Study on Atmospheric Boundary Layer Retrieval Method and Observation Data Analysis Based on Aerosol Lidar

Abstract

The atmospheric boundary layer is the lowest part of the troposphere, directly influenced by the Earth’s surface. The boundary layer’s height is a critical parameter for weather forecasting, air quality monitoring, and climate modeling. Lidar has become a premier tool for continuous boundary layer height detection with its high spatial–temporal resolution. A multi-wavelength aerosol lidar with 355 nm, 532 nm, and 1064 nm has been developed and deployed for operational observations at the Haidian District Meteorological Service of Beijing. The structure design, specifications, observation campaign, and detection principle of the multi-wavelength aerosol lidar are presented and the retrieval method of the boundary layer’s height is introduced. By comparing it with the data of the digital radiosonde, it is verified that the first normalized gradient of the range-corrected signal can more accurately retrieve the boundary layer’s height. The typical daily variation characteristics and influencing factors of urban boundary layer height are analyzed through observational examples and the monthly mean value of the boundary layer’s height in 2019 is acquired and analyzed.

1. Introduction

The atmospheric boundary layer (ABL) is the lowest layer of the atmosphere directly affected by the ground, serving as a bridge for material and energy exchange in the Earth’s atmosphere system [1,2]. The relationship between the ABL and human activities is the closest and most direct, and air pollution problems mainly occur in the ABL [3,4]. Because the ground is the main source of aerosols, the concentration of aerosols in the ABL is significantly higher than that in the free atmosphere above the ABL. The vertical structure of the ABL is closely related to the spatial distribution of aerosols [5,6].

The urban area of Beijing has typical urban underlying surface characteristics [7]. Studying the structure of the ABL in the urban area of Beijing will help to understand the structural characteristics of the urban ABL, playing an important role in evaluating the distribution and diffusion of atmospheric pollutants, and will have important significance for urban meteorological research and meteorological support [8,9].

The atmospheric boundary layer height (BLH) is one of the key parameters in studying the ABL. Multiple methods and techniques have been developed to observe the BLH. In meteorology, for example, the vertical profiles of meteorological elements such as temperature and humidity are used to retrieve the BLH. Data obtained from instruments like weather balloons, meteorological towers, and microwave radiometers can all be employed for this purpose. Typically, a strong temperature inversion exists at the top of the ABL, which restricts the upward diffusion of pollutants and water vapor. Correspondingly, there is a pronounced positive gradient in potential temperature (PT) and a negative gradient in relative humidity (RH) at the top of the ABL. Therefore, temperature and humidity profiles from radiosonde data can be used to retrieve the BLH [10,11]. Atmospheric aerosols mainly exist in the ABL, and their concentration decreases significantly in the free atmosphere above the ABL. As a result, atmospheric aerosols can serve as tracers, and the variations in their vertical distribution can be used to retrieve the BLH [12].

Lidar detection of the ABL relies primarily on atmospheric aerosols as tracers. As a result, under clear-sky conditions with low aerosol concentrations, inaccuracies may occur in estimating the BLH. Furthermore, due to the inherent blind zone in lidar measurements, the low-altitude nighttime stable boundary layer may sometimes remain undetected. Nevertheless, with its high spatiotemporal resolution, lidar offers a novel approach for studying the ABL, enabling the retrieval of the BLH from the vertical profile data of atmospheric aerosols [13,14]. Through continuous observation, lidar can capture the spatiotemporal variations in the BLH with higher temporal and spatial resolution. Research has found that there is a high correlation between the BLH retrieved by lidar and meteorological radiosonde data [15]. By utilizing long-term observation data from lidar, the spatiotemporal distribution characteristics of the ABL can be obtained [16,17].

A multi-wavelength aerosol lidar (MWAL) with 355 nm, 532 nm, and 1064 nm wavelength outputs, designed and developed by the Institute of Oceanographic Instrumentation, Qilu University of Technology (Shandong Academy of Sciences), has been deployed at the Haidian District Meteorological Service of Beijing for observation. To ensure the long-term reliability of the MWAL observations, we carry out regular system calibrations and implement rigorous signal quality assurance procedures [18]. The structure, specifications, observation campaign, and detection principle of the MWAL are presented. The retrieval method of the BLH was introduced and compared with the data of GTS1 digital radiosonde to verify the reliability of the method. Through long-term continuous observations of the MWAL, the daily and monthly variations in the ABL in the urban area of Beijing are analyzed.

