This section presents the results and discusses the findings of this study. Section 3.1 presents the temporal variation of NO2 concentration at the surface level using ground-based observation by AEROS, and Sect. 3.2 presents the temporal variation of aloft NO2 concentration based on MAX-DOAS. The model validations are also presented. Section 3.3 examines the diurnal variation of NO2 concentration at the surface and aloft. Section 3.4 considers the reason for episodic high concentrations of NO2 based on the meteorological field. Finally, Sect. 3.5 provides the correspondence between surface and aloft NO2 pollution as analyzed by the well-validated modeling simulation.
3.1 Temporal variation of surface NO2 concentration
The overall modeling performance during Chiba-Campaign 2015 was validated using ground-based AEROS observation. A comparison of the surface observation by AEROS located within 24 km (maximum horizontal distance measured by MAX-DOAS, see Fig. 1) of Chiba University and the model at the surface level is shown as a scatterplot in Fig. 3. During Chiba-Campaign 2015, surface NO2 concentration around Chiba University ranged mostly within 30 ppbv (dense circles in Fig. 3), and sometimes exhibited high concentrations of up to 70 ppbv. Over 20,000 pairs of observed and modeled hourly NO2 concentration, R of 0.50 showed a moderate linear correlation, NMB was + 9.8%, and NME was 55.9%. It has been reported that R ranged from 0.51 to 0.92 and NME ranged from 25–85% over China based on model results using three different emission datasets and their evaluation against the various surface observation categories (Liu et al. 2018). The Model Inter-Comparison Study for Asia (MICS-Asia) Phase III reported that R was 0.02–0.70 and NMB ranged between − 44% and + 36% (Kong et al. 2020). The results of this study demonstrate that the model could generally capture the surface NO2 concentration around Chiba University through Chiba-Campaign 2015.
For the purpose of detailed comparison using ground-based AEROS observations, the modeling performance at ground-based AEROS observations located along the path of 4AZ-MAX-DOAS measurements (see Fig. 1) was further evaluated. The temporal variations during Chiba-Campaign 2015 at these AEROS sites are shown in Fig. 4. Hereinafter, timeseries data are shown as local time (LT) which defined as Universal Time Coordinate (UTC) + 9 h. In the north, east, west, and south directions, a total of 5, 3, 1, and 4 sites were located along the measurement paths of 4AZ-MAX-DOAS, respectively. These AEROS sites are named in order of closeness from Chiba University in each direction (see Fig. 4). High NO2 concentration is defined as greater than 30 ppbv in this discussion, and is denoted by a yellow highlight at each site in Fig. 4. In the north direction (green in Fig. 4), R was around 0.5, NMB was within ± 25%, and NME was within 50%. At the N2 site, classified as an AEROS RAPMS site, the model performance was lower than that at other APMS sites in the north direction. Because the N1 and N2 sites are located close together, the modeling results were taken from same grid point in this comparison. Compared with the modeling performance found at N1, all statistical scores were worse, and model underestimation was identified at N2. This result indicates that even when the finer horizontal resolution of 1.3 km was used for air quality modeling, it was difficult to capture local NO2 pollution mainly caused by automobile sources. In the east direction (blue in Fig. 4), the comparison at the E2 and E3 sites which are classified as AEROS APMS sites was better than at the E1 site which was classified as an AEROS RAPMS site in terms of NMB and NME. The model also underestimated NO2 concentration at RAPMS sites, and this result in the east direction was similar to that found in the north direction. In the west direction (orange in Fig. 4), only a single APMS site close to Chiba University was available because this direction covers the Tokyo Bay area. The model performance at the W1 site was comparable to that at the N3 and E3 sites. In the south direction (red in Fig. 4), at the S1 site which is classified as an AEROS RAPMS site located near Chiba University, the model performance was mostly comparable to that found in the other directions. In contrast, the model generally tended to overestimate the surface NO2 pollution at the other sites S2, S3, and S4 in the south direction. At the S2, S3, and S4 sites, the model overestimation was suggested from NMB exceeding + 50%. As indicated from the temporal variation at S2, S3, and S4, the model overestimation was seen during 15–16, 17–18, and 20–21 November 2015, when NO2 concentrations were generally high. Because of this tendency of the model for overestimation in the south direction, model underestimation at RAPMS sites as found in the north and east directions was not detected. Based on these detailed comparisons of the temporal variation at AEROS sites located along the path of 4AZ-AMAX-DOAS, it was found that the model had some difficulties with underestimation when capturing local air pollution caused by automobile sources at RAPMSs sites in the north and east direction and overestimation in the south direction. Overall, the model generally captured the temporal variation of surface NO2 pollution around Chiba University along the path of 4AZ-MAX-DOAS.
