Characteristics of Planetary Boundary Layer Height (PBLH) over Shenzhen, China: Retrieval Methods and Air Pollution Conditions

Highlights

• What are the main findings?

• The gradient method and standard deviation method based on Micro-Pulse Lidar and the parcel method based on Microwave Radiometers are more sensitive to abrupt changes in boundary layer height.
• During the observation period, Shenzhen’s PBLH characteristics exhibited significant diurnal variation and high sensitivity to pollution, the daytime mean PBLH ranged from approximately 512 to 1345 m, while nighttime values generally decreased to around 500 to 650 m. And under high aerosol loading conditions, PBLH was significantly suppressed to approximately 500 m, indicating that pollution limits atmospheric mixing by inhibiting boundary layer development.

• What are the implications of the main findings?

• By comparing different retrieval methods, we can identify the strengths and weaknesses of each method, providing guidance on selecting appropriate algorithm for similar urban environments.
• The study of boundary layer characteristics in Shenzhen holds significant scientific value for future research on factors influencing boundary layer height in megacities.

Abstract

The PBLH affects the intensity of the surface turbulence and the state of pollutant dispersion. Current research on PBLH characteristics and their relationship with pollution in coastal megacities remains insufficient. Moreover, existing studies rarely evaluate the consistency of various boundary layer solution methods, making it difficult to identify deviations in single methods. So, we conducted enhanced observation experiments in Shenzhen, a megacity in China, between March and July 2023. The characteristics of the PBLH was analyzed by five months of observations from Micro-Pulse Lidar (MPL) and Microwave Radiometer (MWR). Five retrieval methods (Parcel, GRA, STD, WCT, and Theta) were applied for comparative assessment. The results shows that all methods captured similar diurnal patterns. During daytime, the PBLH ranged from 512 to 1345 m, with Theta yielding the highest and STD the lowest average values. At night, PBLH decreased overall, and method-dependent differences persisted. Under different pollution levels, this study also discussion the properties of PBLH using MPL and microwave radiometer. And aerosol optical depth (AOD) and PBLH showed a strong negative correlation, indicating aerosol-induced suppression of boundary layer growth. The study of boundary layer characteristics in coastal megacities can provide reference for atmospheric physics research in economically developed coastal areas.

1. Introduction

The planetary boundary layer (PBL), the lower part of the troposphere, regulates the exchange of heat, momentum, water, and pollutants between the surface and the free troposphere [1]. The boundary layer height is commonly used to describe the vertical extent of mixing within the atmospheric boundary layer and the level at which, exchange with the free troposphere occurs, making it an important parameter in boundary layer studies [2,3]. The dispersion of anthropogenic pollutants is constrained not only by horizontal transport but is more strongly modulated by vertical diffusion capacity. As a key parameter determining vertical diffusion capacity, the planetary boundary layer height (PBLH), along with its variations and associated dynamic processes (such as turbulence and wind), plays a critical regulatory role in the accumulation and dispersion of pollutants [4,5,6]. Existing studies have shown that turbulence is suppressed when the boundary layer is stably stratified or under temperature inversion conditions [7], and a reduced PBLH restricts pollutants (especially aerosols) within the near-surface layer, leading to aggravated air pollution. Conversely, elevated aerosol concentrations reduce incoming solar radiation and enhance atmospheric stability, further inhibiting boundary layer development and establishing a positive feedback that aggravates pollution [8,9,10]. Therefore, determining the atmospheric boundary layer height is of great importance for forecasting pollution events and understanding their formation mechanisms.

Currently, various remote sensing techniques can be employed to estimate the PBLH, including satellite observations [11,12], ground-based lidar [6,13], wind profiler radar [14,15], ceilometers [16,17], and microwave radiometers [18,19]. Among these, lidars and microwave radiometers are particularly suited for continuous monitoring due to their high temporal resolution and capability for automatic operation. There are various methods for retrieving the PBLH based on aerosol lidar observations, among which the most commonly used include the standard-deviation method [20], gradient method [21,22], wavelet covariance transform method [23,24], and curve fitting method [25]. For microwave radiometers, PBLH is often inferred using the parcel method [26] or by detecting discontinuities in the temperature gradient [27]. Although lidar-based methods are widely used for PBLH retrieval, significant discrepancies can arise when comparing lidar-derived PBLH with thermodynamic retrievals (e.g., from radiosondes), especially in the presence of residual layers or complex aerosol structures [28,29]. This reflects the method-dependent nature of these retrievals, where each algorithm exhibits distinct strengths and limitations. For example, the gradient method captures the diurnal evolution of PBLH reliably and performs best under clean conditions, though it may slightly underestimate PBLH in the afternoon and is vulnerable to interference from lofted layers during transition periods. The standard-deviation tends to be more responsive on strongly convective days but is more likely to misidentify the PBL top under polluted or rapidly changing conditions, and thus benefits from being constrained with gradient information [30]. The wavelet covariance transform method is more robust in the early morning, late afternoon, and under cloudy or residual-layer conditions, but often underestimates PBLH at midday and shows increased sensitivity to local emissions [31]. For thermodynamic approaches based on temperature profiles (e.g., parcel or potential temperature gradient methods), the PBLH is typically inferred from the lifting condensation level or inflection points in the temperature lapse rate, which makes them suitable for convective boundary layers but less sensitive to aerosol-defined mixing heights or weakly forced stable conditions [32].