2. The MWAL and Observation Campaign

The MWAL with three-wavelength outputs at 355 nm, 532 nm, and 1064 nm, has been developed for automatic observations of aerosols and clouds in the troposphere. The echo signals are received by a Schmidt–Cassegrain telescope with a 12-inch diameter. The optical receiving system has six channels, which, respectively, obtain elastic scattering signals at 355 nm and 1064 nm, the 532 nm parallel polarization (532 nm-P) and perpendicular polarization (532 nm-S) signals, and N₂ Raman scattering signals at 387 nm and 607 nm. The schematic diagram and internal structure of the MWAL are shown in Figure 1.

The MWAL is deployed at the Haidian Park Observation Site (39°59′ N, 116°17′ E) of the Haidian District Meteorological Service of Beijing to conduct all-day observations. The GTS1 digital radiosonde carried by balloon is usually released at 7:15 and 19:15 local time every day at Beijing Observatory (39°48′ N, 116°28′ E) to obtain the meteorological elements profiles, such as temperature, humidity, wind, and pressure, etc. The technical specifications of the MWAL are shown in

Table 1. The specifications of MWAL.

Parameters	Specifications
Transmitter
Laser type	Lamp-pumped Nd:YAG
Laser wavelength	355 nm, 532 nm, 1064 nm
Pulse energy	40 mJ@ 355 nm, 30 mJ@532 nm, and 60 mJ@1064 nm
Pulse duration	10 ns
Pulse repetition frequency	20 Hz
Beam divergence	200 µrad
Receiver
Telescope type	Schmidt–Cassegrain
Telescope diameter	12-inch
Field of view	500 µrad
Receiving channels	355 nm, 387 nm, 532 nm-P, 532 nm-S, 607 nm, and 1064 nm
Detector type	APD for 1064 nm PMT for 355 nm, 387 nm, 532 nm-P, 532 nm-S, and 607 nm
Range resolution	15 m
Detection range	0.1~20 km

.

3. The BLH Determination

3.1. The Retrieval Method of BLH

The aerosol lidar equation is the basis for describing the detection principle of lidar and provides the relationship between the lidar echo signal and the optical properties of the aerosols. The atmospheric backscattered signal power P(z) at distance z can be described by the following lidar equation:

$$P\left(z\right)={\displaystyle \frac{k{P_{0}A}_r\beta \left(z\right)}{z^2}}exp\left[-2{\int }_0^z\alpha \left(z^{\prime }\right)d{z}^{\prime }\right],$$

(1)

where P₀ is the laser pulse power, A_r is the telescope’s effective receiving area, and k is the lidar system constant, incorporating the geometric overlap factor, total optical transmittance, and range resolution. Calculation of the geometric factor from MWAL parameters reveals a blind zone of about 100 m and a unity overlap factor beyond 500 m, thus necessitating a geometric correction in our signal processing [19]. The terms α(z) and β(z) denote the altitude-dependent extinction and backscatter coefficients, respectively, characterizing the optical properties of atmospheric molecules and aerosols. The range-corrected signal (RCS) can be expressed as

$$P_{RCS}\left(z\right)=P\left(z\right)z^2.$$

(2)

The aerosol concentration is generally higher within the ABL than in the free atmosphere above. Consequently, the gradient of the lidar’s signal profile exhibits a distinct negative peak at the top of the ABL. Owing to its high spatiotemporal resolution, the MWAL is highly beneficial for investigating the ABL’s structure.

Several methods, notably the gradient method [20], the wavelet covariance transform method [21], and the curve-fitting method [22], are currently used to retrieve the BLH from lidar data. A correlation coefficient of around 0.85 was reported by Zhang and Caicedo, when they compared the BLH retrieved from lidar data (using the gradient method) against the BLH obtained from radiosonde data [23,24]. This study employs the gradient method, as it is simpler than the conceptually similar wavelet method, while the curve-fitting method is mainly effective for simple aerosol structures. The gradient method utilizes the sharp decrease in aerosol concentration at the top of the ABL, which manifests as a strong negative peak in the gradient of the lidar’s signal profile. This robust principle makes the gradient algorithm widely applicable for BLH retrieval from atmospheric lidar data.