From the temporal variation shown in Fig. 4, surface NO2 concentration typically showed diurnal variation with minima during the daytime and maxima during the nighttime. During Chiba-Campaign 2015, elevated NO2 concentrations were detected on 9, 15, 16, 17–18, and 20–21 November 2015, as distinguished by yellow highlights in Fig. 4. For the elevated NO2 concentrations on 9, 15, and 16 November, a short-time peak was found during the nighttime, whereas 17–18 and 20–21 November exhibited continuously higher concentrations over the whole day as the diurnal variations changed such that there was no daytime minima. Diurnal variation and episodic peaks of NO2 concentration are respectively discussed in Sect. 3.3 and 3.4.
3.2 Temporal variation of aloft NO2 concentration
The temporal variation of aloft NO2 concentration obtained by comparing 4AZ-MAX-DOAS and model results averaged over 0–1 km are shown in Fig. 5. The horizontal measurement distance of MAX-DOAS is also plotted. Since the horizontal distance viewed by MAX-DOAS is dependent on aerosol pollution, it varied from 4 to 24 km, with longer distances on clean days, such as during lower NO2 concentration periods on 11–13 November 2015. The modeled NO2 concentration within 24 km from Chiba University from model results is shown as the range in Fig. 5, and statistical analysis of the comparison with MAX-DOAS observation was conducted using the data within the observed horizontal distance to match the view by MAX-DOAS. From the range of modeled NO2 concentration within 4–24 km from Chiba University, the modeled NO2 pollution range within the horizontal direction of MAX-DOAS was generally small except for pollution events. This result implies that NO2 pollution around Chiba University was generally dominated by regional-scale broad pollution over the greater Tokyo area. In the north, east, and west directions, the range of modeled NO2 concentration was small except for high concentration events on 9, 15, 16, 17–18, and 20–21 November 2015. Because MAX-DOAS observation was limited to during the daytime, the higher NO2 concentrations at night on 9, 15, and 16 November were not measured. However, the cases of episodic high concentration on 17–18 and 20–21 November, which were characterized by continuous high concentration, corresponded well between MAX-DOAS and aloft model results. The modeling performance for the north, east, and west directions were R in the range 0.36–0.56, NMB around + 20%, and NME around 50%. These results are comparable with the surface results. In the south direction, the range of modeled NO2 concentration was also large during episodic NO2 pollution events, but also showed larger variation throughout Chiba-Campaign 2015 compared with the other directions. As found in the surface comparison, the model also tended to overestimate aloft NO2 concentration in the south direction. Although the value of NMB was larger in the south direction than in the other directions, NME was comparable and R was better compared with the other directions. In summary, it was validated that the model captured aloft NO2 concentration through comparison with 4AZ-MAX-DOAS observations.