In the Pearl River Delta (PRD), a region characterized by complex meteorology and heavy anthropogenic emissions, numerous studies have examined PBLH and its role in air quality [33,34]. As a representative PRD city, Shenzhen frequently experiences heavy convective rainfall, heat episodes, and high-ozone days, all closely linked to boundary-layer structure. However, current research on PBLH characteristics and their relationship with pollution in coastal megacities remains insufficient. Existing studies have primarily focused on inland [35] or industrial regions [36], while observational evidence remains relatively scarce in coastal megacities characterized by strong land-sea coupling [37], complex thermal structures [38], and high humidity [39]. Moreover, while various PBLH retrieval algorithms are widely used, few studies systematically evaluate the consistency and performance differences among these methods under different pollution conditions [30,40]. This lack of cross-method validation introduces uncertainty into boundary-layer interpretation and makes it difficult to identify potential biases or limitations of individual approaches. As a result, the inferred relationship between PBLH and pollutant concentrations may vary significantly depending on both the region and the retrieval method, further complicating efforts to generalize findings across different urban environments [41]. To address this gap, we use MPL and MWR observations and apply several established algorithms to retrieve PBLH over Shenzhen, systematically comparing their performance, analyzing the diurnal cycle under different pollution conditions, and exploring relationships with aerosol optical depth (AOD) and PM_2.5. In addition, we conduct case studies of typical pollution episodes to provide mechanistic insight into the coupling between boundary-layer dynamics and urban air quality. The structure of this paper is as follows: Section 2 introduces the observational setup and inversion methods, including the gradient method(GRA), standard deviation method(STD), wavelet covariance transformation method (WCT), air parcel method (Parcel), and positional temperature method (Theta). In Section 3, we compare the PBLH calculated by different methods, show the characteristics of diurnal PBLH variations, and further discuss the relationship between PBLH and PM_2.5 and AOD through five months of statistical analysis and case studies. Finally, the conclusion and perspective are given in Section 4.

2. Materials and Methods

2.1. Study Aera

Shenzhen lies on the eastern shore of the Pearl River Estuary, backed by hills and facing the sea (Figure 1). Under the combined influence of sea–land breezes, the East Asian monsoon, and intensive urbanization, convective storms and air-quality events occur relatively frequently [42]. As a high-density coastal megacity with nearly 18 million permanent residents, Shenzhen stands as one of China’s most significant economic and technological hubs. Built along a narrow coastal plain, Shenzhen is surrounded by reservoirs, low hills, and the waterways of the Pearl River Delta, bordering Hong Kong. The combination of high population density, intense industrial and transportation emissions, and complex coastal meteorological conditions makes Shenzhen an ideal location for studying boundary layer structure, pollutant accumulation, and urban atmospheric processes.

The data used in this study were obtained from an experiment conducted at the Shenzhen Meteorological Tower (SZMT, 22.66°N, 113.91°E) located at the Meteorological Station in Shenzhen, from 1 March to 31 July 2023. The experiment was specifically designed to intensify observations during the warm season, with a focus on capturing changes in the Shenzhen boundary layer under warm season conditions. The site is situated within a reservoir surrounded by forest. A busy highway is located approximately 800 m away, with Shenzhen Airport to the west. The northern area is covered by trees and several low-rise residential buildings. Therefore, the primary sources of aerosols in this area are anthropogenic emissions [43].

2.2. Instrumentations

2.2.1. MPL

The MPL operates at a pulse repetition frequency of 2.5 kHz and a wavelength of 532 nm. It provides a vertical resolution of 15 m and a temporal resolution of 10 s for backscatter profiles. The instrument is capable of detecting cloud signals and retrieving a range of atmospheric products, including normalized relative backscatter signals, backscatter coefficient profiles, extinction coefficient profiles, aerosol optical depth, among others. Due to the incomplete formation of laser pulses near the surface, a blind zone of approximately 150 m exists in the lower atmosphere. Therefore, in this study, the lidar value at 150 m is taken as the near-surface value to minimize the impact of this blind zone on the data. As part of the standard data processing protocol, the raw MPL data are subjected to background subtraction, saturation correction, overlap correction, after-pulse correction, and range correction to derive the normalized signal [44].

