-
Increases in surface-level aerosol number concentrations are associated with low-level temperature inversions. Quantifying the effect of low-level temperature inversions on aerosols is important for the study of the dispersion of pollutants (Wallace and Kanaroglou, 2009). Changes in surface aerosol number concentration when temperature inversions (both SBI and EI) were present during the 16-year period were quantified. One-hour mean surface aerosol number concentrations were obtained by averaging measurements over the 30 minutes before and after the radiosonde launch time. The effect of SBIs on aerosols during different seasons was also assessed. In addition, relationships between SBI parameters and aerosol number concentration were examined.
According to the thermal structure of a given temperature profile, the mean aerosol number concentrations of all profiles were divided into three groups, i.e., SBI, only EI, and no temperature inversion (NTI). For SBI episodes, the statically neutral residual layer usually appears above the stable boundary layer and above it is the capping inversion, which is an elevated inversion layer that separates boundary-layer air from free-atmosphere air (Stull, 1988). There is convincing evidence that residual layers have an impact on air quality in the morning (Neu et al., 1994; Fochesatto et al., 2001). To investigate the impact of the residual layer on aerosols near the ground, all SBI episodes were divided into two groups, i.e., only SBI and SBI under an EI. Tests were performed to determine whether changes in aerosol number concentration associated with the two groups were statistically significant. The methodologies used were One-way Analysis of Variance (ANOVA) and Fisher’s Least Significant Distance (LSD) test (Milionis and Davies, 2008). The null hypothesis (H0) was that no difference in mean aerosol number concentration exists due to the presence of different types of temperature inversions. H0 was tested using the F-statistic. When H0 was rejected, the one-way ANOVA could only determine that mean aerosol number concentrations among the groups were significantly different, but it did not necessarily confirm that every set of two groups was sufficiently different. The LSD test is one of multiple comparison tests that can examine this difference between groups. For the application of this test, all differences between the two groups were compared with the LSD. When the difference between two groups’ means was greater than the LSD, the two groups were significantly different at the α-significance level (i.e., 0.05).
Aerosol number concentrations near the ground and temperature inversions have obvious diurnal variations (Sheridan et al., 2001), and they interact with each other (Yu et al., 2002), so the effects of temperature inversions on aerosols are likely to have diurnal variations. To assess the diurnal variation, mean aerosol number concentrations in the presence of SBIs, EIs, SBI/EIs, and when there was no temperature inversion were compared for different times of the day. Table 1 summarizes the results of the one-way ANOVA and the LSD test for differences in the mean aerosol number concentration in the presence of a temperature inversion, an SBI, an EI or an SBI under an EI (SBI/EI), and when there was no temperature inversion. Percentage increases in aerosol number concentration due to temperature inversions are also presented. Differences in mean particle concentration at the four observation times for the four groups were statistically significant at the 95% confidence level, except at noon (the value of the F-statistic was 1.0 and almost no SBIs occurred). The LSD tests results for all observation times showed that there were significant differences (difference in means < LSD) in mean aerosol number concentration for the SBI&EI, SBI/EI&EI SBI&NTI, and SBI/EI&NTI groups, except for SBI&EI and SBI&NTI at dusk. For EI&NTI cases, however, there were no significant differences (difference in means > LSD) for all observation times. The LSD test also showed that there was significant difference for the SBI&SBI/EI group except at dawn (0530 LST). At dawn (0530 LST), mean aerosol concentrations were 49.1%, 49.4% and 4.5% higher when SBIs, SBI/EIs and EIs were present than when there were no temperature inversions, respectively. At midnight (2330 LST), these percentages were 31.3%, 36.9% and −10.0%, respectively. At dusk, the result was similar to that at midnight, whereas the magnitudes of these percentages were smaller (9.8%, 23.2% and −1.9%, respectively). At midnight and dusk, higher aerosol number concentrations appeared when an SBI under an EI was present. This result confirms that the simultaneous occurrence of the residual layers and SBI has a significant effect on the accumulation of aerosols near the ground, but this effect is weak at dawn. Slightly higher and even lower aerosol number concentrations at all observation times were seen in only the presence of an EI. This is partly because only the presences of EIs occur most frequently and are more intense during the passage of a cold front. The passage of a cold front has a strong scavenging effect on aerosols (Li et al., 2015). Aerosol number concentrations in the presence of an SBI at dusk were not significantly different to those measured with an EI or no temperature inversion present. Most SBIs formed around sunset, which were also less intense and had a weaker effect on aerosols. This result is consistent with the relationships between temperature inversion parameters and aerosol number concentration in Figs. 10b and c. At noon, when SBIs were far less frequent (1130 LST), only changes in aerosol number concentration in the presence of EIs were calculated, which amounted to a low of 7.1%.