3.3 Diurnal variation of surface and aloft NO2 concentrations
Sections 3.1 and 3.2 evaluated the temporal variations of surface and aloft NO2 concentrations and confirmed that the modeling system generally captured surface and aloft NO2 pollution. The diurnal variations are further analyzed here in Sect. 3.3. The surface and aloft diurnal variations in NO2 concentration averaged during Chiba-Campaign 2015 for the direction observed by 4AZ-MAX-DOAS are presented in Fig. 6. From the comparison at the surface level (left side of Fig. 6), the model could capture the observed diurnal variation, which consisted of a morning peak, subsequent decline during the daytime, and then an evening peak. In detail, the observed values showed morning maxima during 8 to 9 AM whereas the model showed morning maxima at the sunrise time of 6 AM. In contrast, for the evening peak seen after the sunset time of 5 PM, the timing of maxima was well reproduced by the model, particularly in the north and east directions. As found from the analysis of temporal variations shown in Fig. 3, the model tended to overestimate in the south direction, and this modeling overestimation was identified during the nighttime from the analysis of diurnal variation at the surface level. Apart from this difficulty, the model can capture the diurnal variation of NO2 concentration at the surface level. The evening peak was a good match between surface observation and the model whereas the morning peak showed a slight difference in timing. This suggests that investigation into the chemical mechanisms related to NOx and sunlight is needed in future research.
In the comparison of aloft NO2 concentration, all MAX-DOAS observations are plotted as light colors, and 1 h averaged data are plotted as dark colors (right side of Fig. 6). The hourly averages at 7 AM and 4 PM were not well counted (1–3 times through Chiba-Campaign-2015) and hence the 1 h averaged plot is not shown for these times. All 4AZ-MAX-DOAS showed a slight decrease in concentration during the daytime around noon, and the model reproduced these observations for all directions. As with the surface concentration in the south direction, the aloft NO2 concentration in the south direction was also overestimated. However, the temporal variation consisting of a drop around noon was captured. Based on this comparison, it was concluded that the diurnal variations were well reproduced by the model at both the surface and aloft. To our knowledge, the diurnal variation in surface and aloft NO2 concentrations has not been adequately compared in detail, and this study provides comprehensive analyses for verifying the air quality modeling performance at both the surface and aloft. Considering the better modeling performance in the north and east directions than in the south direction, one possible reason for the difficulty in modeling may be issues in the emission estimates. The coastline of the Tokyo Bay area has many industrial and power plants, which have relatively high stack heights, and the impact of these may be observed over the south direction from Chiba University. Because NO2 concentration was overestimated both at the surface and aloft, the emission amount itself might be overestimated. Moreover, chemical and physical processes related to NOx emissions from point sources could be related to the emission amount itself. These points should be explored in the future to improve the modeling performance around the greater Tokyo area.
3.4 Episodic peaks in NO2 concentration
The diurnal variation presented in Sect. 3.3 was the averaged characteristics during Chiba-Campaign 2015. As shown in the temporal variation presented in Figs. 4 and 5, some episodic peaks in NO2 concentration were detected on 9, 15, 16, 17–18, and 20–21 November 2015. Here in Sect. 3.4, the reason for the increased NO2 concentration is considered based on the meteorological field. The meteorological parameters measured by the Automated Meteorological Data Acquisition System (AMeDAS) of the Chiba special area meteorological observatory, which is located near Chiba University (see Fig. 1), and its modeling performance are shown in Fig. 7. In this figure, high concentrations of NO2 of more than 20 ppbv are also indicated by yellow highlights based on the averaged concentration of APMS sites located along the path of 4AZ-MAX-DOAS for clarity (see also Fig. 4). The meteorological field was well reproduced by the modeling system throughout the Chiba-Campaign 2015. The air temperature ( Fig. 7a), ranged from 10℃ to 20℃ with clear diurnal variation of daytime maxima. During Chiba-Campaign 2015, there were drops in temperature on 13 and 20 November 2015. The observed hourly daylight duration and modeled solar radiation reaching the surface level are plotted in Fig. 7b. These indicate the weather conditions, and the results show that 10, 13–14, and 20 November 2015 had cloudy conditions. Rainy conditions were clarified by analyses of precipitation as shown in Fig. 7c and relative humidity as shown in Fig. 7d. During Chiba-Campaign 2015, precipitation events occurred on 9, 10, 14, 15, and 19 November. In these precipitation events, relative humidity also exhibited high values of close to 100%. Wind components are presented as wind speed in Fig. 7e and wind direction in Fig. 7f. The wind speed was generally around 5 m/s and sometimes reached 10 m/s, and it was found that the high NO2 concentration (yellow highlight) correlated well with weak wind speed. Throughout Chiba-Campaign 2015, wind was in the north (0°) to east (90°) direction. However, wind changes to the west (270°) or northwest direction were found which corresponded to intense emission sources around the Tokyo Bay area. Therefore, the stagnant conditions in the transport of air mass from the dense emission sources around the Tokyo Bay area are thought to be the reason for increased NO2 concentration in terms of meteorological conditions. The highest NO2 concentration was seen on 17–18 November and the longest continuous NO2 concentration was found during 20–21 November with difference diurnal patterns; these are further discussed in Sect. 3.5.
