Research on the Detection of Middle Atmosphere Temperature by Pure Rotating Raman–Rayleigh Scattering LiDAR at Daytime and Nighttime

Abstract

The temperature of the middle atmosphere is of great significance in the coupled study of the upper and lower layers. A pure rotational Raman–Rayleigh scattering LiDAR system was developed for profiling the middle atmospheric temperature at daytime and nighttime continuously by employing an ultra-narrow band interferometer. The comparisons between LiDAR detections and radiosonde data show that the LiDAR system has temperature detection capabilities of 80 km and 60 km at night and during the day, respectively. The results demonstrate that our method can reliably detect the atmospheric temperature in the middle atmosphere. The significant non-uniformity in the horizontal distribution of temperature in the middle atmosphere and the vertical gradient of atmospheric temperature could be observed by using the developed LiDAR.

1. Introduction

The physics and chemical processes in the middle atmosphere (10–100 km) are extremely complex and have always been a hot and difficult topic for scientists to explore [1]. Particularly, the changes in atmospheric environmental factors (temperature, wind field, etc.) and dynamic disturbances in the space of this layer will directly affect human exploration activities [2]. Therefore, high-precision and high-resolution temperature detection in the middle atmosphere is particularly important.

At present, the acquisition of atmospheric temperature distribution mainly includes atmospheric temperature models, in situ detection, and remote sensing detection [3,4,5]. Among these, the atmospheric temperature model has been continuously improved and perfected with the improvement of data acquisition methods and processing technologies, such as NRLMSISE (United States Naval Research Laboratory Mass Spectrometer and Inconsistent Scatter Radar Exosphere), WACCM (the Whole Atmosphere Community Climate Model), etc. [3,6,7]. However, for the varied atmospheric environment, atmospheric temperature models cannot provide high-precision real-time atmospheric temperature distribution. In situ detection can be carried out by installing detection instruments on radiosonde balloons or radiosonde rockets to achieve the detection of atmospheric temperature profiles. Due to their low cost, radiosonde balloons have become one of the main detection methods among meteorological departments and researchers. However, due to the limitations of temporal and spatial resolutions, they cannot meet the requirements of high spatiotemporal detection accuracy (the detection altitude is usually less than 50 km, and radiosonde balloons are released once in the morning and once in the evening). Radiosonde rockets are greatly limited due to their high cost and inability to continuously observe. Remote sensing achieves temperature detection by detecting certain types of radiation in the atmosphere, such as sound waves, microwaves, and light waves. Among them, satellites can achieve relatively accurate detection of large-scale space atmospheric parameters, such as the MLS (Microwave Limb Sounder) carried by the Aura satellite and the SABER (Radiosonde of the Atmosphere using Broadband Emission Radiometry) detector equipped by the TIMED (Thermosphere Ionosphere Mesosphere Energetics and Dynamics) satellite [5,8,9]. However, satellite remote sensing cannot obtain long-term continuous observation data of a certain place.

Ground-based LiDAR can obtain the temperature profile in the middle atmosphere continuously with high spatial and temporal resolutions [10,11]. Due to the aerosols at the altitudes below about 30 km, pure rotational Raman technology could be used for profiling temperature in this range. Martucci et al. [12] used rotational Raman LiDAR to achieve temperature detection within the range of 1 km–30 km. Rayleigh scattering detection technology can obtain atmospheric temperature profiles within a range of 30 km to 100 km. Li et al. [10] developed a ground-based rotational Raman–Rayleigh LiDAR system to detect atmospheric temperature and density at 15 km–80 km. For temperature observation in the range above 75 km, some scientific researchers developed Na atomic resonance fluorescence or Fe atomic resonance fluorescence techniques. The atmospheric temperature detection LiDAR system of the Leibniz Institute of Atmospheric Physics (IAP) in Germany has achieved detection of atmospheric temperature profiles ranging from 1 km to 105 km [13]. However, the complexity of the resonant fluorescence system has limited its development [13,14].

