The contribution of transport and chemical processes on coastal ozone and emission control strategies to reduce ozone

Abstract

The interaction between transport and chemistry is pivotal for local ozone (O₃) concentration, especially for a coastal region where the upstream sources might change diurnally. In the current emission control policy, most pollutants, such as particulate matter, SO₂, NOx, and CO, decrease while the annual O₃ trend might increase due to the complex feedbacks of precursors. In this study, we investigate the influence of transport upon the wintertime O₃ diurnal trend over ZuoYing Kaohsiung, an urban coastal site in southern Taiwan, by constructing a two-dimensional numerical model coupling both physical mechanisms and core chemical processes and provide a feasible emission control strategy. The transport process (i.e., import vs. export) for the daytime is determined using the Leighton Ratio (Φ), the ratio of O₃-production over O₃-loss rate, under the pseudo-steady-state condition. Φ shows a deviation of -9 to +13% from the photo-stationary state, and experiences a transition from import effect before 10:15 to weakening import or net export effect afterward associated with a net O₃ production as sea breeze starts developing. The significantly higher Φ derived from observation than from simulation by a factor of 1.35 might be resulted from the over-reported NO₂ due to NOy contribution on the NO₂ measurement, and the influence of aerosol and cloud possibly reducing ∼30% on applied NO₂ photolysis rate constant, associated with aerosol optical depth of 0.75 ± 0.15 and single scattering albedo of 0.85 ± 0.15. In this studied NOx-saturated regime, the addition of sea breeze convergence over the land enhances the maximal O₃ by ∼10%, mainly due to the O₃ accumulation (∼88%). Furthermore, the ozone isopleth analysis as a function of non-methane hydrocarbons and NOx emissions provides an achievable strategy to decrease both maximum daily ozone and the increment of ozone from morning to maximum by reducing hydrocarbons and NOx emissions, which can also eliminate the additional nitrate contribution on the aerosols.

Highlights

• •Leighton Ratio variation illustrates the transition of the ozone transport process.
• •Transport causes the noontime O₃ time lag under pseudo-steady state conditions.
• •AOD and SSA are derived under the apparent photo-stationary condition.
• •Sea breeze convergence increases O₃ primarily via physical accumulation.
• •A feasible emission control strategy is proposed by superposed isopleth diagrams.

Ozone, Photochemistry, Leighton Ratio, Sea Breeze, Ozone Isopleth

1. Introduction

Surface ozone (O₃) is one of the strong atmospheric oxidants and produced via photochemical processes driven by a well-known NOx-VOCs (volatile organic compounds) framework happening mainly in urban areas [1,2]. High O₃ concentration is harmful to human health and has negative impacts on the environment [3, 4, 5]. Sustainable Development Goals 11 in 2019 (SDG11) stated that 90% of humans living in the cities did not enjoy good air quality, which meets the World Health Organization's (WHO) guideline value [6,7]. The government has applied regulations on controlling anthropogenic emissions, and the long-term concentrations of CO and NOx have revealed a decreasing trend [8]. However, the O₃ concentration may not have a corresponding response due to the increasing background O₃ concentration [9] and the complex feedbacks of precursor gases [10-12]. In addition to the photochemistry, physical mechanisms including the transport process and the evolution of planetary boundary layer height (PBLH) can also play crucial roles in local surface ozone levels by ventilation or accumulation [13-15].

The classical Leighton relationship is frequently applied to estimate O₃ concentration under the photostationary state as follows:

[O₃] = j_NO2 [NO₂]/(k_NO-O3 [NO]) (Eq. 1)

where k_NO-O3 is the rate constant for NO + O₃, j_NO2 is the photolysis rate constant of NO₂. The Leighton relationship holds when ozone is balanced by the rapid chemical production and loss processes [1]. Many studies addressed the ozone problem by assessing NO/NO₂; computation modeling relies on the Leighton relationship to lower the computational cost [1]. The Leighton Ratio (Φ) is defined as

Φ = j_NO2 [NO₂]/(k_NO-O3[NO][O₃]) (Eq. 2)

and is ∼1 for a system composed of NOx and O₃ only. For a zero-dimensional box model, the photo-stationary state can be easily reached with Φ = 1 due to the fast chemical processes. The deviation of Φ from 1 might be caused by the additional pathways such as the conversion of NO to NO₂ in the presence of RO₂, or O₃ loss reactions other than NO + O₃ to have Φ > 1 while a significant local NO emission might lead to Φ < 1 [1, 16]. The sufficiently slow photolysis of NO₂ at sunrise and sunset might cause the failure of the photo-stationary state [17]. Moreover, Khalil suggested the deviation of Φ from 1 as an indicator to interpret the transport contribution for an open system using the observation data in Yanbu [15]. He pointed out the invalidity of photo-stationary state in real cases and suggested that the Leighton Ratio, defined as the rate ratio of O₃-production/O₃-loss, can be applied to differentiate the role of advection and chemistry under the pseudo-steady state assumption [14]. In other words, Leighton Ratio describes the deviation of O₃ from the hypothetical photo-stationary state of the system composed of NOx and O₃ mainly due to transport. If Leighton Ratio is larger than 1, the chemistry favors the O₃ production, which suggests a net export of O₃ to sustain the pseudo-steady state. On the other hand, if the net transport O₃ is positive (i.e., flux in > flux out), chemistry production is weakened to lead Leighton Ratio less than 1 in response [14, 15].