3.5 Correspondence between surface and aloft NO2 pollution
Section 3.1 and 3.2 validated the modeling at the surface and aloft and Sect. 3.3 evaluated the diurnal variation at the surface and aloft by combining surface AEROS and aloft MAX-DOAS measurements. Section 3.4 examined the meteorological field causing episodic high NO2 concentrations, and stagnant conditions with a westerly wind direction can be considered as a reason for the high concentrations. Finally, here in Sect. 3.5, the relationship between surface and aloft NO2 pollution was analyzed based on well-evaluated modeling simulation results. The hourly NO2 concentration averaged during Chiba-Campaign 2015 is plotted at each hour for surface and aloft in Fig. 8. Throughout the day, high NO2 concentration was found over the Tokyo Bay area both at the surface and aloft. As seen for the diurnal variation reported in Fig. 6, the range of diurnal variation was larger at the surface than aloft. At the surface, high concentrations of greater than 20 ppbv (shown as orange to red in Fig. 8) were distributed over eastern Tokyo and Kanagawa prefecture and western Chiba prefecture, but these areas of high concentration were limited to only along the coastline of Tokyo Bay area from 10–14 LT. Compared with the features of the spatial distribution pattern of the high concentration area found at the surface, a moderately high concentration (shown as green in Fig. 8) was continuously seen aloft through the day. In order to find the relationship between the surface and aloft NO2 concentrations, linear regression was performed using all grid point data over this modeling domain for surface and aloft NO2 concentrations. The results are listed in Table 1 with domain averaged concentration (mean ± standard deviation), slope, intercept, and R. The results show clear linearity with R exceeding 0.918 through the day. The value of slope in the linear regression was around 0.4 during the nighttime and reached around 0.55 during the daytime. This result indicates that the aloft concentration was linearly scaled to 0.4 times the surface concentration during the nighttime and also linearly scaled to 0.55 times the surface level around noon (10 LT to 14 LT). The higher fraction during the daytime was due to the well-mixed air pollution within the PBL. This linear relationship between surface and aloft NO2 concentrations through the day was a key finding of this study; and this relation variated with a lower scaling value of around 0.4 during the nighttime and a higher value reaching 0.55 during the daytime.
