Changes in particle characteristics that influence CCN activity are shown through a series of probability density distributions in Figure 4.2. The effective hygroscopicity parameter and critical supersaturation are computed according to the approach described in Section 3.2.1. At each time, the normalized
two-0.25
cond. only coag. only both cond. and coag.
number [cm−3 ] dry$mass$[µg$m−3 ] hygroscopicity parameter, κ dry diameter, Ddry [µm]
ˆndsl (Ddry,κ) cloud supersaturation, s ΔNCCN [cm−3 ] fdsl s50 κgm Ddry,gm [µm] Ψ
number [cm−3 ] dry$mass$[µg$m−3 ] hygroscopicity parameter, κ dry diameter, Ddry [µm]
ˆndsl
Figure 4.3: Evolution of (a) number fraction of diesel particles susceptible to nucleation scavenging, fdsl, and (b) difference in CCN concentrations between simulations with and without diesel emissions, ∆NCCN. The change in number concentration is indicated by the horizontal grey lines in Figure 4.3.b
Table 4.2: Time periods representing aging regimes time characteristics
6:00 am–11:00 am fresh particle emissions, slow aging by condensation
11:00 am–4:00 pm fresh particle emissions, rapid aging by condensation of ammonium nitrate 4:00 pm–6:00 pm fresh particle emissions continue after condensational aging slows
6:00 pm–6:00 am no particle emissions, slow aging by condensation
0.05 0.06 0.07 0.080.090.1
10−3 10−2 10−1
12:00 pm 6:00 pm 12:00 am 6:00 am 0.1%
0.3%
1%
3%
cond. only coag. only both cond. and coag.
exact only size changes only comp. changes
a)!
b)!
c)!
number [cm−3 ] dry$mass$[µg$m−3 ] hygroscopicity parameter, κ dry diameter, Ddry [µm]
ˆndsl
(Ddry,κ) cloud supersaturation, s
ΔNCCN [cm−3 ] fdsl s50 κgm Ddry,gm [µm] Ψ
number [cm−3 ] dry$mass$[µg$m−3 ] hygroscopicity parameter, κ dry diameter, Ddry [µm]
ˆndsl
(Ddry,κ) cloud supersaturation, s
ΔNCCN [cm−3 ] fdsl s50 κgm Ddry,gm [µm] Ψ
number [cm−3 ] dry$mass$[µg$m−3 ] hygroscopicity parameter, κ dry diameter, Ddry [µm]
ˆndsl
(Ddry,κ) cloud supersaturation, s
ΔNCCN [cm−3 ] fdsl s50 κgm Ddry,gm [µm] Ψ
Figure 4.4: For each set of aging conditions, evolution of diesel particles’ (a) geometric mean diameter and (b) geometric mean hygroscopicity parameter. For simulations including both condensation and coagulation, evolution of (c) supersaturation at which 50% of diesel particles become CCN active for exact PartMC-MOSAIC results (black line), assuming particles size changes but not their composition (red line), and assuming their composition changes but not their size (blue line). Vertical lines show the time periods outlined in Table 4.2.
dimensional distribution ˆndsl(Ddry, κ) is shown for each set of aging conditions. The unnormalized distribu-tion ndsl(Ddry, κ) is defined as:
ndsl(Ddry, κ) = ∂2Ndsl(Ddry, κ)
∂ log Ddry∂ log κ, (4.1)
where Ndsl(Ddry, κ) is the cumulative number distribution with respect to Ddryand κ. Because we compare particle populations with very different number concentrations, it is useful to use the normalized number distribution ˆndsl(Ddry, κ):
ˆ
ndsl(Ddry, κ) = ndsl(Ddry, κ) R∞
−∞
R∞
−∞ndsl.(Ddry, κ)d log Ddryd log κ (4.2) The individual contributions of condensation and coagulation are shown through comparisons between the simulation that includes both aging mechanisms (top row of Figure 4.2) and simulations in which each mechanism operated alone (middle and bottom rows of Figure 4.2). The state of diesel particles is shown every four hours from 7:00 am until 11:00 pm. The selected times were chosen to show how particle properties change during the key time periods outlined in Table 4.2.