The air characteristics of the marine atmosphere are typically different from that of the inland atmosphere due to the lack of human activities [18]. The air quality of coastal regions may be subject to the sea-land breeze circulation, and the influence is estimated to be a few kilometers. On one hand, the developed sea breeze together with coastal outflow may ventilate pollution from the boundary layer even under high-pressure days [19]. On the other hand, in some areas, the sea breeze leads to severe ozone episodes under some meteorological conditions [18, 20]. Due to the complex feedbacks of precursor gases on ozone production, the causal linkage of local emissions and ozone concentration relies on the local environment and meteorological conditions.

Taiwan is an island with Central Mountain Range running from the North to the South. Approximately 90% of the population lives on the west side, having flat to gently rolling plains. In the wintertime, Taiwan is under the influence of northeast monsoon, which might bring polluted air to the residential area in the downwind [21, 22]. Kaohsiung, a heavy industrial city in southern Taiwan, suffers from severe air pollution issues and has the worst air quality in Taiwan, especially in winter, due to the meteorological conditions, such as low PBLH, persistent but weak north wind as monsoon, and limited rainfall, leading to poor air quality [22, 23]. Despite the diminishment of some pollutants such as particulate matter (PM), SO₂, and CO, ambient O₃ concentration has been revealed an increasing trend over the past decades [24, 25, 26]. Nonetheless, the increasing O₃ trend may be viewed as a penalty of NOx reduction [27] and the daily maximum O₃ increment can serve as an indicator to evaluate the effect of NOx reduction [27, 28, 29]. Hence, the increased O₃ and decreased maximum O₃ increment may imply a successful NOx reduction with overall O₃ concentration restoring the background value. Additionally, the decreased maximum O₃ increment reduces the hospital admission rate since the short-term enhancement in ozone corresponds to increased hospital admissions of respiratory diagnosis [30]. The previous study relies on a synoptic model or statistical approaches [27, 31, 32]; however, there is less investigation of the detailed causal relations.

To characterize the interaction of the ozone transport and chemical production, this study constructs the emission diurnal pattern of precursors and the reaction processes over a coastal region (ZuoYing area as shown in Figure 1) in southern Taiwan using a two-dimensional numerical model with time-dependent PBLH. Leighton Ratio is applied to distinguish the chemical and physical processes. Moreover, the influence of local emission and convergence of sea breeze upon winter ozone is discussed for the regional characteristics. With the analysis, a feasible local emission reduction strategy on precursors is proposed by analyzing the maximum daily O₃ and the daily maximum O₃ increment under different precursor emission scenarios.

Figure 1.

The map of Taiwan with pins specifying the location of TWEPA sites as stated in section 2. The red-framed area is the studied domain.

2. Methodology

A two-dimensional numerical model, consisting of both physical mechanism (including advection, emission, and entrainment) and core chemical processes, was developed using Matlab2019b (The MathWorks, Inc) to simulate the mean winter diurnal O₃ profile in Southern Taiwan. The simulated results are compared with the observation of the ZuoYing Environmental Protection Administration (EPA) site (22.6749 °N, 120.2929 °E), ∼2.8 km from the sea, in 2018 winter. The developed model is further used to evaluate the contribution of transport and emissions.

2.1. Model description and setup

For a small domain study with no topography and the resident time of nonvanishing advection process <6 hr, the concentration X_i of a given species i (in a unit of molecules/volume) in the Eulerian framework under the well-mixing assumption is governed by the differential mass balance equation,

$$\frac{\partial X_i}{\partial t}=-\nabla \cdot \left(X_i\mathit{u}\right)+\frac{E{m}_i}{h}+\frac{\left(X_{B,i}-X_i\right)}{h}{h}^{\prime }\theta \left(h^{\prime }\right)+(P_i-L_i)$$ (Eq. 3)

with negligible diffusion and dry decomposition, where u is the wind vector, $\nabla$ is the gradient, Em_i is the local emission, X_B,i represents the background concentration, h and h' are the mixing layer height and its time-derivative, respectively, θ is the Heaviside step function, and P_i and L_i stand for the chemical production and loss, respectively. Simply speaking, the four terms in Eq. (3) correspond to the advection, emission, entrainment, and chemistry processes, respectively. Emission term with a unit of molecules km⁻² hr⁻¹ describes the anthropogenic source from the surface. Entrainment characterizes the effect from the free troposphere when the mixing layer is rising, i.e., θ(h') = 1 if h' ≥ 0 and θ(h') = 0 otherwise. The detailed chemical processes are illustrated in section 2.1.1. In this study, advection and chemical production are solved using the backward Euler method, while other processes are treated by the forward method.

The model takes time-dependent meteorological parameters (including wind field, temperature, and relative humidity (RH)), PBLH, emission profile, and the boundary and background conditions of each measured species (CO, NMHCs, CH₄, NO, NO₂, and O₃) as input for Eq. (3). The boundary and background conditions for non-measured species (e.g., radicals and HNO₃) are assumed to be zero. The spatial domain of this study takes over the region of 6.65 km × 7.5 km (22.7042 ± 0.0295 °N, 120.2928 ± 0.0354 °E) as shown in the map of Figure 1, and is partitioned into 11 × 11 rectangles with a spatial resolution of 604 m × 680 m. Because the system is independent of the initial condition after 6-hr simulation, the model is implemented at a 0.01-hr time step for a 48-hr simulation with the last 24-hr data for further analysis.