Table 1
Hourly correspondence between modeled surface and aloft NO2 concentrations (averaged over 0–1 km)
LT
|
Surface
|
0–1 km
|
Slope
|
Intercept
|
R
|
0
|
8.73 ± 7.00
|
4.37 ± 2.88
|
0.398
|
0.897
|
0.966
|
1
|
8.33 ± 6.68
|
4.15 ± 2.82
|
0.409
|
0.753
|
0.968
|
2
|
8.19 ± 6.72
|
4.03 ± 2.91
|
0.418
|
0.604
|
0.966
|
3
|
8.31 ± 6.91
|
4.00 ± 3.03
|
0.421
|
0.503
|
0.959
|
4
|
8.72 ± 7.04
|
4.08 ± 3.09
|
0.415
|
0.459
|
0.945
|
5
|
9.65 ± 7.45
|
4.28 ± 3.14
|
0.391
|
0.511
|
0.926
|
6
|
10.99 ± 8.10
|
4.65 ± 3.26
|
0.370
|
0.577
|
0.918
|
7
|
10.57 ± 7.44
|
4.67 ± 3.11
|
0.391
|
0.536
|
0.933
|
8
|
9.50 ± 6.59
|
4.64 ± 3.08
|
0.447
|
0.394
|
0.957
|
9
|
8.63 ± 6.13
|
4.63 ± 3.10
|
0.493
|
0.381
|
0.972
|
10
|
7.84 ± 5.80
|
4.55 ± 3.17
|
0.531
|
0.382
|
0.975
|
11
|
7.15 ± 5.42
|
4.39 ± 3.12
|
0.558
|
0.397
|
0.969
|
12
|
6.72 ± 4.95
|
4.23 ± 2.92
|
0.567
|
0.417
|
0.962
|
13
|
6.76 ± 4.89
|
4.26 ± 2.86
|
0.561
|
0.462
|
0.958
|
14
|
7.12 ± 5.19
|
4.44 ± 2.98
|
0.548
|
0.536
|
0.954
|
15
|
8.12 ± 6.04
|
4.91 ± 3.32
|
0.523
|
0.654
|
0.951
|
16
|
10.04 ± 7.83
|
5.52 ± 3.87
|
0.469
|
0.809
|
0.950
|
17
|
11.24 ± 8.96
|
5.64 ± 3.99
|
0.424
|
0.881
|
0.952
|
18
|
11.44 ± 8.85
|
5.58 ± 3.81
|
0.408
|
0.913
|
0.947
|
19
|
11.61 ± 8.66
|
5.49 ± 3.53
|
0.383
|
1.038
|
0.940
|
20
|
11.47 ± 6.34
|
5.32 ± 3.28
|
0.371
|
1.063
|
0.942
|
21
|
10.79 ± 7.93
|
5.03 ± 3.17
|
0.380
|
0.933
|
0.950
|
22
|
9.82 ± 7.69
|
4.71 ± 3.11
|
0.388
|
0.902
|
0.959
|
23
|
9.23 ± 7.60
|
4.53 ± 3.10
|
0.393
|
0.903
|
0.964
|
Note: Total grid numbers in modeling domain are 11025. |
The previous discussion focused on the diurnal variation. Because episodic high concentrations were found during Chiba-Campaign 2015, the daily-averaged surface and aloft NO2 concentrations over the modeling domain were further analyzed in order to investigate the temporal change in the linear relationship between surface and aloft. The daily average was based on the average over 0–23 LT. The spatial mappings of each day are plotted in Fig. 9, and the period average during Chiba-Campaign 2015 are also plotted at the bottom-right corner of Fig. 9. NO2 concentration showed day-to-day variation and the spatial distribution pattern between surface and aloft are in generally good agreement. For example, in the case of high NO2 concentration on 9 November 2015 as suggested by temporal variation (Figs. 4 and 5), the higher NO2 concentration stretched into northern Chiba and up to Ibaraki and Saitama prefectures at the surface level, and this feature was also seen aloft. On 9 November 2015, a southerly wind close to 10 m/s prevailed (Fig. 7). In another case of high NO2 concentration during 20–21 November 2015, higher NO2 concentrations were found over a broad area of Kanagawa prefecture, eastern Tokyo, and western Chiba prefecture on these days at the surface, and the aloft concentration also showed a similar spatial distribution. To confirm these characterizations of the surface and aloft NO2 pollution, linear regression was performed for each day and period average in the same way as for the diurnal variation listed in Table 1. The result for daily and period averages are listed in Table 2. The daily averaged concentrations varied from 6 ppbv to 15 ppbv during Chiba-Campaign 2015. In spite of this variation in concentration, comparison of surface and aloft found an R exceeding 0.924, which suggests clearly linear correspondence between surface and aloft NO2 concentrations through the Chiba-Campaign 2015. Although the intercept of the linear regression ranged between 0.0 and 0.75, it was found that the slope ranged within 0.4–0.5 except on 16 November 2015 when it was 0.37. This result indicates that the aloft NO2 concentration is linearly scaled to 0.4–0.5 times the surface NO2 concentration on both clean and polluted days.