Curves of constant sc for T =293.15 K are plotted as a function of Ddry and κ in Figure 4.2. Particles that would be CCN-active at a specified supersaturation s are those with sc,i ≤ s. Crossing the lines of constant sc from left to right means particles transition from CCN-inactive to CCN-active at that s. For the particle concentrations simulated in this paper, a cloud supersaturation of s = 0.1% is consistent with updraft velocities of 0.2–0.5 m s−1 (stratus clouds), and s = 0.6% is consistent with updraft velocities of 2–5 m s−1 (convective clouds).
Each aging mechanism caused the evolution of the particle population to proceed in a different way. An hour after emissions started (7:00 am), most diesel particles were too small and hydrophobic to activate at any s < 1%. When particles aged only by condensation, Ddry,i and κi increased incrementally over time as particles accumulated secondary aerosol coating, shown in the middle row of Figure 4.2. Photochemical production of secondary aerosol between 11:00 am and 4:00 pm caused rapid condensational aging during these hours. Particles emitted after 4:00 pm did not experience this rapid daytime aging and, consequently, remained small and hydrophobic with high values for sc,ithroughout the night, shown in the middle row of Figure 4.2.
Without condensation Ddry,iand κiwere altered only for particles that participated in coagulation events, shown in the bottom row of Figure 4.2. Because we assume that particles coagulate by Brownian motion, coagulation events were most likely between the largest and smallest particles in the distribution. The smallest of the diesel particles were often incorporated into much larger background particles through such
coagulation events, forming a mode of diesel particles with κi approximately the same as the background (blue circle in Figure 4.2). These combined particles were often CCN-active at low values of s, enabling the removal of the small diesel particle contained in the combined particle through nucleation scavenging.
The particles that did not participate in coagulation events maintained their initial hygroscopic properties throughout the simulation, shown by the mode of particles in the bottom row of Figure 4.2 that maintain κ = 3 × 10−4 throughout the entire simulation (red circle in Figure 4.2).
As particles increase in size (greater Ddry,i) and hygroscopicity (greater κi), they become more susceptible to CCN activation and wet removal. The number fraction of diesel particles that are able to activate into CCN (fdsl) is shown as a function of cloud supersaturation in Figure 4.3.a. The value of fdsl(s) at a specific s is the portion of particles with sc≤ s, corresponding to particles above that scline in Figure 4.2. Because fdsl is related to the time-dependent number concentration of diesel particles, the same value of fdsl at different times could have a very different effect on the number concentration of CCN.
Freshly emitted particles were CCN-active only at very high supersaturation levels (s > 1%), but their CCN activity increased as they aged by condensation and coagulation. Condensation of semi-volatile sub-stances between 11:00 am and 3:00 pm caused sc,i to decrease for all diesel particles (Figure 4.2), enabling their activation and removal under a greater range of cloud supersaturation levels. For simplicity, we discuss changes in the fdsl spectrum using a single value, the supersaturation required to activate 50% of parti-cles (s50). While fdsl at a specific time and supersaturation indicates the portion of particles that could be removed at that supersaturation (Figure 4.3.a), the quantity s50 provides information on the fdsl curve with only a single value by indicating the supersaturation level needed to activate most of the diesel parti-cles. In simulations including condensation alone (dashed lines) s50 decreased from s50 ≈ 1% at 11:00 am to s50 ≈ 0.3% at 3:00 pm. On the other hand, a small portion of diesel particles combined with hygro-scopic particles through coagulation events, increasing fdsl(s) at 0.1% < s < 1% in simulations including coagulation without condensation (dotted lines of Figure 4.3.a).