Because of the radiation from the sun during daytime, the distribution and variation of temperature during daytime is also very important in the study of physical and chemical processes in the middle atmosphere. However, there are few reports on the detection of temperature distribution in the middle atmosphere during daytime because of the strong skylight.

This article presents a LiDAR technology that combines pure rotational Raman scattering and Rayleigh scattering for temperature detections in the middle atmosphere during daytime and nighttime. Ultra-narrowband interferometer technology with a bandwidth of 10 pm was developed to solve the detection during daytime. The composition of the system is introduced in the second part. The methodology of data processing is introduced in Section 3. Some detections and comparisons are shown in Section 4. Some conclusions are given in the last part.

2. Instrument

In this study, a vehicle-based LiDAR system, combining the technology of pure rotational Raman scattering and Rayleigh scattering, was developed. The structure and principle of the system are shown in Figure 1. The pure rotational Raman–Rayleigh scattering LiDAR is located inside the cabin of the vehicle. As it is a vehicle-mounted system, it facilitates our subsequent detection of temperature in the middle atmosphere over different regions. The system mainly consists of five parts: a laser emission subsystem, optical receiving subsystem, data acquisition and control unit, data analysis and processing unit, and vehicle platform subsystem.

A Nd: YAG solid-state laser was selected for the system. The single longitudinal mode 1064 nm wavelength laser emitted by a narrow linewidth laser was used as seeded light. After frequency tripling, a 355 nm output laser pulse was obtained. In order to effectively receive weak laser scattering echo signals, a Cassegrain telescope with a diameter of 1000 mm was used. The received signal by the telescope proceeds into the subsequent optical pathway after undergoing collimation and filtering. Initially, the reflected component undergoes color separation via a narrow-band interference filter (wavelength: 354.7 nm, bandwidth: 0.3 nm) with an incident angle of 6.5°, allowing the passage of Mie scattering and Rayleigh scattering signals at a wavelength of 354.7 nm. The optical signal reflected by this filter then undergoes spectral separation through another interference filter (wavelength: 354.05 nm, bandwidth: 0.3 nm) with an angle of 6.5°, transmitting the low-order rotational Raman scattering signal, which subsequently enters a photomultiplier tube for detection via a converging lens. The optical signal reflected once more undergoes color separation through an interference filter (wavelength: 353.2 nm, bandwidth: 0.5 nm) with an angle of 5°, transmitting the high-order rotational Raman scattering signal, which also enters a photomultiplier tube for detection via a converging lens. Meanwhile, the transmitted Mie/Rayleigh scattering signal is directed into the corresponding detection channel via a mirror. During nighttime detection, the signal is further refined by passing through an interference filter (wavelength: 354.7 nm, bandwidth: 0.15 nm) to minimize background noise before entering a photomultiplier tube for detection via a converging lens. Conversely, during daytime detection, the mirror is removed from the optical path, and the signal light undergoes additional refinement by passing through an interference filter (wavelength: 354.7 nm, bandwidth: 0.15 nm) and an F-P (Fabry–Perot) interferometer (wavelength: 354.7 nm, bandwidth: 10 pm) to minimize background noise before entering a photomultiplier tube for detection via a converging lens. The specific technical parameters of the system are shown in

Table 1. Main technical parameters of the pure rotational Raman–Rayleigh scattering LiDAR system.

Technical Index	Performance Parameter
Wavelength (nm)	355
Single pulse energy (mJ)	600
Laser linewidth (MHz)	200
Laser divergence angle (mrad)	0.5
Beam Expansion Factor	6.75
Telescope aperture (mm)	1000
Viewing angle of telescope (mrad)	0.3
Filter 1 central wavelength (nm)	354.7
Filter 1 bandwidth (nm)	0.3
Low J filter central wavelength (nm)	354.05
Low J filter bandwidth (nm)	0.3
High J filter central wavelength (nm)	353.2
High J filter bandwidth (nm)	0.5
Filter 2 central wavelength (nm)	354.7
Filter 2 bandwidth (nm)	0.15
F-P central wavelength (nm)	354.7
F-P bandwidth (pm)	10
Sampling precision (MHz)	150

.