Table 2
Daily/period averaged correspondence between modeled surface and aloft NO2 concentration (averaged over 0–1 km)
Date
|
Surface
|
0–1 km
|
Slope
|
Intercept
|
R
|
9 November 2015
|
9.49 ± 9.21
|
4.65 ± 4.69
|
0.483
|
0.061
|
0.949
|
10 November 2015
|
9.04 ± 6.87
|
5.15 ± 3.50
|
0.486
|
0.752
|
0.954
|
11 November 2015
|
6.13 ± 6.00
|
3.46 ± 3.03
|
0.478
|
0.530
|
0.946
|
12 November 2015
|
5.97 ± 6.31
|
3.15 ± 3.16
|
0.469
|
0.361
|
0.933
|
13 November 2015
|
6.61 ± 6.58
|
3.40 ± 3.28
|
0.464
|
0.329
|
0.932
|
14 November 2015
|
6.17 ± 6.26
|
3.28 ± 3.17
|
0.478
|
0.329
|
0.944
|
15 November 2015
|
13.36 ± 8.65
|
6.62 ± 4.23
|
0.470
|
0.332
|
0.961
|
16 November 2015
|
9.30 ± 8.51
|
4.17 ± 3.40
|
0.371
|
0.714
|
0.928
|
17 November 2015
|
12.11 ± 10.11
|
6.19 ± 5.16
|
0.483
|
0.347
|
0.946
|
18 November 2015
|
12.97 ± 7.66
|
6.54 ± 3.96
|
0.496
|
0.107
|
0.959
|
19 November 2015
|
5.50 ± 5.44
|
2.95 ± 2.62
|
0.448
|
0.485
|
0.927
|
20 November 2015
|
15.48 ± 12.15
|
7.34 ± 6.22
|
0.474
|
−0.001
|
0.926
|
21 November 2015
|
8.97 ± 7.13
|
4.48 ± 3.24
|
0.425
|
0.674
|
0.935
|
22 November 2015
|
7.81 ± 7.81
|
4.01 ± 3.90
|
0.461
|
0.409
|
0.924
|
Period average
|
9.20 ± 6.74
|
4.67 ± 3.06
|
0.436
|
0.665
|
0.961
|
Note: The total number of grid cells in the modeling domain is 11025. |
Finally, the vertical profile of NO2 concentration from the model simulation is shown in Fig. 10. This curtain plot shows the vertical structure for NO2 concentration during the Chiba-Campaign 2015 with modeled PBL height at the grid point of Chiba University. The PBL height showed variation from the near surface up to 1.0 km with a general diurnal profile featuring daytime maxima. As discussed for temporal variations in the surface and aloft NO2 concentrations, high NO2 concentration events were found on 9, 15, 16, 17–18, and 20–21 November 2015. The vertical profiles show that high NO2 concentrations (red in Fig. 10) were mostly limited within 0.2 km, and also within the PBL. In the vertical direction, NO2 concentrations were lower than 5 ppbv (light blue in Fig. 10) above 0.5 km, even though PBL height sometimes reached up to 1.0 km. Note the highest NO2 concentration on 17–18 November 2015 and the longest continuous NO2 concentration during 20–21 November 2015. During these events, the PBL height was close to the near-surface level. Taking into consideration analyses of the meteorological field as shown in Fig. 7, the stagnant air mass with weak wind speed and cloudy conditions with a lower PBL close to the surface were possible factors causing the increased NO2 concentration. Through the validated modeling performance, the vertical NO2 profile with its important connection to the PBL height was clarified.