2.1.1. Core ozone chemical processes

The ozone-related chemistry involves the well-known NOx-VOCs (or CO) framework [2, 11, 15, 33]. In this study, only gas-phase reactions are taken into accounts for simplification since the resident time via advection is approximately 1 hr. The daytime O₃ is governing by the following cycle:

NO + O₃ → NO₂ + O₂ (R1)

$${\text{NO}}_2+\text{hv}\stackrel{{\text{O}}_2}{\to }{\text{NO}+\text{O}}_3$$ (R2)

P_i + OH → → → RO₂ + other molecules (R3)

$${\text{RO}}_2\text{ }+\text{ NO}\stackrel{{\text{ O}}_2}{\to }{\text{RO}+\text{NO}}_2$$ (R4)

RO → → → HO₂ + other molecules (R5)

where P_i denotes the precursor of i, such as CO, CH₄, or non-methane hydrocarbons (NMHCs), and RO₂ denotes HO₂ or organic peroxyl radicals. Propene is applied as the representative NMHCs to initialize the chain reactions during daytime [2]. On the other hand, nighttime O₃ is dominated by the titration of NOx via R1 and R6.

NO₂ + O₃ → NO₃ + O₂ (R6)

The rate constants of applied major reactions are summarized in Table S1. The photolysis rate constants are assumed to be proportional to the cosine of solar zenith angle and calculated with the reported actinic flux [1,2], while the cross-section and quantum yield are obtained from SAPRC-07 [34].

2.1.2. The use of observation data

The applied diurnal meteorological parameters are the mean winter condition of the ZuoYing EPA site as shown in Figures 2a and 2b. Temperature is in the range of 18–24 °C while the RH is 65–85%. There is ∼1.73 m s⁻¹ of the prevailing north wind mainly due to monsoon, while the sea-land breeze circulation induces the west wind in the range from -0.19 m s⁻¹ to 2.36 m s⁻¹. The applied time-dependent mixing layer height shown in Figure 2c is based on the analysis of the Central Weather Bureau database [35]. Except for the case of sea-breeze convergence stated in sections 2.3 and 3.4, all meteorological parameters within the studied domain are assumed to be spatially invariant but time-dependent.

Figure 2.

Applied meteorological parameters and emission profile in the model. (a) Temperature (T) and relative humidity (RH), (b) wind speed (WS), (c) mixing layer height (MLH) [35], and (d) CO emission profile with a unit of 10⁴ mol km⁻² hr⁻¹.

The boundary and background conditions of measured chemical species are based on the observation results (Figure S1) of the nearby Taiwan EPA sites with PengHu data for the whole west side, NanZih data for the northeastern corner, and FuGuiJiao data as the background value. For NMHCs and CH₄, the west side boundary data are adapted from FuGuiJiao's data due to no measurement in PengHu. The northern boundary is estimated using linear interpolation between PengHu (or FuGuiJiao) and NanZih data (locations are shown in Figure 1). The wind field applied in this study has no south wind and the span of east wind is within an hour, which has non-observable influence over ZuoYing, so the concentration of the south side is assumed to be zero and the east side in the domain is extended as the east boundary condition.

2.1.3. Emission profile

The precursor emissions for Kaohsiung summarized in Table 1 are based on the annual emission from Taiwan Emission Data System (TEDS9.0) reported by Taiwan Environmental Protection Administration (TWEPA) [36] by assuming the same daily pattern through the whole year. To downscale the annual emission of Kaohsiung into a small domain, the relative daily emission trend is firstly retrieved based on the following assumptions, 1) the vehicle source is proportional to transportation flux reported by the local governments, 2) the industrial source is independent of time and the other sources are modeled by a sinusoidal wave, and 3) the relative amount of emission matches the emission ratio in Table 1. Then the diurnal emission rate is quantified by minimizing the deviation of modeled nighttime CO concentration. In particular, the studied domain is ∼10 km away from the heavy industrial parks with limited diesel trucks passing, which might lead to a lower ratio of NO over CO emissions than the reported ratio in Table 1. Thus, an extra emission adjustment for NOx by a factor of 0.1 of the averaged Kaohsiung emission rates (Table 1) is applied to provide a similar quantity of NOx as observation. Similar daily emission adjustment was applied in the Community Multiscale Air Quality Model (CMAQ) study by Lin et al. [37] but with lower diurnal amplitude. For NOx emission, 90% is assumed to be NO, while the remaining 10% is NO₂. As an example, the CO emission profile is shown in Figure 2d.

Table 1.

Emission in 2013 over Kaohsiung [36].

Sources∖Species (103 kg yr−1)	CO	NMHCs	NOx
Industry	19092	18428	38433
Vehicle	78784	20594	27384
Non-road transportation	911	270	5827
Commercial	937	13338	260
Construction	0	5884	0
Burning	2553	282	349
Other	941	443	1808

Note: In the model, sources other than Industry and Vehicle in Table 1 are summed up and considered as "Others".