In order to meet the detection requirements at various altitudes, an air-gap F-P interferometer is designed. As shown in Figure 2, the space where the F-P interferometer is located achieves airtightness. It has a pressure tuning function to achieve tuning of its center wavelength.

Table 2. Main technical parameters of air gap F-P interferometer.

Indicator Name	Performance Parameter
Effective aperture (mm)	50.8
Free Spectral Range (pm)	90
3 dB bandwidth (pm)	10
Peak transmittance (parallel beam)	>70%
Mirror optical reflectance	78%

presents the main technical parameters of the air-gap F-P interferometer, with a bandwidth of 10 pm, peak transmittance greater than 70%, and a free spectral range of 90 pm. The developed interferometer can well meet the application requirements of this LiDAR system, effectively suppressing the background light during the daytime sky and improving the detection capability during the day.

3. Methodology

The data acquisition system used in this study is a Licel transient recorder, which achieves simultaneous acquisition of analog counting and photon counting data [15]. The near end photon counting data often exhibits supersaturation effects. Although it can be alleviated after dead time correction, there is still a significant deviation when the near end photon counting rate approaches the photon counting detection limit. The analog counting mode was employed for detection mode of larger signals, and it has a poor signal-to-noise ratio (SNR) in the far end. In order to improve the overall detection performance of LiDAR, especially to enhance the SNR of LiDAR echoes, it is usually necessary to splice the analog counting and photon counting data of LiDAR echo signals.

Figure 3 shows the integrating results of Rayleigh channel signals detected over the Hefei area (E 117.16°, N 31.90°) at 18:37 on 18 December 2022. Among them, red, green, and blue represent the analog channel signal, dead-time-corrected photon counting signal, and integrated signal, respectively. The integrated signal combines the advantages of analog counting and photon counting, improving the SNR of the entire signal profile.

The expression for the intensity of pure rotational Raman scattering spectral lines of atmospheric molecules can be written as follows [16,17,18]:

$$I\left(J,T\right)=Pc\tau N{\left({\displaystyle \frac{d\delta }{d\mathsf{\Omega }}}\right)}_{rot,J,T}^{\pi }$$

(1)

where J and T are rotational quantum numbers and atmospheric temperature; P, c, τ, and N corresponds to the laser emission power, the speed of light in vacuum, the laser pulse width, and the number density of atmospheric molecules, respectively; and ${\left({\displaystyle \frac{d\delta }{d\mathsf{\Omega }}}\right)}_{rot,J,T}^{\pi }$ is the pure rotational Raman backscattering differential cross-section at atmospheric temperature T.

Starting from the LiDAR equation, the received pure rotational Raman atmospheric echo signal can be obtained as follows [16,17]:

$$S_{RR}\left(z\right)=S_0\epsilon {\displaystyle \frac{A\mathrm{G}\left(z\right)}{z^2}}\mathsf{\Delta }zn\left(z\right)T_{atm}^2\left(z\right)\left[{\displaystyle \sum }_{i=O_2,N_2}{\displaystyle \sum _{J_i}{\tau }_{RR}\left(J_i\right)}{\eta }_i{\left({\displaystyle \frac{d\delta }{d\mathsf{\Omega }}}\right)}_{RR,i}^{\pi }\left(J_i\right)\right]$$

(2)

In the formula, z, S₀, ε, A, and Δz represent the detection distance, number of emitted photons, photodetector efficiency, effective receiving area of the telescope, and detection distance resolution, respectively; G(z), n(z), and T²_atm(z) are the geometric overlap factor, atmospheric molecular number density, and atmospheric two-way transmittance of the LiDAR system at height z, respectively; η_i is the volume fraction of N₂ and O₂ molecules in the atmosphere; and τ_RR(J_i) is the optical transmittance of the spectral line corresponding to the wavelength of the rotating quantum number J_i. The last term of Formula (2) is the differential backscattering cross-section of the J-order pure rotational Raman spectral line.

