 33, 609623.
  • Kerminen, V.-M., H. Lihavainen, M. Komppula, Y. Viisanen, and M. Kulmala (2005), Direct observational evidence linking atmospheric aerosol formation and cloud droplet activation, Geophys. Res. Lett., 32, L14803, doi:10.1029/2005GL023130.
  • Kulmala, M. (2003), How particles nucleate and grow, Science, 302, 10001001.
  • Kulmala, M., M. Dal Maso, J. M. Mäkelä, L. Pirjola, M. Väkevä, P. Aalto, P. Miikkulainen, K. Hämeri, and C. D. O'Dowd (2001), On the formation, growth and composition of nucleation mode particles, Tellus, Ser. B, 53, 479490.
  • Kulmala, M., H. Vehkamäki, T. Petäja, M. Dal Maso, A. Lauri, V.-M. Kerminen, W. Birmili, and P. H. McMurry (2004), Formation and growth rates of ultrafine atmospheric particles: A review of observations, J. Aerosol. Sci., 35, 143176.
  • Kulmala, M., K. E. J. Lehtinen, and A. Laaksonen (2006), Cluster activation theory as an explanation of the linear dependence between formation rate of 3 nm particles and sulphuric acid concentration, Atmos. Chem. Phys., 6, 787793.
  • Laaksonen, A., A. Hamed, J. Joutsensaari, L. Hiltunen, F. Cavalli, W. Junkermann, A. Asmi, S. Fuzzi, and M. C. Facchini (2005), Cloud condensation nucleus production from nucleation events at a highly polluted region, Geophys. Res. Lett., 32, L06812, doi:10.1029/2004GL022092.
  • Lauros, J., E. D. Nilsson, M. Dal Maso, and M. Kulmala (2007), Contribution of mixing in the ABL to new particle formation based on observations, Atmos. Chem. Phys., 7, 47814792.
  • Mäkelä, J. M., M. Dal Maso, L. Pirjola, P. Keronen, L. Laakso, M. Kulmala, and A. Laaksonen (2000), Characteristics of the atmospheric particle formation events observed at a boreal forest site in southern Finland, Boreal Environ. Res., 5, 299313.
  • McMurry, P. H., et al. (2005), A criterion for new particle formation in the sulfur-rich Atlanta atmosphere, J. Geophys. Res., 110, D22S02, doi:10.1029/2005JD005901.
  • Pierce, J. R., and P. J. Adams (2007), Efficiency of cloud condensation nuclei formation from ultrafine particles, Atmos. Chem. Phys., 7, 13671379.
  • Saltelli, A. (1999), Sensitivity analysis: Could better methods be used? J. Geophys. Res., 104, 37893794.
  • Sihto, S.-L., et al. (2006), Atmospheric sulphuric acid and aerosol formation: Implications from atmospheric measurements for nucleation and early growth mechanisms, Atmos. Chem. Phys., 6, 40794091.
  • Spracklen, D. V., K. S. Carslaw, M. Kulmala, V.-M. Kerminen, G. W. Mann, and S.-L. Sihto (2006), The contribution of boundary layer nucleation events to total particle concentrations on regional and global scales, Atmos. Chem. Phys., 6, 56315648.
  • Tatang, M. A., W. Pan, R. G. Prinn, and G. J. McRae (1997), An efficient method for parametric uncertainty analysis of numerical geophysical models, J. Geophys. Res., 102, 21,92521,932.
  • Tunved, P., H. Korhonen, J. Ström, H.-C. Hansson, K. E. J. Lehtinen, and M. Kulmala (2006a), Is nucleation capable of explaining observed aerosol integral number increase during southerly transport over Scandinavia? Tellus, Ser. B, 58, 129140.
  • Tunved, P., H.-C. Hansson, V.-M. Kerminen, J. Ström, M. Dal Maso, H. Lihavainen, Y. Viisanen, P. P. Aalto, M. Komppula, and M. Kulmala (2006b), High natural aerosol loading over boreal forests, Science, 312, 261263.

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Methods
  5. 3. Performed Simulations
  6. 4. Results and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
grl24393-sup-0001-t01.txtplain text document0KTab-delimited Table 1.
grl24393-sup-0002-t02.txtplain text document1KTab-delimited Table 2.
grl24393-sup-0003-t03.txtplain text document1KTab-delimited Table 3.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.