Aerosol number concentration (cm−3) SGP 0530 LST 1130 LST 1730 LST 2330 LST SBI 4048/790 −/− 5055/158 4647/988 EI 2836/1318 4455/3860 4510/2926 3187/1204 SBI/EI 4053/2238 −/− 5668/505 4845/2365 NTI 2713/39 4157/520 4599/923 3539/78 F-statistic 153.6 1.0 16.8 156.9 LSD Test MD & LSD SBI & NTI 510 < 1333 − 554 > 455 488 < 1107 SBI & EI 147 < 1210 − 563 > 544 170 < 1459 SBI & SBI/EI 148 > 6 − 609 < 613 173 < 198 EI & NTI 525 > 123 374 > 299 257 > 89 444 > 352 EI & SBI/EI 120 < 1217 − 332 < 1158 157 < 1658 SBI/EI & NTI 583 < 1340 − 363 < 1068 543 < 1305 SBI increase 49.1% − 9.8% 31.3% EI increase 4.5% 7.1% −1.9% −10.0% SBI/EI increase 49.4% − 23.2% 36.9% Table 1. Mean aerosol number concentration (in cm−3) in the presence of an SBI, an EI, an SBI under EI, and NTI. Numbers in red are the number of profiles. DM: difference in means; LSD: Fisher’s Least Significant Distance. The significance levels of the F-test and LSD test are 5%. The LSD test results that have no significant difference are marked in bold.
Figure 10. Aerosol number concentration as a function of (a) depth, (b) dT, and (c) dT/dz. Vertical dashed lines represent standard deviations.
Because there were no significant differences in mean aerosol number concentration for SBI&SBI/EI and EI/NTI cases, all profiles were then divided into two categories: profiles containing an SBI, and profiles without an SBI. The effects of SBIs on aerosol number concentration were then reanalyzed for different times of the day. Table 2 shows mean aerosol number concentrations for cases with and without an SBI based on ANOVA statistics. Mean surface aerosol number concentrations increased by 43.0 %, 21.9 % and 49.2 % when SBIs were present at 0530, 1730 and 2330 LST, respectively. All the F-statistic values were greater than F0.05/2 (5.02), especially for the midnight and dawn categories. Because most profiles obtained at 0530 LST were launched before sunrise, SBI episodes at 0530 and 2330 LST were combined for the following analysis. When considering only midnight and dawn profiles, the increase in mean aerosol number concentration due to SBIs was 47.2%. The F-value was 854.9, which was much greater in magnitude than the other F-values. The distributions of aerosol number concentration frequencies at midnight and at dawn with SBIs and without SBIs present are shown in Fig. 8. All distributions seen in Fig. 8 are similar to the Fisher–Snedecor distribution, but the magnitude and location of the distribution peaks are markedly different. The aerosol number concentration peaked at 3500–4500 cm−3 when an SBI was present and reached a maximum at 1500–2500 cm−3 when there was no SBI.
SGP Aerosol number concentration (cm−3) Cases ANOVA SBI No SBI Increase % SBI No SBI F-statistic 0530 LST 4052 2833 43.0 3028 1381 450.1 1730 LST 5522 4531 21.9 663 3839 46.2 2330 LST 4786 3208 49.2 3353 1282 462.7 Midnight and dawn 4438 3014 47.2 6381 2663 854.9 Table 2. Mean aerosol number concentrations for cases with and without an SBI based on ANOVA statistics. The significance level of the F-test is 5%.