The gradient algorithms predominantly include the first gradient of the raw signal (G-Raw), the first normalized gradient of the raw signal (NG-Raw), the first gradient of the RCS (G-RCS), and the first normalized gradient of the RCS (NG-RCS). The height where the gradient reaches a minimum is identified as the BLH. The calculation formulas for the gradient algorithms are shown in

Table 2. Gradient algorithms for retrieving the BLH.

Algorithm	Calculation Formula
G-Raw	$D_{\mathrm{G}\text{-}\mathrm{Raw}}\left(z\right)={\displaystyle \frac{\Delta P\left(z\right)}{\Delta z}}$
NG-Raw	$D_{\mathrm{NG}\text{-}\mathrm{Raw}}\left(z\right)={\displaystyle \frac{\Delta P\left(z\right)}{\Delta z\times P\left(z\right)}}$
G-RCS	$D_{\mathrm{G}\text{-}\mathrm{RCS}}\left(z\right)={\displaystyle \frac{\Delta P_{\mathrm{RCS}}\left(z\right)}{\Delta z}}$
NG-RCS	$D_{\mathrm{NG}\text{-}\mathrm{RCS}}\left(z\right)={\displaystyle \frac{\Delta P_{\mathrm{RCS}}\left(z\right)}{\Delta z\times P_{\mathrm{RCS}}\left(z\right)}}$

.

In this work, we mainly use a 1064 nm echo signal to retrieve the BLH, due to its stronger penetration for atmospheric particulate matter compared to 532 nm and 355 nm. Figure 2 shows the five-minute accumulated RCS from the MWAL at 04:30 on 22 February 2019. In summary, the 1064 nm laser provides better penetration and a stronger signal response.

In the study of the BLH using lidar, four gradient methods are employed to retrieve the BLH. Meanwhile, the temperature and humidity profiles from radiosonde data are utilized to validate the BLH-retrieval algorithm. Generally, there are very strong positive gradients for potential temperature and negative gradients for relative humidity at the top of the ABL [25]. The results are compared against the PT and RH profiles obtained from GTS1 digital radiosonde. Figure 3 shows the five-minute accumulated 1064 nm RCS profile at 19:15 on 10 February 2020, the corresponding gradient profiles obtained by four gradient methods, and the PT and RH profiles obtained by the GTS1 digital radiosonde.

Evidently, the minimum gradient identified by the G-Raw method leads to a misidentification of the BLH at 0.255 km. In contrast, the minimum gradients detected by the other three methods consistently occur at approximately the same altitude—around 0.845 km. This result is further supported by the RH and PT profiles from the GTS1 digital radiosonde, which also exhibit significant gradient changes at 0.845 km, confirming that this altitude corresponds to the top of the ABL.

Figure 4 compares the gradient profiles from the five-minute accumulated 1064 nm RCS with the PT and RH profiles from the digital radiosonde at 19:15 on 16 June 2020. The minimum values in the gradient profiles obtained by the G-Raw and NG-Raw methods are identified at 0.255 km and 0.24 km, respectively, leading to a misidentification of the BLH. In contrast, the other two methods produce consistent minimum gradient positions, corresponding to an altitude of approximately 1.375 km. The result is corroborated by the PT and RH profiles, which also exhibit distinct gradient changes at around 1.375 km, confirming this altitude as the BLH.

3.2. BLH-Retrieval Process

In practical applications, the retrieval method proves relatively reliable when the aerosol structure in the ABL remains stable. However, several factors can adversely affect the accuracy of BLH determination.

The first scenario is that inhomogeneous of vertical aerosol distribution causes fluctuations in the backscatter signal profile, thereby introducing multiple peaks in the gradient signal. This fluctuation can lead to misinterpretation of the gradient minimum. In comparison, the gradient minimum obtained by the G-RCS method is the most distinct, making it more suitable for determining the BLH, while the NG-RCS method more effectively suppresses signal fluctuations. Therefore, this study primarily employs the G-RCS method for BLH retrieval, with the NG-RCS method applied additionally under conditions of irregular aerosol distribution in the lower atmosphere to improve determination accuracy.