2.2.2. MWR

The RPG-HATPRO passive microwave radiometer for temperature and humidity profiling utilizes multi-channel parallel measurements, employing a receiver system with seven temperature channels and seven humidity channels to detect brightness temperatures. The instrument is also equipped with sensors to measure surface temperature and relative humidity (RH). This MWR provides a variety of atmospheric products, such as brightness temperature, humidity profiles, and liquid water path (LWP). It can detect up to a maximum altitude of approximately 10 km and generates temperature and humidity profiles from the surface to 10 km in about 2 min. Additionally, it can retrieve Level 2 data such as integrated water vapor (IWV, the total column water vapor content per unit area) and liquid water path (LWP, the vertically integrated liquid water content per unit area) with a temporal resolution of 1–2 s.

For consistency, the data used in this study were uniformly averaged to a 5 min resolution. To ensure the quality of the data, observations taken during rainfall were excluded.

2.3. Methods

2.3.1. Retrieving Method PBLH from MPL

The raw data received by the MPL consist of photon counts returned after the laser beam is scattered by atmospheric constituents. These photon counts are converted into signals by a photon counter, then transmitted to a computer and stored as a function of time. However, the raw data collected by the system always contain instrument noise and sky background noise. Therefore, corrections such as after-pulse correction, overlap factor correction, delay factor correction, and background noise subtraction are required to obtain the normalized relative backscatter signal (NRB). According to the algorithm proposed by He (2006) [45], the raw MPL signal can be expressed as Equation:

$$p(z)={\displaystyle \frac{\frac{Qc(z)CE\beta (Z)T^2}{z^2}+n_b(z)+n_{ap}(z)}{DTC[P(Z)]}}$$

(1)

where $p\left(z\right)$ represents the number of detected photoelectrons at range $z$; $E$ is the emitted pulse energy of the lidar; $C$ is a system constant; $T^2=exp(-2{\int }_0^z\sigma \left(r\right)dr)$ denotes the atmospheric transmittance; $\sigma \left(z\right)$ is the extinction coefficient of the atmosphere at range $z$; $\beta \left(z\right)$ is the atmospheric backscatter coefficient; $n_b\left(z\right)$ is the background noise; $n_{ap}\left(z\right)$ is the after pulse for detector run on; $Q_c\left(z\right)$ is the overlap correction factor of the lidar system and DTC refers to the detector timing delay correction. After applying the above series of corrections, the normalized relative backscatter signal can be obtained, and the corrected equation is given as Formula (2):

$$NRB(z)=\{p(z)\times DTC\left[P(z)-n_b(z)-n_{ap}(z)\right]\}\times z^2$$

(2)

Based on the above introduction, radar data can be used to determine the PBLH using the GRA, STD, and WCT methods. The following provides a detailed description of these three methods:

The GRA method defines the PBLH as the height where the gradient of the NRB reaches its minimum value [46]. The $D\left(z\right)$ can be expressed as follows:

$$D(z)={\displaystyle \frac{dNRB}{dz}}$$

(3)

In the STA method, the PBLH is the height where the lidar signal changes the most. This is found by checking where the NRB has the highest standard deviation (σ) [47]. The function can be expressed as follows:

$$\sigma ={[{\displaystyle \frac{1}{N}}\sum _{i=1,N}{(NR{B}_i-NRB)}^2]}^{{\displaystyle \frac{1}{2}}}$$

(4)

The WCT method is to calculate the wavelet coefficients at each scale factor in each NRB profile, observe the step change in the return signal, and the height corresponding to the maximum value of the wavelet coefficient in each set of NRB data is the required PBLH [48]. This method is based on the harr function, which is defined as follows:

$$h({\displaystyle \frac{r-b}{a}})=\left\{\begin{array}{c}+1,b-{\displaystyle \frac{a}{2}}\le r\le b\\ -1,b\le r\le b+{\displaystyle \frac{a}{2}}\\ 0,\mathrm{elsewhere}\end{array}\right.$$

(5)

In this context, $r$ is the altitude, $a$ refers to the spatial extent of the Haar wavelet, and $b$ is its center position, acting as the translation parameter in the wavelet transform. These variables are used to define the function for calculating the wavelet covariance transform confficient:

$$W_{NRB}(a,b)={\displaystyle \frac{1}{a}}{\int }_{r_b}^{r_t}NRB(r)h\left({\displaystyle \frac{r-b}{a}}\right)dr$$

(6)

In the equation, $W_{NRB}(a,b)$ represents the similarity between the NRB signal and the Haar function within the range of $b\pm {\displaystyle \frac{a}{2}}$. A higher value of $W_{NRB}(a,b)$ indicates a greater resemblance between the NRB signal and the Haar function, meaning a more pronounced step change. Therefore, the altitude corresponding to the maximum $W_{NRB}(a,b)$ is considered to be the PBLH.