For a single spectral line detection system, the ratio of signal intensity Q between two channels and temperature T satisfies the following relationship:

$$T={\displaystyle \frac{b}{\ln Q-a}}$$

(3)

where a and b are related to the corresponding LiDAR system. Usually, the two parameters, a and b, can be obtained by performing nonlinear least squares fitting on the ratio of temperature data obtained from radiosonde balloons to pure rotational Raman signals obtained from LiDAR. Once a and b are determined, the inversion of the atmospheric temperature profile can be achieved.

A previous study has suggested that the optimal fitting function for a pure rotational Raman scattering temperature detection system is as follows [19]:

$$T={\displaystyle \frac{-2a}{b\pm \sqrt{b^2-4a\left(c-\ln Q\right)}}}$$

(4)

The Rayleigh scattering LiDAR equation can be expressed as follows [20]:

$$N_R(\lambda ,z)=\left[{\displaystyle \frac{P_L(\lambda )\mathsf{\Delta }t}{hc/\lambda }}\right]\left({\displaystyle \frac{A}{z^2}}\right)\left[\eta (\lambda )G(z)\right]\left[{\sigma }_R(\pi ,\lambda )n_R(z)\mathsf{\Delta }z\right]T_{atm}^2(\lambda ,z)$$

(5)

The number density of atmospheric molecules at height z can be expressed as follows:

$$n_R(z)={\displaystyle \frac{N_R(\lambda ,z)z^2}{\left[\frac{P_L(\lambda )\mathsf{\Delta }tA\lambda }{hc}\right]\eta (\lambda )G(z)\mathsf{\Delta }z{\sigma }_R(\pi ,\lambda )T_{atm}^2(\lambda ,z)}}$$

(6)

According to formula (5), the number density of atmospheric molecules is directly proportional to the square of the distance from the echo signal. In addition, besides being related to atmospheric parameters such as σ_R(π, λ) and T²_atm(λ, z), it is also related to the system parameters of LiDAR [11,21].

$$T(z)={\displaystyle \frac{n_R\left(z_0\right)}{n_R\left(z\right)}}T(z_0)+{\displaystyle \frac{M}{R}}{\displaystyle \underset{z}{\overset{z_0}{\int }}\frac{g(z^{\prime })n_R\left(z^{\prime }\right)}{n_R\left(z\right)}}d{z}^{\prime }$$

(7)

In the equation, n_R(z₀) is the number density of atmospheric molecules at z₀ altitude, and T(z₀) is the atmospheric temperature at z₀.

4. Results and Analysis

4.1. Comparative Detection and Error Analysis

Figure 4 shows the comparison results between the detection data of the LiDAR (300 m vertical resolution, 20 min temporal resolution) and the radiosonde balloons from Anqing station during 8:00 to 8:20 at daytime on 31 January 2023 (the distance between the radiosonde balloon at Anqing Station and the LiDAR is about 170 km). Figure 4a shows the atmospheric temperature detection results of the entire layer from 20 km to 60 km, and Figure 4b shows the corresponding statistical temperature error. It can be seen that within the comparison range of radiosonde balloons around 20 km to 27 km, the trend of radiosonde data and LiDAR detection results is consistent, especially the calibrated pure rotational Raman temperature inversion results, which overlap completely with the radiosonde data results.