In the second case, when there are clouds above the ABL, the lidar signal gradient at the cloud top will also have a minimum value, which affects the judgment of the BLH. Therefore, during the retrieval process, if a clear separation exists between the ABL and the clouds, the BLH is determined by setting an appropriate altitude range for selecting the gradient minimum. Taking the data of 16 October 2019 as an example, the Time–Height Indicator (THI) figure of the 1064 nm RCS in Figure 5a shows multiple aerosol layers below 3 km and a distinct cloud layer above. If the gradient method is applied directly, the retrieved BLH will be overestimated. By constraining the selection range for the gradient minimum, a more accurate BLH value (indicated by the black dotted line in the THI) can be obtained. When clouds are low and there is no clear separation between the clouds and aerosols in the ABL, the clouds’ top height is generally taken as the BLH—meaning that the cloud is considered to be within the ABL. As shown in Figure 5b of the THI figure of the 1064 nm RCS from 12 October 2019, a persistent, thin cloud layer is present at around 1 km altitude. The aerosol concentration below the clouds is high, and the clouds themselves suppressed aerosol diffusion. In this case, the clouds’ top height can be defined as the BLH [26,27].

The third case involves the direct impact of precipitation on lidar observations, which can lead to a degradation in data quality or result in the data primarily reflecting rainfall information. This makes it difficult to accurately determine the BLH from the lidar measurements. As shown in Figure 5c, using the 1064 nm RCS from 9 July 2019 as an example, the THI indicates that rainfall begins around 10:00 and persists throughout the night. Although the MWAL is able to capture the rainfall process, the precipitation significantly affected near-surface observations. Consequently, it is challenging to retrieve the BLH during this period. Therefore, when performing statistical analysis of the ABL using lidar data, the data during rainfall events will be excluded.

In summary, the main BLH-retrieval process using lidar observational data is illustrated in Figure 6.

The process begins with the raw 1064 nm signal of the MWAL. Firstly, background subtraction processing is performed to remove the interference of background noise on the signal. Generally, the average value of the last 50 points of the signal is selected as the background signal. Subsequently, data smoothing is performed using a five-minute temporal average combined with Locally Weighted Scatterplot Smoothing (LOWESS), which applies weighted linear least-squares regression based on a second-degree polynomial model. This smoothing method effectively reduces signal noise and fluctuation, thereby enhancing the accuracy of subsequent data retrieval without significantly altering the signal’s detailed features. A comparison of the signals before and after smoothing is shown in Figure 7. The processed signal then undergoes range correction. Following this, a check for precipitation is conducted to remove invalid data affected by rain. Cloud base height is then identified. For cloudy conditions, a specific retrieval altitude range is defined. Finally, the BLH is determined by using the gradient method.

4. Results and Discussion

4.1. Analysis of the Diurnal Variation in BLH

The BLH varies both temporally and spatially. Under clear-sky conditions, the urban atmospheric ABL typically exhibits distinct diurnal variations. Specifically, during the daytime, solar radiation heats the ground, triggering turbulent motions that transport heat upward and result in an unstable atmospheric state. This phase is referred to as the convective boundary layer (CBL), with heights ranging from several hundred meters to a few kilometers. At night, surface cooling stabilizes the atmosphere, often leading to strong temperature inversions at lower altitudes. These inversion layers significantly inhibit the diffusion of material and energy within the boundary layer. The nocturnal boundary layer is generally termed the stable boundary layer (SBL), with a height lower than that of its daytime layer [28,29]. However, due to the presence of a residual layer, the BLH retrieved by the lidar gradient method at night is usually higher than that of the SBL.

Figure 8 shows the THI figures of 24 h of continuous observation of the 1064 nm RCS of the MWAL on 27 October 2019 and 16 June 2020, respectively, along with the variation curve of the mean value of the BLH (black dotted lines in the THI figures). The wind direction in these two days is southerly, with an average wind speed of grade 2. The maximum temperature on 27 October 2019 is 17 °C, while the maximum temperature on 16 June 2020 is 34 °C. Both figures have diurnal variation characteristics of the BLH. During the day, the BLH rises with the increase in temperature, with the mean value of the BLH being higher than that at night. With the temperature decreasing at night, the BLH gradually decreases. Considering the weather conditions on the respective days, the rise in BLH and change in aerosol concentration should be primarily attributed to enhanced turbulent exchange caused by surface warming, rather than wind effects. Under similar wind speed and direction conditions, the temperature on 16 June 2020 is significantly higher than that on 27 October 2019. Consequently, the convective effects are more pronounced during the daytime on June 16, 2020, resulting in a higher BLH and greater variability compared to 27 October 2019.

