2.1.

Pulsed CDL System

The pulsed CDL transmits narrow bandwidth laser pulses into the atmosphere, and then the line of sight (LOS) wind velocity can be acquired by calculating the Doppler shift caused by moving aerosols. This pulsed CDL system is mainly composed of high-energy single-frequency pulsed laser source, optical antenna system, highly sensitive heterodyne detector, signal acquisition, and real-time processing section. The system diagram is shown in Fig. 1.

Fig. 1

System diagram of the pulsed CDL (PMTI, polarization-maintaining tap isolator; AOM, acousto-optical modulators; EDFA, erbium-doped fiber amplifier; EYDFA, erbium–ytterbium codoped fiber amplifier; FPGA, field programmable gate array).

The pulsed laser unit is a master oscillator power amplification system, and the seeder laser is a distributed feedback single-frequency continuous wave laser.¹² The seeder laser beam is divided into two parts by a $1\times 2$ optical fiber splitter. One part (about 0.8 mW) is used as the local oscillator (LO) for heterodyne detection, and the other part is pulse modulated and frequency (intermediate frequency: 160 MHz) shifted by an acousto-optic modulator. The laser pulse was amplified by multistage optical fiber amplifiers and finally up to $300\text{-}\mu \mathrm{J}$ pulse energy output with pulse duration of 400 ns at 10-kHz pulse repetition rate was obtained. The central wavelength was 1540 nm.¹⁷

The optical antenna system is mainly composed of a fiber optical circulator, a coaxial telescope, and a scanner. The fiber circulator is applied to insulate transmitting laser beam and backscattered signal from aerosols. The outgoing laser beam goes into telescope from the output port of the circulator, and then is collimated to 80 mm in diameter. The effective aperture of the telescope is 100 mm. The scanner is used to realize velocity–azimuth display (VAD) scanning by rotating one optical wedge with elevation angle of 20 deg. The LOS wind speeds detected from eight azimuths are used to retrieve the wind vector within 16 s.

The highly sensitive heterodyne detector is used to detect the frequency beating signal caused by mixing the backscattered signal and the LO beam. The frequency beating signal is obtained using a $2\times 2$ (50:50) single-mode polarization-maintaining fiber coupler. The heterodyne signal is sampled by an acquisition card with 1 GSPS sampling rate with 10 bits resolution. The total sampling time is about $40\mu \mathrm{s}$, which is divided into 200 time gates, corresponding to 30-m space–distance per time gate. Fast Fourier transform (FFT) and power spectrum accumulations are implemented in every time gate. The power spectra are finally sent to a computer via RS422 interface, and the LOS wind velocity can be obtained by the traditional peak detection method. The main parameters of the CDL system are listed in Table 1.

Table 1

Main parameters of the pulsed CDL system.

To extract the peak value of the power spectra corresponding to the Doppler shift, gravity method is used here, and the LOS wind velocity can be expressed as⁴

The multiazimuth LOS wind velocities usually have the sine function in VAD scanning mode, which is applied to eliminate the exceptional points and reduce random error. The 3-D wind profiles can be rebuilt by

2.2.

Calibration Method of Wind Velocity between Collocated Lidar and the CDL

The collocated lidar (Molas B300), scaled by German WindGuard testing organization, is a Doppler lidar with distance detection range of 40 to 300 m and wind velocity detection range of 0 to $60\mathrm{m}/\mathrm{s}$. The accuracy of 10-min averaged wind speed and direction is $0.1\mathrm{m}/\mathrm{s}$ and 1 deg, respectively.

Molas B300 can detect from 40 to 260 m in distance with range gate of 20 m, and its zenith angle is 28 deg. Considering the distance detection range of 0.1 to 5 km for our CDL, it is believed that the wind velocity in range of 100 to 260 m is suitable for calibration.

Molas B300 can export both 1-s and 10-min averaged wind speed dataset, and the 1-s dataset can be utilized in time and height matching with the CDL. On the basis of correct time sequence and similar height, the comparison results of 10-min averaged wind profiles can be obtained.

2.3.

Comparative Method for Horizontal Wind Velocity between Sounding Balloon and the CDL

Sounding balloon can directly detect wind vector by tracking the balloon with Beidou/GPS navigation module. The dataset from sounding balloon includes 30-s averaged horizontal wind velocity with 1-s updating period. Its measurement repeatability of the 30-s averaged wind speed and direction of sounding balloon is about $0.2\mathrm{m}/\mathrm{s}$ and 2 deg, respectively.^29,30 The vertical range resolution is about 5 m in both ascent and descent periods.