2.2. Leighton Ratio

To evaluate the contribution of chemical processes or transport to the O₃ profile, the Leighton Ratios for both observation (j_NO2 adapted from the model) and simulation as a function of time are estimated using Eq. (2). In addition, a revised Leighton Ratio as the ratio of the overall production rate over the overall loss rate of O₃ derived from the model in the following [38]:

Φ = ∑Rate_Prod/∑Rate_Loss (Eq. 4)

reduces the influence of chemical processes such as NMHCs. In this study, the pseudo-steady state condition holds during the daytime, with $\left|\frac{\partial O_3}{\partial t}\right|$less than 3.5% of the major transport and chemical processes, fulfilling the validity of a pseudo-steady state with $\left|\frac{\partial O_3}{\partial t}\right|$< 10% of major terms in Eq. (3) as stated by Khalil [15]. The detailed information regarding the contribution of each term is shown in Figure S2. As a remark, we emphasize Khalil's framework in which the Leighton Ratio is reinterpreted as the contribution of the transport process since O₃ is controlled only by transport and chemical processes. In addition to the advection, the Leighton Ratio should be more precisely reinterpreted as the contribution of nonchemical processes due to the additional entrainment term in Eq. (3) in our framework. Nevertheless, the entrainment term is considerably smaller than others, as shown in Figure S2a.

2.3. Sensitivity analysis

With the established model, we multiply a factor from 0 – 2.0 to NMHCs and NOx emission, individually, in the “Full Processes” case to evaluate the influence of NOx-VOCs framework on the maximum daily O₃, denoted as O₃(Max), and the maximum O₃ increment from 8:00 to the maximum O₃ of the day, denoted as ΔO₃ (i.e., ΔO₃ = O₃(Max) - O₃(8:00)). Furthermore, the measured u-wind over RenWu (an EPA site, 4.1 km in the east of ZuoYing) of 0.58 m s⁻¹ smaller than that over ZuoYing during 14:00–15:00, might suggest a significant convergence as wind traversing toward the east. The convergence is estimated at ∼ -1.41 × 10⁻⁴ s⁻¹ in the east-west direction likely due to the land roughness and the interaction with the northerly monsoon. For a clean marine condition, sea breeze might ventilate the pollution and act as a factor causing convergence to accumulate pollution in ZuoYing. The impact of convergence on O₃ concentration is evaluated by assuming that a hypothetical convergence happens at u > 0.5 m s⁻¹ and is proportional to the development of sea breeze, i.e., D_max × (u - 0.5)/(u_max - 0.5) with D_max = - 5.2 × 10⁻⁵ s⁻¹ as the maximum convergence.

3. Results and discussion

3.1. Model results

Figure 3 shows the diurnal variation of the monitored species for both observation and simulation results. The simulated emitted species considering all processes (i.e., “Full Processes” case) show a similar diurnal trend with the observation and have better consistency in the nighttime but underestimation over the afternoon period. Table 2 summarizes the correlation (r), the normalized mean bias factor (B_NMBF) and the normalized mean absolute error factor (E_NMAEF) [39] between simulation and observation. The modeled values are highly correlated to observation with r > 0.90 for most species, but a lower correlation r = 0.78 for NO₂ due to more substantial deviation during noontime. The negativity of B_NMBF indicates the overall underestimation of simulation, which might be attributed to the model well-mixing assumption. The emission is mainly sourced from the surface level and requires time or turbulence motion to mix. The concentrations, in general, have a decreasing trend with respect to altitude even within PBL. The well-mixing assumption averages the higher concentration near the observation sites (∼15 m above the ground level) along with the decreasing profile within the PBLH. Therefore, it might be expected that the observed concentrations are higher than the simulated concentrations. Additionally, part of NMHCs underestimation might arise from the lack of natural emission sources in TEDS9.0. E_NMAEF of CO, NMHCs, CH₄, NO₂, and O₃ has a value <0.22 suggesting the bias is acceptable. As the only exception, NO seems to have a larger E_NMAEF due to its significantly low concentration (≤1 ppbv) during nighttime, near the instrument detection limit of 0.5 ppbv as summarized in Table S2 [40]. Overall, the simulation reveals major ozone characteristics, the start of ozone elevation and the peak occurrence time, which are used in the following analysis.

Figure 3.

The simulation results of two cases ("Full processes" and "No Chemistry") and observation over ZuoYing for (a) CO, (b) CH₄, (c) NMHCs, (d) NO, (e) NO₂, and (f) O₃. The error bar of observation stands for ± σ and the average is taken after removing outlier (>± 1.5σ). "Full Processes" considers both physical processes and chemistry, while "No Chemistry" only takes physical processes into account. The crimson dashed line in (d) and (e) corresponding to the right axis is the simulation result with no NOx correction factor, as illustrated in section 2.1.

Table 2.

The normalized mean bias factor of each monitor species.

Species∖metrics	r	BNMBF	ENMAEF
CO	0.9099	-0.0604	0.0840
NMHCs	0.9500	-0.2689	0.2119
CH4	0.9508	-0.1046	0.0097
NO	0.9738	-0.1046	0.4335
NO2	0.7828	-0.1973	0.2200
O3	0.9917	-0.0126	0.1400

The normalized mean bias factor (B_NMBF) and the mean normalized absolute factor error (E_NMAEF) [39] are calculated as follows.${\text{B}}_{\text{NMBF}}=\text{S}\left(\exp \left|\ln \frac{\sum M_i}{\sum O_i}\right|-1\right)$${\text{E}}_{\text{NMAEF}}=\frac{\sum |M_i-O_i|}{{\left(\sum O_i\right)}^{\frac{1+S}{2}}{\left(\sum M_i\right)}^{\frac{1-S}{2}}}$ and $\text{S}=\frac{\sum M_i-\sum O_i}{\left|\sum M_i-\sum O_i\right|}$ where M_i and O_i is the modeled and observed values, respectively, and the summation is taken over the samples over the day.