The SNR of pure rotational Raman signals above 33 km is poor and almost ineffective. Within the range of 33 km to 60 km, by comparing the Rayleigh signal and NRLMSISE model, it can be concluded that they have the same trend of change. It is worth mentioning that Rayleigh signals can reflect more detailed changes in temperature structure, which provides the possibility for gravity wave detection. Due to the use of a series of methods such as ultra-narrowband interferometers and data integration, we were able to control the atmospheric temperature error within ±2 K even at a detection altitude of 60 km during the day.

Figure 5 shows the comparison results between the detection data (18:00–18:20) of the LiDAR (20 min temporal resolution) and the radiosonde balloons from Anqing station at nighttime on 30 January 2023 (the distance between the radiosonde balloons at Anqing station and the LiDAR is about 170 km). To improve the accuracy of temperature detection and reduce statistical errors, this can be achieved by reducing vertical resolution and increasing cumulative time. Figure 5a shows the atmospheric temperature detection results of the entire layer from 20 km to 90 km. It can be seen that the atmospheric temperature profile inverted from Rayleigh scattering signals is consistent with the results measured by pure rotational Raman spectroscopy within the range of 30 km to 40 km. Within the range of 50 km to 60 km, there is a significant deviation due to the influence of pure rotational Raman SNR, but the trends of both are consistent. Figure 5b shows the corresponding statistical temperature error. The statistical error of pure rotational Raman channels is less than 1 K below 30 km, 4 K around 40 km, and approximately 10 K at 50 km. The statistical error of the Rayleigh detection channel is less than 3 K below 60 km, about 5 K at 70 km altitude, about 10 K at 80 km altitude, and about 20 K at 90 km altitude.

Figure 6 shows the correlation analysis of the results derived from the Raman signals, Rayleigh signals, and radiosonde during daytime and nighttime shown in Figure 4 and Figure 5. Figure 6a shows the results from the Raman signals and sonde during daytime in the range from 19.8 km to 27 km, and the correlation coefficient is 0.908. Figure 6b shows the results from the Raman signals and sonde during night in the range from 19.8 km to 33 km, and the correlation coefficient is 0.963. Figure 6c shows the results from the Raman signals and Rayleigh signals during nighttime in the range from 27.9 km to 40.2 km, and the correlation coefficient is 0.986. The correlation analysis results show the reliability of the LiDAR detections of temperature in the middle atmosphere during daytime and nighttime.

In order to further verify the accuracy of the Raleigh temperature detection data, the results of the LiDAR system were compared with SABER (Figure 7). Among them, the detection time of SABER’s data is 1:08 am local time, with a longitude and latitude of E 117.18° and N 33.17°; the vertical resolution of LiDAR inversion data is 600 m, and the time resolution is 60 min (00:38–01:38); and the distance between the two is about 140 km. Overall, the two have good consistency throughout the entire comparison range, and the bimodal structure within the range of 40 km to 50 km also fits well. However, at the temperature conversion position between 55 km and 65 km, there is a significant deviation between the two, with a maximum of 8 K. Above 75 km, there is a significant difference in the overall trend, mainly due to the influence of temperature differences at the reference point and the SNR of the LiDAR.

4.2. Continuous Atmospheric Temperature Distribution in the Middle Atmosphere

Starting from 20:00 on 30 January 2023 and ending at 7:00 on 31 January, an 11 h detection experiment was conducted on the temperature of the middle atmosphere in the Hefei area. The temperature spatiotemporal distribution in the middle atmosphere obtained from this detection experiment is shown in Figure 8 (300 m vertical resolution, 20 min temporal resolution). It can be seen that the temperature in the middle atmosphere has a very complex disturbance structure. From Figure 6, it can be seen that there is good consistency in the vertical temperature distribution detected using two different detection principles, pure rotational Raman scattering and Rayleigh scattering. Two temperature peaks were detected near 48 km and 56 km. This indicates that the disturbance structure of the atmospheric temperature distribution measured by the LiDAR in Figure 8 is reliable.