This study established a height threshold (3 km) during the inversion process to prevent cloud layers from being misidentified as the boundary layer top. This threshold was determined through visual inspection of backscatter profiles and aligns with the typical development characteristics of convective boundary layers in the region. It effectively filtered out the influence of high-altitude clouds (exceeding the threshold height) appearing after 15:00 on the PBLH inversion results.

2.3.2. Retrieving Method PBLH from MWR

In this study, based on previous research using MWR observational data, two methods are adopted to determine the PBLH: the parcel method and the potential temperature gradient method. The Parcel method identifies the PBLH as the highest altitude a surface air parcel can reach when lifted dry adiabatically under the ambient temperature profile. This is determined by the intersection point between the temperature profile and the dry adiabatic lapse line, starting from the surface temperature. On the other hand, the Potential temperature gradient method defines the PBLH as the height at which the vertical gradient of virtual potential temperature ${\displaystyle \frac{\partial {\theta }_v}{{\partial }_z}}$ first exceeds a critical threshold, typically capped by an inversion layer in the convective or mixed layer, where temperature increases more rapidly with height [49]. In this study, based on previous research, the threshold is set to 6 K km⁻¹.

$${\theta }_v={T_v\left({\displaystyle \frac{p_0}{p}}\right)}^k$$

(7)

In Formula (7), $T_v$ represents the virtual potential temperature; $p_0$ is the standard atmospheric pressure, taken as 1013.25 hPa; $p$ is the atmospheric pressure. The constant $k$ is defined as $R/C_p$ $\approx$0.286, where R is the gas constant for dry air, and $C_p$ is the specific heat capacity of air at constant pressure.

2.3.3. Comparison of PBLH Calculation Methods

In the previous two sections, we introduced five different methods to calculate PBLH, each with its own unique principles and applications. To provide a clearer understanding of their relative performance and applicability, the

Table 1. Advantages and Disadvantages of PBLH Calculation Methods.

Method	Advantages	Disadvantages
GRA	Applicable to lidar data, especially suitable for use under clear-sky conditions	In the case of rapid gradient changes, it may lead to abrupt or discontinuous results
STD	Wide applicability	It may yield inconsistent results under complex boundary layer or unstable conditions
WCT	Effectively identifies PBLH based on temperature gradients, suitable for both stable and unstable atmospheric conditions.	It may produce errors when identifying inversion layers.
Parcel	Directly reflects the potential temperature structure of the boundary layer	Sensitive to surface temperature errors, and less accurate under strong convection or non-adiabatic conditions
Theta	Effectively reflects changes in potential temperature gradients, especially suitable for cases with clear temperature profiles	It may overestimate under strong vertical temperature gradients.

summarizes the advantages and disadvantages of each method, offering a detailed comparison of their strengths and limitations.

2.4. Other Datasets

The Air Quality Index (AQI) data used in this study were obtained from the China Environmental Monitoring Center (CEMC), which continuously updates and releases hourly data. At present, the national urban ambient air quality monitoring network covers 338 prefecture-level and above cities, providing hourly concentrations of six major pollutants, including PM_2.5, PM_2.5, SO₂, O₃, NO₂, and CO, as well as the corresponding AQI values.

The AOD product from the MODIS satellite sensor (MCD19A2-v6) was used to obtain daily AOD data. MCD19A2 is a MAIAC (Multi-angle Implementation of Atmospheric Correction) Level-2 gridded land AOD product with a high spatial resolution of 1 km, integrating images from both the MODIS Terra and Aqua satellites. This product provides daily AOD measurements at two wavelengths, 470 nm (blue band) and 550 nm (green band). For this study, AOD at 550 nm (green band) was selected, and the grid data were averaged within a 5 km radius around the Lidar observation site. MCD19A2 data are freely available for download from the MODIS website (https://modis-land.gsfc.nasa.gov/MAIAC.html (accessed on 8 December 2024)).

3. Results

3.1. Comparison of PBLH Retrieved by Different Instruments and Methods

We applied five methods to estimate the PBLH. Figure 2 shows the diurnal variation in PBLH on 17 March 2023, as derived from MPL and MWR using these methods. The results indicate that noticeable differences exist among the PBLH values obtained from different methods. Between 08:00 and 12:00, during the development stage of the boundary layer, the PBLH values estimated by the Parcel, GRA, and STD methods are significantly lower than those derived from the WCT and the Theta methods. At 15:00, the PBLH begins to decline. However, when a distinct “cap-shaped” red band appears in the MPL backscatter signal at an altitude of 1.5–2.0 km (indicating the presence of a high aerosol layer) the PBLH values given by the Parcel, GRA, and STD methods increase rapidly, while those from the WCT and Theta methods remain at low levels. This discrepancy may be attributed to the fact that the GRA and STD methods identify the strongest gradients within the profile, whereas the WCT and Theta methods rely on the averaged signal characteristics over a certain height range to detect step changes.