Figure 8. Distributions of aerosol number concentration frequencies during midnight and at dawn in the presence of SBIs (solid line) and when there are no temperature inversions (dashed line).
Figure 9 shows a seasonal comparison of mean aerosol number concentration for cases with and without an SBI at midnight and at dawn. The difference was smallest in the summer and larger during the other seasons. When there was no SBI, the mean particle concentration reached its peak in the summer and a minimum in the winter. The seasonal difference was smaller when no SBI was present than when an SBI was present. The presence of SBIs led to increases in mean aerosol number concentration of 56.8%, 18.1%, 55.2% and 58.4% for spring, summer, autumn and winter, respectively.
Figure 9. Seasonal comparison of mean aerosol concentrations for cases with (red) and without (black) an SBI.
The relationships between SBI parameters and aerosol number concentrations at midnight and dawn were examined quantitatively. SBI parameters (i.e., depth, dT and dT/dz) were divided into 8–10 bins, based on the magnitudes of these parameters. The mean and standard deviation of values in each bin were then calculated. Figure 10 shows aerosol number concentration as a function of the three parameters. There was almost no relationship between the depth of the SBI and aerosol number concentration (Fig. 10a), but a quadratic correlation existed between aerosol number concentration and both dT and, to a stronger degree, dT/dz, at the 95% confidence level (Figs. 10b and c). This means that the temperature difference and temperature gradient across a temperature inversion correlates fairly well with aerosol concentrations near the ground. For SBI episodes at midnight and dawn, thermally stable conditions occurred when warmer air overlay cooler, denser air. In this situation, an SBI introduces an important vertical stratification dT/dz where buoyancy is suppressed, and leads to the accumulation of aerosols near the ground. This can partly explain the relationships between SBI parameters and aerosol number concentration. The regression equations for dT and dT/dz are quadratic equations of the form: y = −12.9x2 + 297.9x + 3402.0 and y = −12.8x2 + 330.3x + 3641.0, where y is the aerosol number concentration, and x is dT or dT/dz. The surface aerosol number concentration increased with dT or dT/dz, but the mean rates of increase in aerosol number concentration decreased with dT or dT/dz. The aerosol number concentration almost never increased at dT greater than 10°C or dT/dz greater than 12°C (100 m)−1. This shows that the relationship between surface aerosol number concentration and the strength of temperature inversion is not a simple positive proportional relationship, and when temperature inversion is sufficiently strong, the effect of temperature inversion on aerosol accumulation does not increase with the increase of temperature inversion strength. These equations are helpful for the modeling of aerosol dispersion in the lower atmosphere.
The temperature gradient is the main parameter characterizing the strength of a temperature inversion and it determines low-level atmospheric stability. It is also the most relevant temperature inversion parameter associated with changes in aerosol number concentration. The combination of this strong positive relationship and the seasonal variation in dT/dz (Fig. 6d) can partly explain the seasonal differences in the effect of SBIs on aerosol number concentration at the SGP site (Fig. 9). In the summer, the temperature gradient of SBIs is at a minimum, so the mean aerosol number concentration when an SBI is present is small. For other seasons, the temperature gradient of SBIs is large with little variation, so the mean aerosol number concentration when an SBI is present is high and nearly constant with time.
-
Surface measurements of aerosol optical properties are usually used to estimate the aerosol contribution to radiative forcing, and the applicability of such calculations depends on how well surface measurements represent column properties (Andrews et al., 2004). It is important then to know the relationship between surface and aircraft measurements of aerosol optical properties. The aircraft flew 698 vertical profile flights between April 2000 and December 2007. For each profile flight, the aircraft flew over the SGP site at heights of 15, 295, 600, 905, 1210, 1515, 2125, 2735, and 3345 m. Because low-level temperature inversions with bases below 3000 m were the focus of this study, only those flight legs below this height were considered. All flight data were checked to ensure that there was valid surface data and an IAP temperature profile during the flight, and SONDE temperature profiles within one hour of the beginning or end of the flight. The total number of flights available for this study was 514. To determine how a temperature inversion (an SBI or EI) affected aerosol properties, each flight was categorized as one with an EI, an SBI, or without a temperature inversion using SONDE and IAP temperature profiles. Each flight with a temperature inversion was further categorized as one within or above the temperature inversion. The number of flights with an EI, an SBI, and without a temperature inversion was 287, 64 and 163, respectively. Mean aerosol properties measured during flight legs that were under (within) and above EIs and SBIs were compared to surface measurements.