Starting from 19:00 on 9 March 2023, and ending at 15:00 on March 10, an additional 20 h experiment was conducted in the Hefei area, covering both daytime and nighttime periods (Figure 9) (600 m vertical resolution, 60 min temporal resolution). To ensure the accuracy of the data, we kept data below 60 km during the day. Data below 80 km was retained at night. There is a clear layered structure between 42 km and 60 km at night, with two temperature peaks appearing near the heights of 42 km and 54 km. As the height of the higher temperature peak gradually decreases, around 2:00 am, these two peak structures gradually merge into one peak. It can be seen that the atmospheric temperature structure measured day and night has good continuity, which confirms the correctness of the daytime temperature detection results. The temperature detection results at an altitude of 40–50 km during the day are significantly higher than those at night, and the peak temperature gradually increases from 7:00 with the continuous increase of sunrise time.

From the two continuous detection cases shown above, it can be seen that the spatiotemporal distribution of temperature in the middle atmosphere is quite complex. On the one hand, the temperature peak is not unique, and the height of the peak also varies over time. On the other hand, the temperature of the middle atmosphere undergoes significant changes over time and space. Therefore, using equipment such as LiDAR to conduct long-term and continuous high-precision detection of the temperature structure of the middle atmosphere is beneficial for studying the spatiotemporal distribution characteristics of the middle atmosphere temperature.

Combining the results of Figure 4 and Figure 5, we can conclude that the error between pure rotational Raman data and radiosonde data below 30 km is less than 1 K, and the statistical error of Rayleigh data below 60 km is less than 3 K. By using data processing methods such as multi-vertical resolution, temperature errors below 80 km can be controlled within ±5 K. Therefore, for the data of the entire system, we believe that the detection results at 30 km–80 km have high accuracy and are effective. Therefore, our continuous detection results retained data within the range of 30 km–80 km.

In addition, the background light signal during the day is very large, which can lead to a poor SNR. From our plan, it can be seen that the signal channels during the day and night are separate (Figure 1). During the day, we used an ultra-narrowband F-P interferometer to improve the SNR of the Rayleigh signal. The above methods, such as accumulating multiple vertical resolutions and reducing temporal resolution, ensure that our daytime detection data remains valid even at 60 km. Obtaining the vertical distribution of atmospheric temperature below 60 km during the day can help us understand the diurnal variations of temperature in the middle atmosphere and even the changes in gravity waves. We will continue to conduct research on this issue in the future to further improve our daytime detection capabilities.

5. Conclusions

A vehicle-based pure rotational Raman–Rayleigh scattering LiDAR system has been developed, which uses a 1 m aperture optical telescope to receive atmospheric backscatter echo signals and suppresses strong sky background light during the day using an ultra-narrowband interferometer with a bandwidth of 10 pm. The daytime temperature detection height of the Rayleigh channel can reach 60 km, and the temperature statistical error below 60 km is less than ±2 K. By accumulating signals with multiple vertical resolutions, the nighttime temperature detection height of the pure rotational Raman channel can reach 50 km, and the temperature statistical error below 40 km is less than 3 K. The temperature detection height of the Rayleigh channel at night can reach over 80 km, and the statistical error of inversion is less than 3 K below 60 km. The statistical error within 60 km to 80 km is less than 7 K, and approximately 10 K between 80 km and 90 km. Through continuous all-day observations, it has been verified that the LiDAR system has temperature detection capabilities of 80 km and 60 km at night and during the day, respectively.

The developed LiDAR system was used to continuously detect the temperature in the middle atmosphere during day and night, obtaining a refined spatiotemporal distribution of the temperature in the middle atmosphere. The disturbance structure characteristics of the temperature of the middle atmosphere were clearly and intuitively obtained from the LiDAR observation data. There is significant non-uniformity in the horizontal distribution of temperature in the middle atmosphere, and the vertical gradient of atmospheric temperature has significant layer structure characteristics and obvious inversion structures. This provides technical reference for subsequent research on the temperature of the middle atmosphere.