As a result, WCT and Theta methods are less responsive to abrupt variations, while GRA, STD, and Parcel methods are more sensitive to such changes. Overall, all five methods can reasonably capture the diurnal variation in PBLH. However, when the boundary layer undergoes sudden changes, the WCT and Theta methods tend to be less capable of capturing these variations, while the Parcel, GRA, and STD methods show higher sensitivity to signal changes. Although MPL and MWR are based on different detection principles, within the group of more sensitive methods and the group of less sensitive methods, the results derived from each group are internally consistent with relatively small differences among the methods. In addition, the five retrieval methods yield statistically significant differences in PBLH. In addition, the five retrieval methods yield statistically significant differences in PBLH. Based on the WCT method as a reference, we further evaluated the correlation coefficients, mean absolute errors (MAE), and biases of the other four methods. The correlation coefficient between WCT and QKF was 0.283, with an MAE of 435.7 m and a bias of 6.61 m. The WWF method showed a correlation coefficient of 0.073, an MAE of 629.48 m, and a bias of 80.47 m. Both the GRA and STD methods exhibited stronger consistency with the WCT method, with correlation coefficients exceeding 0.4. These discrepancies arise because the instruments rely on different sensing principles and have different vertical resolutions, leading to variations in their sensitivity to boundary-layer structures and thermodynamic conditions. The algorithms also define PBLH differently in terms of physical meaning and parameter choices. As a result, the PBLH estimates from different methods are not mutually consistent.

To further compare the diurnal variations in boundary layer height obtained by the five methods, we plotted box plots of the results from each method, as shown in Figure 3. As shown in the figure, the Parcel and Theta methods more clearly exhibit the diurnal variation in PBLH. During daytime, solar radiation strengthens after sunrise, air temperature rises, and the mixed layer develops rapidly; around noon the boundary layer peaks, with mean heights exceeding 1 km, and the Theta method yields a maximum of 1565 m. As solar radiation weakens later in the day, surface cooling leads to a gradual decline in mixed-layer height. The Parcel method shows a relatively small diurnal contrast: nighttime PBLH generally remains above 500 m, the daytime mean is around 1000 m, and a midday maximum exceeds 1500 m. The three Micro-Pulse Lidar (MPL)–based methods exhibit a less pronounced diurnal cycle than the other two methods. This indicates that the GRA and STD methods are more sensitive to sharp gradient transitions, consistent with our earlier conclusions. The WCT method also shows a relatively smooth diurnal range (about 500–1000 m), but features a marked increase in PBLH near sunrise.

In Shenzhen, influenced by a subtropical monsoon climate, summer temperatures are high and surface heating is strong, promoting active daytime convection that typically favors boundary layer development. However, due to Shenzhen’s coastal location, it is frequently affected by sea–land breeze circulation, high humidity, and urban heat island effects [50]. These combined factors can lead to the formation of elevated aerosol residual layers or inversion suppression in the afternoon, posing challenges for accurately determining PBLH [37,38]. Of the five methods shown in the figure, the Parcel and Theta methods which based on temperature profiles retrieved from the MWR are better reflect the thermodynamic structure of the boundary layer. The Parcel method is particularly responsive to thermodynamic forcing, accurately capturing the growth and decay of the daytime boundary layer, and being especially sensitive to abrupt changes in the afternoon, often showing a pronounced increase in PBLH. The Theta method, by contrast, produces relatively stable results but tends to overestimate PBLH in the afternoon. The PBLH values derived from three lidar-based methods are significantly lower than those obtained from MWR [51,52]. The MPL method is particularly sensitive in detecting the aerosol boundary layer or the material boundary layer height, as demonstrated by, who showed that MPL identifies the aerosol layer top based on aerosol backscatter signals. In contrast, temperature or virtual potential temperature profiles retrieved from microwave radiometers characterize the thermodynamic boundary layer height, which represents the maximum altitude attainable by aerosols through convective mixing [53]. As illustrated in Figure 2, from 07:00 to 11:00, thermodynamic PBLH gradually increases due to convection driven by solar radiation. Specifically, during the early morning and evening, under the influence of sea–land breezes, the surface rapidly loses heat, leading to cooling and the formation of a temperature inversion layer around 200–300 m above the ground. The thermodynamic PBLH associated with this inversion layer is significantly lower than the PBLH derived from the microwave radiometer.