During the observational process, statistical analysis of the data revealed that in many cases, the diurnal variation in the ABL does not necessarily follow the diurnal variation law. The structure and evolution of the ABL are influenced by local cloud distribution and meteorological conditions such as wind, temperature, humidity, and pressure, as well as human activities.

4.2. Analysis of the Monthly Variation in BLH

A statistical analysis is conducted using data collected by the MWAL from January to December 2019.

Table 3. Statistics of MWAL data (2019).

Month	Days with Complete Data	No-Data Days	Valid Hourly Data Count
1	27	3	667
2	20	0	593
3	21	5	587
4	22	2	611
5	18	3	551
6	19	4	541
7	18	1	624
8	14	3	527
9	18	4	540
10	23	2	621
11	17	6	452
12	11	13	306

summarizes the number of days with complete data, days with missing data, and the number of valid hourly observations for each month in 2019. During the observation period, factors such as precipitation and system maintenance affected the acquisition of BLH data. Specifically, August is primarily impacted by rainy weather, resulting in only 14 days of complete data. In December, due to system comparison experiments, testing, and maintenance activities, both the number of days with complete data and the count of valid hourly data points are the lowest for the year.

Figure 9a displays the maximum (Max-BLH), mean (Mean-BLH), and minimum (Min-BLH) values of the BLH at different hourly time points in January. The Max-BLH curve indicates that the maximum values at some nighttime hours are greater than those during daytime hours, primarily due to the influence of low clouds or the elevation of the aerosols caused by nighttime winds. The Min-BLH curve shows that the minimum values between 13:00 and 17:00 are slightly higher than those at other times. Meanwhile, the Mean-BLH curve demonstrates that the BLH begins to rise from 11:00 and reaches its peak at 15:00. The variation trend of the Mean-BLH aligns well with the diurnal characteristics of the ABL. Based on data analysis, the average BLH in January 2019 is approximately 0.75 km.

Figure 9b presents the Max-BLH, Mean-BLH, and Min-BLH values of the BLH at different hourly time points in July. The Max-BLH values at some nighttime hours (e.g., 20:00) are relatively high, mainly due to the effects of aerosol diffusion. The Mean-BLH curve indicates that the BLH generally decreases gradually from 00:00 to 06:00, increases from 06:00 to 15:00, and decreases overall from 15:00 to 23:00. Additionally, the Mean-BLH in the early night is higher than that in the late night. The average variation trend in July is consistent with the diurnal pattern of the ABL. Based on these calculations, the average BLH in July 2019 is determined to be approximately 1.11 km.

Through analysis, it can be observed that the Mean-BLH in July is significantly higher than that in January 2019, which further illustrates that solar radiation energy is stronger in summer than in winter, resulting in more pronounced convective activity, and consequently, a higher BLH.

Using effective observation data of the MWAL in 2019, the monthly Mean-BLH values and variation ranges of the BLH are acquired, as shown in Figure 10. Figure 10 clearly shows the monthly variations in the maximum, minimum, and mean boundary layer heights throughout 2019. The Mean-BLH of August is the highest; the BLH is lower in winter and autumn, and higher in spring and summer. As a result, the ABL is more stable in autumn and winter, while its height exhibits a larger variation range in spring and summer.

5. Conclusions

This study primarily employs the MWAL to investigate ABL-retrieval methods and analyze structural characteristics, offering research approaches and technical solutions for lidar-based studies of ABL structures. First, the methods for retrieving the BLH using lidar are examined, comparing four gradient-based techniques. Through case analyses and comparisons with simultaneous radiosonde data, the first gradient method applied to the RCS is identified as the primary approach for determining the BLH. Several scenarios affecting the retrieval accuracy in practical applications are discussed, and corresponding solutions are proposed. A retrieval workflow is established to enhance the method’s suitability for accurate and automated retrieval of the BLH, thereby obtaining high spatiotemporal resolution information on ABL structures. Subsequently, continuous lidar observation data are used to analyze the ABL structure. The typical diurnal variation characteristics of the BLH are discussed, and case studies are conducted to examine the daily evolution of ABL structures and their potential influencing factors. Furthermore, comparisons of the BLHs across different months in summer and winter are performed, further illustrating the diurnal variation patterns and confirming that summer exhibits a higher BLH than winter. The monthly Mean-BLH values and variation ranges in the BLH for 2019 are generated, revealing that August 2019 has the highest Mean-BLH, while the ABL is more stable and lower in autumn and winter.