The distribution of flight legs with respect to EIs and results from the correlation analysis for surface and vertically averaged aerosol properties for each flight leg were examined (Table 3). Relatively small samples were not analyzed because statistically significant results of small samples are usually ignored. Table 3 shows that 70% of the flights flown near 600 m (i.e., 915 m above sea level) were within an EI and that 16% of the flight legs were near 1220 m. Aerosol properties for flight legs under an EI more closely matched surface measurements than those measured above an EI. The σsca, σext and b measured during all flight legs under an EI were highly correlated with surface measurements, but poorly correlated with surface measurements when the flight leg was above an EI. Aerosol properties obtained under an EI at the two lowest flight levels correlated well with values obtained at the surface. Little correlation between surface measurements and measurements made above an EI, especially during the flight legs between 1000 and 3000 m, was seen. Results for SBIs are shown in Table 4. There was an 83% chance that flight legs below ~610 m were within an SBI. Aerosol measurements made within an SBI were highly correlated with surface measurements. The correlation at the lowest flight level was slightly better than that when an EI was present. Aerosol properties measured above an SBI and at the surface did not agree well. The correlation coefficients decreased with height and were higher in magnitude than those above an EI.
Mean height (m) Flight legs under/above EI Num σabs σsca σext SSA b å 145 285/2 0.74/− 0.96/− 0.96/− 0.71/− 0.82/− 0.55/− 295 266/21 0.71/− 0.95/− 0.94/− 0.67/− 0.82/− 0.58/− 600 200/87 0.68/0.56 0.93/0.65 0.93/0.65 0.51/0.25 0.77/0.51 0.58/0.35 915 110/177 0.59/0.53 0.90/0.44 0.90/0.45 0.29/0.26 0.67/0.11 0.34/0.26 1220 47/240 0.50/0.49 0.88/0.32 0.89/0.34 0.33/0.15 0.74/0.00 0.43/0.12 1540 12/275 −/0.31 −/0.36 −/0.37 −/0.02 −/0.00 −/0.04 2190 0/287 −/0.60 −/0.36 −/0.39 −/0.06 −/0.00 −/0.00 2780 0/287 −/0.47 −/0.36 −/0.39 −/0.00 −/0.00 −/0.00 Table 3. Correlation analysis of surface and vertically averaged aerosol properties for each flight leg over the SGP site at the 95% confidence level. The numbers of flight legs and the correlation coefficients between surface and vertically averaged aerosol properties under an EI are marked in bold.
Mean Height (m) Flight legs within/above SBI Num σabs σsca σext SSA b å 140 59/5 0.81/− 0.96/− 0.96/− 0.76/− 0.94/− 0.58/− 300 48/16 0.63/0.42 0.95/0.88 0.94/0.86 0.59/0.44 0.87/0.50 0.56/0.42 610 11/53 −/0.51 −/0.74 −/0.72 −/0.64 −/0.67 −/0.39 930 2/62 −/0.29 −/0.77 −/0.76 −/0.52 −/0.62 −/0.28 1240 0/64 −/0.31 −/0.72 −/0.70 −/0.34 −/0.39 −/0.04 1560 0/64 −/0.37 −/0.67 −/0.66 −/0.49 −/0.29 −/0.00 2200 0/64 −/0.43 −/0.49 −/0.48 −/0.36 −/0.26 −/0.00 2830 0/64 −/0.28 −/0.36 −/0.36 −/0.06 −/0.02 −/0.02 Table 4. As in Table 3, but for SBIs.