Subsequently, we compared the PBLH between daytime (08:00–18:00) and nighttime (19:00–07:00 of the following day) using five different methods, as shown in Figure 4. During the daytime period, the PBLH values ranged from 512 m to 1345 m. Among them, the Theta method produced the highest values, with an average of 1342 m. This may be attributed to the fact that the Theta method estimates PBLH based on temperature gradients, and during the observation period, temperature inversions were relatively rare in the lower atmosphere, resulting in higher PBLH estimates. Moreover, the Theta method showed a higher number of low-value outliers, indicating potential estimation errors. The Parcel method resulted in a PBLH of 897 m, while the GRA, STD, and WCT methods produced relatively similar results, with values of 755 m, 588 m, and 697 m, respectively. During the nighttime period, the GRA and STD methods estimated slightly higher PBLH values than during the daytime, reaching 926 m and 747 m, respectively. In contrast, the remaining three methods (Theta, WCT, and Parcel) all showed lower PBLH values at night compared to daytime, with average values ranging between 500 m and 650 m. Additionally, all five methods exhibited more outliers at night than during the day, suggesting a tendency for overestimation of PBLH during the nighttime period.

To further understand the characteristics of the PBLH in Shenzhen and its relationship with pollution, we compared observational data from Shenzhen with other coastal megacities, such as the Jungnang Station in Seoul and the Tianjin Science Park Observation Station. The diurnal variation in Shenzhen’s PBLH exhibits higher daytime values and lower nighttime values, consistent with the typical diurnal patterns observed in these cities. This variation is primarily driven by solar radiation and coastal thermal circulation and correlates with pollution levels. However, our study reveals that Shenzhen’s PBLH range (400–1000 m) is significantly narrower than Seoul’s (with a daytime maximum PBLH of approximately 1580 m). Its average PBLH (645 m) is lower than Seoul’s and closer to Tianjin’s level, with daytime peaks about 33% lower than Seoul’s. This PBLH suppression phenomenon is not evident in similar studies of Seoul or the PRD region [54,55,56]. We attribute Shenzhen’s PBLH characteristics to its subtropical monsoon climate, frequent summer sea breezes, topographical constraints from adjacent mountain ranges, and intense urbanization. These factors collectively suppress vertical PBL development, resulting in structural differences in PBLH between Shenzhen and other coastal megacities or cities in the PRD region.

3.2. PBLH Variation Properties Under Different Particulate Pollution Condition

To investigate the variations in PBLH under different particulate pollution conditions, we classify air quality based on the “Technical Regulation on Ambient Air Quality Index (Trial)” (HJ633-2012 [57]) issued in China. According to this standard, the Air Quality Index (AQI) is divided into six levels: Excellent (0–50), Good (51–100), Slightly Polluted (101–150), Moderately Polluted (151–200), Heavily Polluted (201–300), and Severely Polluted (>300). During the observation period, no days with moderate or higher pollution levels occurred in Shenzhen. Therefore, this study focuses on the first three categories (Excellent, Good, and Slightly Polluted), and AQI values above 100 are defined as polluted conditions. In addition, due to missing MWR data under both Good and Slightly Polluted conditions, only the PBLH values derived from MPL are discussed in this section.

Figure 5 compares the PBLH calculated using the GRA, STD, and WCT methods under Excellent, Good, and Slightly Polluted conditions. As shown in the figure, all three methods indicate that the PBLH under Excellent conditions is the highest, with average values of 807 m (STD), 1143 m (WCT), and 1062 m (GRA). The median PBLH values estimated by the three methods are also relatively high, with a wide distribution range and some values exceeding 2000 m, suggesting strong atmospheric convection and enhanced vertical mixing that facilitates pollutant dispersion. Under Good conditions, the PBLH levels decrease, and the boxes shift downward, reflecting slightly weaker but still effective dispersion conditions. In Slightly Polluted conditions, PBLH decreases further, with average values significantly lower than those in the previous two categories: 486 m (STD), 460 m (WCT), and 608 m (GRA). This indicates a more stable atmospheric stratification, limited boundary layer development, and greater accumulation of pollutants near the surface. Additionally, extremely low PBLH values (below 1.5 times the interquartile range) are mostly concentrated under polluted conditions, suggesting that boundary layer suppression is more pronounced during unfavorable dispersion conditions. At the same time, PBLH variability appears to increase as pollution levels decrease—under Excellent conditions, the wider range of the boxes and whiskers reflects a more sensitive response of the boundary layer to different meteorological processes.