The concentrations of CO, CH₄, and NMHCs are relatively unaffected by the chemical processes in this small domain study since their lifetimes in the daytime are 60.8 days, 6.6 years, and 13.5 hr, respectively, significantly longer than the residence time of air in the domain, ∼ 1 hr. Even though NMHCs have the lowest concentration compared with CO and CH₄, it is the major contributor to ozone production due to the higher reactivity. In contrast, NO, NO₂ and O₃, having short lifetimes responding to the chemical processes (54 s, 2.7 min, and 8.7 min, respectively), are subject to chemistry even within the applied small domain. As anthropogenic activities progress, such as morning traffic, precursors can reach their maximum in the morning due to the lower PBLH. For example, NO, despite titrated by O₃, reveals a peak in the morning [41]. On the other hand, O₃ achieves the maximum around 14:00 due to the combined effect of chemistry and transport, which is discussed in section 3.2. The relatively low simulated precursor concentrations in the afternoon are likely contributed by the high PBLH with the well-mixing assumption and additional adjustment through the convergence as discussed in section 3.4.

3.2. Ozone profile

3.2.1. Time-lag

Photochemistry is a key process to produce O₃ in the troposphere. However, the ambient O₃ profile usually shows a certain time lag with solar radiation [42]. As shown in Figure 4a, there is a 2-hr time-lag between the simulated O₃ profile and the applied NO₂ photolysis rate constant, j_NO2. This 2-hr time-lag can be deconvoluted into the contribution of several factors such as the transport, transformation of other chemical forms, and the quantity of oxidation production.

Figure 4.

(a) The diurnal profile of j_NO2 (dotted line), the observation (line with error bar, as in Figure 3) and simulated (solid line) O₃, and the simulated Ox (dashed line). (b) Leighton Ratio Φ profile for observation (solid line, via Eq. 2), for observation with a 15-min j_NO2 shift (dotted line, via Eq. 2), and for simulation results via Eq. 2 and Eq. 4. The gray dashed line specifies Φ = 1.

First, under no wind condition, the model almost reduces to a chamber system. It shows a 30-min time-lead than the maximum radiation (Figure S3), possibly due to the time-dependent emissions and PBLH. On the other hand, as all chemical processes are turned off, O₃ achieves a peak value at 15:15 due to transport as shown in Figure 3f ("No Chemistry" case). In this case, O₃ is governed by mere advection with negligible emission and entrainment terms. Intuitively, the two-dimensional wind field blows the northern boundary conditions (with peak ∼ at 14:00) to ZuoYing, with the resulting O₃ profile delayed due to transportation time. Overall, the “Full Processes” case shows the combined effect of chemical processes (peak at 11:30) with the transport (peak at 15:15) leads to the peak at 14:00.

The chemical processes in the NOx-VOCs framework generate the HOx (i.e., OH + RO + RO₂) cycle, which not only produces O₃ but also partially stays as NO₂ for approaching the photo-stationary state via the fast NO + O₃ reaction. Some HOx might be terminated as HNO₃ or H₂O₂, having lower reactivity to affect ozone production. To access the net photochemical formation of O₃, we modified the definition of Ox as Ox = O₃ + NO₂ + 2×NO₃ + 3×N₂O₅ + OH + 1.5×HNO₃, stated by Womack et al. [43]. Ox has a maximum value around noontime as shown in Figure 4a and the model result suggests that the major contributors of Ox are O₃ and NO₂.

The transport pattern over ZuoYing, together with the composition of Ox, explains the O₃ time-lag. During the period from 12:00 to 14:00, O₃ is exported by the possible ventilation effect of the sea breeze as Φ > 1 (see section 3.2.2). To retain the pseudo-steady state, it results in higher NO₂ concentration to raise the production rate, balancing the exported O₃. Therefore, despite the Ox maximum at noon, it partially stays in the form of NO₂ instead of converting to O₃. Likewise, a higher reactivity or a higher amount of reactive NMHCs can shift the O₃ peak toward noontime (Figure S3) by adjusting the distribution of NOx-O₃ while retaining the pseudo-steady-state. Overall, the observed and simulated postponed O₃ peak away from the maximal photolysis rate is caused by the combined effect of transport and chemical processes (i.e., the production efficiency and the distribution of Ox to various species other than O₃).

3.2.2. Leighton Ratio

With the detailed chemical processes simulation, the derived Leighton Ratio (Φ) using Eq. (4) is less than that using Eq. (2) as shown in Figure 4b. That is mainly caused by an additional contribution of O₃ loss reactions (reactions 3, 5, 6, 12a, and 28, in Table S1) other than NO + O₃ in the denominator of Eq. (4). Nevertheless, the overall trends of Φ between Eqs. (2) and (4) are consistent despite the slightly different transition time for Φ = 1. The simplified version from Eq. (2) gives a fair estimation as compared with that derived from Eq. (4) to evaluate the pseudo-steady state of O₃ and to illustrate the net transport, T_net (= Flux in - Flux out) (Figure S2b), so Φ discussion in the rest of this study is based on Eq. (2).