Vertical profiles of aerosol optical properties from flight legs made under different temperature inversion scenarios are shown in Fig. 11. For flights without a temperature inversion, leg-averaged σabs at the lowest flight level was greater than that measured at the surface, but σsca and σext were slightly less than the equivalent values at the surface. The leg-averaged σabs, σsca and σext tended to decrease with height from the lowest flight level upward, and became less variable with increasing height. The leg-averaged SSA, b, and å did not change much with height, had smaller differences between the different temperature inversion scenarios for flight legs below 2 km height, and were the same as at the surface except that the SSA was significantly larger. The vertical distribution of aerosol properties under an EI was different from that above an EI. The results of the one-way ANOVA and the LSD test confirmed that all aerosol optical properties measured at the surface under an EI were not significantly different to those measured without a temperature inversion present, which is consistent with the results of Table 2 for aerosol number concentration.
Figure 11. Leg-averaged (a) σabs, (b) σsca, (c) σext, (d) SSA, (e) b, and (f) å. Blue, red, and black solid lines represent SBI, EI, and NTI episodes. Dark blue (red) and light blue (red) represent aerosol properties within (under) and above SBIs (EIs). Horizontal bars represent standard deviations.
For all flight legs under an EI, leg-averaged values of aerosol optical properties appeared to be consistent, except that there was an increase in σsca and σext above 1 km. Leg-averaged σabs, σsca and σext were significantly greater than those measured when there was no temperature inversion. This difference was not seen for SSA, b and å. For flight legs above an EI, all leg-averaged values of aerosol properties except b slightly decreased with height and were significantly lower than those measured when there was no temperature inversion. The impact of EI on the vertical distribution of aerosol properties can be partly explained by the fact that the height of the EI base denotes the mixing height after sunrise (Godowitch et al., 1985) and convective mixing leads to consistent aerosol properties in a well-mixed convective boundary layer. By contrast, it is the free atmosphere above EI where aerosol particles are much lower and decrease with height. The σabs, σsca and σext measurements made at the surface during SBI episodes were significantly higher than those during EI episodes and when there was no temperature inversion. These aerosol properties within an SBI tended to decrease with height and were significantly lower than those during EI episodes near the SBI top. This result is consistent with Fig. 10. The leg-averaged values of, and trends in, σabs, σsca, σext and SSA above an SBI were similar to those above an EI. Values of b and å above an SBI were almost the same as those without a temperature inversion present. In general, the differences between aerosol properties within (under) and above temperature inversions (both EIs and SBIs) were quite large, especially for σabs, σsca and σext.
Aerosol number concentration (cm−3) | ||||
SGP | 0530 LST | 1130 LST | 1730 LST | 2330 LST |
SBI | 4048/790 | −/− | 5055/158 | 4647/988 |
EI | 2836/1318 | 4455/3860 | 4510/2926 | 3187/1204 |
SBI/EI | 4053/2238 | −/− | 5668/505 | 4845/2365 |
NTI | 2713/39 | 4157/520 | 4599/923 | 3539/78 |
F-statistic | 153.6 | 1.0 | 16.8 | 156.9 |
LSD Test | MD & LSD | |||
SBI & NTI | 510 < 1333 | − | 554 > 455 | 488 < 1107 |
SBI & EI | 147 < 1210 | − | 563 > 544 | 170 < 1459 |
SBI & SBI/EI | 148 > 6 | − | 609 < 613 | 173 < 198 |
EI & NTI | 525 > 123 | 374 > 299 | 257 > 89 | 444 > 352 |
EI & SBI/EI | 120 < 1217 | − | 332 < 1158 | 157 < 1658 |
SBI/EI & NTI | 583 < 1340 | − | 363 < 1068 | 543 < 1305 |
SBI increase | 49.1% | − | 9.8% | 31.3% |
EI increase | 4.5% | 7.1% | −1.9% | −10.0% |
SBI/EI increase | 49.4% | − | 23.2% | 36.9% |