To further investigate the dynamic relationship between PBLH and PM_2.5 concentration, Figure 6 presents their temporal variations and correlation analysis. As shown in Figure 6a, under “excellent” conditions, PBLH generally remains at relatively high levels (often exceeding 1500 m), corresponding to low PM_2.5 concentrations (10–20 µg/m³); under “good” conditions, PBLH stays between 500–1500 m, corresponding to moderate PM_2.5 levels (20–30 µg/m³); whereas under “slightly polluted” conditions, PBLH mostly remains below 1000 m, while PM_2.5 concentrations significantly increase, reaching up to 57 µg/m³. The observed peak in PM_2.5 concentration at 09:00 under slightly polluted conditions is primarily due to the combination of a low PBLH resulting from early-morning stable stratification, which traps pollutants accumulated overnight, and elevated anthropogenic emissions (e.g., traffic and industrial activity) in the morning. Currently, the PBLH is still insufficient to support vertical mixing, leading to the maximum PM_2.5 concentration. After 12:00, enhanced solar radiation drives a sharp rise in PBLH (exceeding 1000 m), significantly boosting vertical mixing and dispersing the accumulated PM_2.5. Since the slightly polluted group has a higher initial concentration, the diffusion effect leads to a more pronounced reduction in PM_2.5, resulting in afternoon concentrations lower than those in the good and excellent conditions, which already have lower initial concentrations. Correlation analysis further indicates a significant negative relationship between the two variables, with PBLH decreasing by an average of approximately 28 m for every 1 µg/m³ increase in PM_2.5 concentration. This result is not only consistent with the static distribution characteristics of different pollution levels revealed in Figure 5 but also further confirms, from the perspective of dynamic interactions, the regulatory role of boundary layer height in pollutant dispersion and accumulation processes.

3.3. Correlation Between PBLH and AOD

Building on the analysis of PBLH variations under particulate pollution conditions, we further investigated the relationship between PBLH and AOD. In polluted atmospheres, the distribution and concentration of aerosols significantly influence boundary layer stability and development. As a key indicator of aerosol concentration, AOD is therefore crucial for analyzing PBLH variations during pollution events. In this study, AOD values were derived from MPL measurements based on extinction coefficients.

To further evaluate the reliability of the MPL_AOD, we compared it with AOD values retrieved from the MODIS satellite. Due to cloud contamination, MODIS AOD was missing on several days. Ultimately, a total of 49 days of MODIS data were successfully matched with MPL observations. As shown in Figure 7a, a significant linear relationship was found between MPL_AOD and MODIS_AOD (R = 0.72), indicating good agreement between the two. Therefore, the AOD retrieved from MPL is considered reasonably reliable. Following the classification of different particulate pollution levels described earlier, we applied the same categorization to the MPL_AOD, as shown in Figure 7b. It is evident that AOD values under slightly polluted conditions are significantly higher than those under good and excellent air quality conditions, with an average value of 0.984. In contrast, the average AOD values under good and excellent conditions are 0.731 and 0.583, respectively. In addition, we compared the monthly mean AOD values over the five-month observation period, as shown in Figure 7c. The results indicate that AOD levels were relatively high in March and April, but gradually decreased as temperatures rose. This trend is closely linked to the seasonal transition between spring and summer in the Shenzhen region, as well as the gradual development of the South China Sea monsoon and the South Asian monsoon. Specifically, as the monsoon establishes itself in mid-to-late May, significant changes occur in atmospheric circulation conditions: the persistently strengthening southerly winds not only enhance horizontal advection but also synergize with local sea-land breeze circulation. These airflows effectively transport clean, moist marine air masses (relative humidity > 85%) from the South China Sea to the coastal regions of South China. This process not only dilutes and replaces anthropogenic aerosols accumulated during Shenzhen’s spring dry season but also promotes hygroscopic growth of aerosols, altering their scattering efficiency and indirectly influencing AOD inversion sensitivity [58]. Simultaneously, monsoon-driven atmospheric instability intensifies, triggering more frequent convective activity and precipitation. Enhanced wet scavenging effects directly remove substantial particulate matter (especially PM_2.5 and PM₁₀) from the atmosphere [59]. Furthermore, convective updrafts associated with monsoon precipitation aid in the vertical transport of aerosols to higher altitudes, reducing their concentration within the planetary boundary layer and further driving down surface AOD values. Notably, the monsoon’s modulation of AOD exhibits nonlinear characteristics: During the early monsoon phase (early May), when monsoon intensity is weak and intermittent, AOD may exhibit short-term fluctuations due to competition between aerosol growth driven by initial moisture and limited removal effects. As monsoon intensity strengthens and stabilizes, the combined effects of dilution, wet scavenging, and vertical transport become dominant, ultimately leading to a sustained and significant downward trend in AOD [60].

Following the analysis of monthly AOD variations, we further examined the diurnal relationship between AOD and PBLH. As shown in Figure 8, a negative correlation is observed between AOD and PBLH, indicating that higher aerosol concentrations are generally associated with a shallower boundary layer. This inverse relationship is primarily attributed to the radiative effects of aerosols, which suppress boundary layer development by reducing surface solar radiation (i.e., the dimming effect) and enhancing atmospheric stability. The hourly mean analysis further reveals that under high AOD conditions, especially during daytime when solar heating dominates, the growth of the boundary layer is significantly inhibited. This result is consistent with findings from previous studies.