Figure 4b shows that the simulation has daytime Φ in the range of 0.9–1.13, slightly less than 1 in the morning (∼0.9 at 7:00), and Φ slowly increases with time to have a value larger than 1 starting at ∼ 10:15. Φ ∼ 1 might imply that NOx-O₃ chemistry tends to adjust itself via photolysis and titration to approach the photo-stationary state and the deviation from unity accounts for the pseudo-steady state with the transport process in the environment. With the condition of O₃ in the pseudo-steady state, the Φ transition suggests a net chemical consumption with a net import for O₃ in the morning but a net production combining a net export in the afternoon. In other words, the transport process compensates for the chemical loss in the morning while exports the produced O₃ out of the box to retain the pseudo-steady state for O₃ [14,15]. From a physical perspective, the Φ transition time almost coincides with the development of sea breeze starting at 10:00 (u-wind in Figure 2). Before the transition, the upstream source brings higher O₃ into the box to have slightly more titration, resulting in Φ < 1, while the established sea breeze with the increased radiation leads to a net O₃ production and fluxes out higher O₃ to have Φ > 1. However, the detailed simulation shows that the T_net transition from T_net > 0 to T_net < 0 happens at ∼ 12:00 (Figure S2c). Therefore, during the period of 10:15 to 12:00, the transport contribution might stay as import but keep weakening by net chemical production (Φ is in the range of 0.9986–1.0482) and then converse to export effect. Hence, in the simulation, Φ > 1 indicates not only the export but also including the transitional process of import to export under the pseudo-steady state. T_net transition process is supported with the animation plot in the supplementary material (Figures S4 and S5).

In contrast to the model result, Φ derived from the observation is mostly greater than 1.3 during the daytime (i.e., significant net export of O₃). This phenomenon is not consistent with the observed lower O₃ concentration in the downwind before the transition under the assumption of linear interpolation from the seaside to inland (Figure S4). Such inconsistency might be caused by the measurement error, the accuracy of the applied photolysis rate constant (j_NO2), and the well-mixing assumption. In the measurement error aspect, NO has a relatively low concentration except for the morning peak as stated in section 3.1. Therefore, the round-off error of NO concentration might lead to an unignorable deviation of Φ. Moreover, the measured NO₂ might be over-reported due to the interference of other NOy species (e.g., HNO₃, organic nitrogen species) on the instrument (molybdenum converter, Ecotech 9841) to cause a higher Φ. It is then estimated a maximum of ∼30% contribution from other NOy species on the over-reported NO₂.

For the photolysis rate constant issue, the same j_NO2 in the simulation was adopted for the Leighton ratio calculation for observation due to the lack of solar radiation measurement. Because the Leighton Ratio from the simulation is 0.9–1.13, not too far away from the photo-stationary state, the significantly higher Φ from observation data by a factor of 1.35 than the model calculation might suggest an overestimation of j_NO2 corresponding to the observation concentration likely due to the influence of aerosols and cloud cover, which was not considered in this study. In particular, aerosols might significantly affect the photolysis rates over ZuoYing, since high PM condition is common in winter. If the system is retained at the photo-stationary state, aerosols and clouds might reduce the radiation responding on j_NO2 by a factor of ∼0.70 ± 0.08. The aerosol optical depth (AOD) in Southern Taiwan derived from MODIS satellite data was >0.6 [44], while the single scattering albedo (SSA) was estimated in the range of 0.8–0.9 [45]. If aerosols have a major influence on j_NO2 reduction, (AOD, SSA) for the winter condition can be estimated as (0.75 ± 0.15, 0.85 ± 0.05) using Tropospheric Ultraviolet and Visible (TUV) radiation model [46], consistent with the reported values [44,45].

Moreover, Φ from observation reveals an absurd peak (∼1.7) in the early morning, which may be caused by various factors. For instance, as mentioned above, Kaohsiung has suffered from severe air quality issues, especially in the early morning. In winter, the abundant aerosols might constantly lower the morning j_NO2 to cause applied j_NO2 highly overestimated. Also, the applied j_NO2 is mathematically estimated with the intensity of solar radiation, which reaches its highpoint at 12:00, as stated in section 2.2.1. In reality, the peak may slightly differ due to the difference between local time and solar time. With a 15-min delay adjustment in j_NO2 (i.e., Φ(t) = j_NO2(t+15min) [NO₂](t)/(k_NO-O3(t)[NO](t)[O₃](t)), the re-calculated Φ for observation has a similar trend as simulation without the absurd morning peak, as shown in Figure 4b. Nonetheless, the overall profile still exceeds 1.3, consistent with the aerosol influence stated above.

The proximity of the observation site to the emission sources might have a stronger influence than the well-mixing process in the model since the fresh emission would cause a lower Φ [1]. The emitted NOx has a ratio NO/NO₂ = 9 while the daytime simulated NO/NO₂ ranges from 0.005 to 0.61, roughly in the observed range from 0.081 to 0.38. This may imply that the NO transforms rapidly to NO₂ once emitted if there is sufficient O₃. If the observation site captures the information close to the source, a lower Φ is expected, and the corresponding import is possibly due to vertical motion or other processes instead of horizontal transport. Nonetheless, a regular Taiwan EPA site, ∼ 15 m above the ground level, may be able to capture the concentration close to photo-stationary state due to the rapid transition of NO, as the observed ratio NO/NO₂ is similar to the simulated ratio, obtained under the ideal well-mixing assumption. Hence, the higher derived Φ from observation data is more likely resulted from the over-reported NO₂, and the applied higher j_NO2, without considering the scattering or absorption of aerosols and clouds.