3.4. Case Study

We further explore the relationship between the PBLH and PM_2.5 concentration during a case period from 28 May to 2 June 2023, which serves as an example. Influenced by the combined effects of the subtropical high and the subsiding airflow on the periphery of Typhoon “Mawar”, Shenzhen experienced persistent high-temperature weather, with maximum daily temperatures exceeding 36 °C. Figure 9 presents the hourly averaged temperature, relative humidity, wind direction, wind speed, PM_2.5 and PM_2.5 concentrations, as well as O₃ concentration during this period. A significant rise in PM_2.5 and PM_2.5 concentrations on May 30 coincided with a pronounced reduction in PBLH, suggesting a potential link between boundary layer suppression and pollutant accumulation. From May 28 to May 30, the average PBLH declined significantly, from 920 m to 320 m. This suppression of PBLH development was accompanied by a general reduction in wind speed, which mostly remained between 2–3 m/s, with some periods dropping below 1.5 m/s. Additionally, wind direction varied frequently, lacking a consistent prevailing pattern. The combination of a shallow boundary layer and weak winds limited the dispersion of pollutants and water vapor, thereby promoting local accumulation of pollutants and hindering the inflow of clean air. Meanwhile, temperatures remained high throughout the event, frequently exceeding 32 °C. The intense solar radiation and high temperatures provided favorable conditions for photochemical reactions, leading to a marked increase in ozone levels during the daytime. At night, increased relative humidity contributed to the hygroscopic growth of fine particles, which in turn sustained elevated PM_2.5 and PM_2.5 concentrations into the following day. The pollution episode persisted until June 1, when the development of the boundary layer facilitated the dispersion of particulate matter. On this day, PBLH peaked at 1079 m, and PM_2.5 concentrations dropped below 20 µg/m³, accompanied by higher wind speeds and lower relative humidity. By June 2, both PM_2.5 and PM_2.5 levels had decreased significantly, marking the end of the pollution event.

4. Conclusions

In this study, we analyzed the diurnal variations in PBLH in Shenzhen from March to July using five months of observations from the MPL and the MWR. Five retrieval methods were applied, including the GRA, STD, WCT, Parcel, and Theta methods. We compared the performance and applicability of these methods in identifying the boundary layer, and further explored the relationships between PBLH, AOD, and PM_2.5 concentrations.

The results showed that although all five methods exhibited consistent overall diurnal trends, they differed notably in their sensitivity and estimated values. The Parcel, GRA, and STD methods responded more rapidly and prominently to boundary layer transitions or the presence of elevated aerosol layers, while the WCT and Theta methods, due to their smoother processing or reliance on averaged features, showed weaker sensitivity to abrupt changes.

During the daytime period (08:00–18:00), the PBLH values estimated by the five methods ranged from 512 to 1345 m, with the Theta method yielding the highest average (1342 m) and the STD method the lowest (588 m). At night (19:00–07:00), PBLH was generally lower, with Parcel, Theta, and WCT producing similar average values (approximately 500–650 m), while GRA and STD yielded relatively higher results, suggesting differences in their ability to detect stable layers.

Further comparisons under varying pollution levels revealed a clear decrease in PBLH with increasing pollution severity. Under “excellent” air quality conditions (AQI 0–50), the average PBLH estimated by the three MPL methods reached 807, 1143, and 1062 m, respectively; under light pollution conditions (AQI > 100), these values dropped to 486, 460, and 608 m. AOD and PBLH also showed a negative correlation, indicating that high aerosol loading reduced incoming solar radiation and surface heating, leading to a suppression of boundary layer development.

The case study from 28 May to 2 June 2023, further illustrated this relationship. Under the combined influence of the subtropical high and subsidence flow from Typhoon “Mawar”, PBLH in Shenzhen dropped to 320 m on 30 May. Meanwhile, PM_2.5 and PM_2.5 concentrations increased to 78 μg/m³ and 46 μg/m³, respectively, ozone concentrations exceeded 200 μg/m³, and wind speeds fell below 2 m/s, resulting in poor dispersion conditions. As PBLH rose above 1000 m on 1 June, air quality improved and the pollution gradually dissipated.

In summary, this study, based on multi-source observational data during the typical warm season in Shenzhen, revealed the diurnal variation characteristics of PBLH and its response to aerosol and particulate matter concentrations. The results demonstrate that PBLH plays a critical regulatory role in the evolution of atmospheric conditions over subtropical coastal cities. The combined application of multiple retrieval methods provides strong support for assessing atmospheric mixing conditions and pollution formation processes in such regions. In future research, we will conduct enhanced observations in Shenzhen, and simultaneously utilize multi-site observational data to carry out more investigations on the PBLH in this region, aiming to comprehensively clarify the spatiotemporal variation patterns of PBLH and its complex interaction mechanisms with aerosols, pollutants, and synoptic systems under different seasonal backgrounds.