3.3. Ozone isopleths for emission control

Emission usually dominates the concentration of precursors and NOx in urban areas. Figure 5a shows that NMHCs emission primarily affects O₃ around 10:00–14:00, at which the photolysis is closed to its peak of the day, and has negligible influence on O₃ for the rest of time, in particular, the morning time ozone O₃(8:00). Therefore, both O₃(max) and ΔO₃ respond positively to NMHCs emission change. On the other hand, Figure 5b suggests that the overall ozone profile responds negatively to NOx emission change in the whole day, since the titration effect is independent of the existence or the intensity of light. Therefore, it indicates that ZuoYing region is under NOx-saturated regime and O₃(max) is negatively related to NOx emission change. Nonetheless, under a half NOx emission change as an example (green line in Figure 5b), the noontime O₃ increases slightly less than ozone in the rest of time: though NOx emission is reduced by half in the whole day, O₃ partially converts back to NOx in the daytime. Thus, the difference of O₃ changes at different times leads to a decrease of ΔO₃ so that ΔO₃ responds positively to NOx emission change. Similar results were reported by Jhun et al., who suggested that NOx reduction narrowed the gap between O₃(Max) and non-peak O₃ [27].

Figure 5.

The simulated diurnal O₃ profile (solid lines) and Leighton Ratio (dotted lines) under different emission conditions; (a) normalized NMHCs emission, and (b) normalized NOx emission. The gray dashed line specifies Φ = 1.

The dependence of O₃(Max), as a function of NMHCs and NOx emissions, was applied to investigate the different reduction strategies [29,43,47]. Figure 6a presents the ozone isopleth plot of a typical NOx-saturated regime, but with linearized ozone isopleth. The behavior of ozone isopleths indicates that the environment of ZuoYing is still NOx-saturated even with the imported precursors only, consistent with the study of Peng et al. [48]. To shift the environment to NOx limited, we would need to adjust the boundary condition, which was not taken into accounts for simplification. Figure 6b shows ΔO₃ isopleths under different NMHCs and NOx emission scenarios. As discussed in the preceding paragraph, the O₃(Max) and ΔO₃ isopleths as a function of NMHCs and NOx emissions show different trends. By visualizing the emission control strategy as an arrow emanating from the base case (1, 1) on Figures 6a and 6b, the simultaneous decrease of O₃ and ΔO₃ can be achieved if the arrow lies in the shaded area; that is, the following criterion is satisfied: -0.25 ≤ Δ_NNOx/Δ_NNMHCs ≤ 0.5 and Δ_NNMHCs < 0, where Δ_N is the change of normalized emission. However, it may not be wise to lower O₃ and ΔO₃ via more NOx emission, so the criterion should be refined as follows:

0 ≤ Δ_NNOx/Δ_NNMHCs ≤ 0.5 and Δ_NNMHCs < 0 (Eq. 5)

Figure 6.

The simulated isopleth of max and the increment of O₃ under different emission conditions (1 is the control emission condition). (a) O₃(Max) and (b) ΔO₃. The shaded region represents the emission conditions corresponding to both O₃(Max) and ΔO₃ decreases.

Once the emission reduction satisfies Eq. (5), the simultaneous decreases of NMHCs, NOx, O₃, and ΔO₃ can be achieved.

Moreover, to further examine the influence of NOx and NMHCs emission during transport, Figures 5a and 5b show the diurnal O₃ profile and Leighton Ratio under different emission rates, respectively. The transported or chemically produced O₃ is always partially titrated by NOx molecules. To retain a pseudo-steady state, higher O₃ results in a stronger net export effect, leading to a higher Leighton Ratio, and causes higher ozone concentration downstream due to O₃ production overcoming the titration via the emission adjustment. In terms of species, the Leighton Ratio is more sensitive to NOx emission change than to NMHCs change, corresponding to the fact that O₃(Max) is more sensitive to NOx emission change. Besides, Leighton Ratio is more sensitive and achieves a maximum value around 14:00, possibly due to high j_NO2, the O₃ peak concentration, and the strongest sea breeze.

3.4. Influence of convergence and sea breeze

With the addition of convergence, Figure 7 shows a maximum of 10.0% O₃ increase during 10:45 to 19:30. This enhancement is attributed to the net effect of ozone accumulation and the chemical processes via the accumulated precursors. By taking 14:00 data for comparison, the O₃ increases 9.6% by convergence. As shown in Figure S6, the deviation of $\frac{\partial O_3}{\partial t}$ due to convergence mainly follows from the deviation of net transport contribution, as transport primarily accounts for an 8.4% O₃ increase while chemistry contributes to the remaining 1.2% increase. The chemical response to convergence event is the combined effect of accumulated NOx and NMHCs: the former lowers O₃ in a NOx-saturated regime while the latter rises O₃ in the noontime. As a result, the opposite chemical effects on ozone due to NOx and NMHCs accumulation (6.3% and 6.5%, respectively, at 14:00) are mostly canceled out in this NOx saturated environment. Therefore, the convergence increases local ozone concentration over ZuoYing mainly physically with minor influence by the chemical processes.

Figure 7.

The simulated O₃ profile (solid lines) and Leighton Ratio (dotted lines) with and without a hypothetical convergence, stated in section 2.3. The gray dashed line specifies Φ = 1. The Leighton ratio profiles for the two cases are indistinguishable.

The Leighton Ratio in Figure 7 does not show an observable change under convergence. The 6.3% increase of NOx at 14:00 can be decoupled as the combination of a 1.2% decrease for NO and an 8.0% increase for NO₂. The non-observable response of Leighton Ratio due to convergence might reflect the cancellation of opposite chemical effects of NMHCs and NOx. Besides, with the convergence, the accumulation leads to increased import contribution during the transition period 10:15–12:00 as shown in Figure S6. The chemical processes are slightly modified to reflect the accumulation of precursors and ozone, which is relatively small (less than 10%) to affect the Leighton Ratio.

The influence of convergence over the coastal ozone is strongly dependent on the environment. Liu and Chan reported that NOx dominated a severe air-pollution episode in Hong Kong while O₃ retained low concentration during the convergence event caused by sea-land breezes [20]. In this Hong Kong case with NOx-saturated regime, because the upstream location was a power plant with significant NOx emission, the titration effect by excessive NOx accumulation was more significant than the possible opposite effects by NMHCs and O₃ accumulation in the convergence zone, resulting in a much lower O₃ concentration than the background values. More specifically, NOx dramatically increased during the convergence episode (e.g., NO ∼ 480 ppbv & NO₂ ∼ 100 ppbv during the episode while the averaged concentration was NO ∼ 24 ppbv & NO₂ ∼ 39 ppbv over Central/Western, a pollutant monitoring station in Hong Kong). In contrast, the convergence induced by the sea breeze and clean upstream source moderately increase O₃ by a maximum of 10.0% in our study. Hence, the convergence chemically decreases ozone in the study of Liu and Chan [20]. Moreover, Wang et al. showed that the sea breeze circulation might lead to severe ozone episodes at the coastal site in Hong Kong [18]. In their domain of study, the marine site revealed a higher ozone concentration than the coastal site due to a stronger photochemical process and weaker titration effect. The cyclone or continental anticyclone might carry the air parcel containing human-emitted primary pollutants outsides the land, and later the photochemistry resulted in a high ozone concentration. Under this setup, the ozone would be transported into the coastal sites by the developed sea breeze; in contrast, the sea breeze, in general, brings cleaner air over ZuoYing in the afternoon, combining with stronger photo-production of ozone, which results in a net export effect.

4. Conclusions and atmospheric implications

In this study, we evaluate the effect of transport on the ozone-related chemistry and provide a feasible emission control strategy over a NOx-saturated regime using a two-dimensional numerical model coupling both physical mechanisms and core chemical processes for the coastal wintertime O₃ concentration simulation. The simulation results suggest that the daytime O₃ chemistry is close to the photo-stationary state Φ ∼ 1, with a deviation of -9% to +13% mainly due to the transport process. Φ experiences a transition from import effect before 10:15 to weakening import or net export effect afterward associated with a net O₃ production as sea breeze starts developing with ventilation effect. During the period of 12:00 to 14:00, the transport process postpones the O₃ peak by keeping Ox in the form of NO₂ under the pseudo-steady state condition. In addition, the significantly higher Φ derived from observation data than from the simulation by a factor of 1.35 suggests the over-reported NO₂ and the possible aerosol and cloud influence, which reduces ∼30% on j_NO2, associated with AOD of 0.75 ± 0.15 and SSA of 0.85 ± 0.15, under the apparent photo-stationary condition (i.e., Φ ∼ 1).

Under the hypothetical sea breeze convergence, the influence of sea breeze convergence leads to a maximum of 10.0% O₃ increase primarily by the accumulation via the transport process in this study. The chemical response to convergence reflects the cancellation of accumulated NMHCs and NOx, leading to no significant changes of Leighton ratio with convergence. With the analysis of ozone isopleths under differential local emission control strategies, the annual increase of O₃ may be attributed to the improper ratio of normalized NMHCs and NOx reductions due to their opposite effects on O₃. The feasible emission control strategies as Eq. (5) can achieve simultaneous decreases of O₃(Max), ΔO₃, NMHCs, and NOx. Furthermore, the same concept could be applied to evaluate the influence of feasible emission control strategies on the mitigation of other species such as Ox, and HNO₃, which are also controlled by NOx and NMHCs and might have significant contributions to particulate matter [43, 49, 50]. The superposition of profiles for different species might propose a better emission control strategy to reduce all species.

Declarations

Author contribution statement

Justin Lien: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Hui-Ming Hung: Conceived and designed the experiments; Analyzed and interpreted the data; Wrote the paper.

Funding statement

This work was supported by the Ministry of Science and Technology, Taiwan (108-2111-M-002-003 and 109-2111-M-002-003).

Data availability statement

Data will be made available on request.

Declaration of interests statement

The authors declare no conflict of interest.

Additional information

No additional information is available for this paper.

Appendix A. Supplementary data

The following is the supplementary data related to this article:

Supplementary_Material_Lien_spl_Hung_R1