The comparison data from sounding balloon and the CDL should also be matched at time and altitude. Since the horizontal distance between starting point of sounding balloon and the CDL is $<50\mathrm{m}$, the wind measured by sounding balloon is identified as the wind field detected by the CDL. The comparative scheme of horizontal wind velocity based on the time and altitude characteristics of these two instruments are listed in Table 2.

Table 2

Comparative scheme of horizontal wind velocity.

3.1.

Calibration of the CDL

To calibrate the wind vector measured by the CDL, Molas B300 was set in campaign. The similar outline of 1-s level wind velocity of both Molas B300 and the CDL reflects correct time and height matching of two lidars. To eliminate random error and utilize the secondary datasets (10-min averaged wind velocity of Molas B300), the 10-min averaged wind velocity of two lidars are compared and similar tendency is obtained and plotted in Fig. 3. The time and height matching of two lidars shows good accordance after time–height matching.

Fig. 3

Calibration of the CDL by Molas B300 (sample at 165-m height). (a) Horizontal wind speed calibration of the CDL, (b) horizontal wind direction calibration of the CDL, and (c) vertical wind velocity calibration of the CDL.

Statistics analysis of measured vertical wind velocity between the CDL and Molas B300 is shown in Fig. 4. The correlation coefficients ($R$: 0.964 of horizontal wind speed, 0.992 of horizontal wind direction, and 0.963 of vertical wind velocity) revealed a significant correlation between two lidar datasets. The discrepancy of 10-min averaged horizontal wind speed and direction is $0.164\mathrm{m}/\mathrm{s}$ and 1.719 deg, respectively. The standard deviation of those two kinds of 10-min averaged vertical velocity is decreased to $0.076\mathrm{m}/\mathrm{s}$. The bias is $-0.069\mathrm{m}/\mathrm{s}$, which reveals a systematic bias due to the intermediate frequency shift. The sample of horizontal wind speed profile also displays high similarity as shown in Fig. 4(d). These results provide the calibration accuracy of the CDL.

Fig. 4

Statistic analysis of 10-min averaged wind velocity measured by Molas B300 and the pulsed CDL. (a) Horizontal wind speed correlation, (b) horizontal wind direction correlation, (c) vertical wind velocity correlation, and (d) horizontal wind speed profile.

3.2.

Intercomparison Experiments of the CDL and Sounding Balloons

The time–altitude matching comparison of horizontal wind velocity between the CDL and sounding balloon, including ascent and descent periods, is presented in Fig. 5. The overall similarity of wind velocity of two sensors emerges while tiny distinction exists inevitably. The common discrepancy of horizontal wind speed of two sensors is $<1\mathrm{m}/\mathrm{s}$. The 10-min error bar is attached to the measurement dataset of our CDL. At altitude from 1500 to 3000 m, the error bar is so short due to the high SNR and relatively steady wind field. In the weak detection regime, the error bar increases with low SNR. However, the central measurement data retrieved from the CDL is still close to the measurement data of sounding balloons. The accuracy of the CDL is affected by weak signal regime, atmospheric turbulence, etc. At the end of descent period, the discrepancy of wind velocity measured by sounding balloon (dot) and the CDL (cross) mainly ascribes to the measured wind field by two sensors. The sounding balloon flies away (about 10 km) as time goes by (after 19 min).

Fig. 5

Comparison of horizontal wind velocity between the CDL (blue cross, with 10-min error bar) and sounding balloon (red dot). (a) and (b) The horizontal wind speed comparison and horizontal wind direction comparison during the ascent period of sounding balloon, respectively. (c) and (d) The horizontal wind velocity comparison during the descent period of sounding balloon.

Comparison of horizontal wind velocity is shown in Fig. 6. Comparison of wind direction shown in the first row shows good accordance in most situations. In the second row, wind speed detected by two instruments shows similar tendency, and a divergence of about 1 to $2\mathrm{m}/\mathrm{s}$ of 30-s averaged wind speed is found between the CDL and sounding balloon.

Fig. 6

Comparison of horizontal wind velocity detected by CDL and sounding balloon. Comparison of (a) wind direction and (b) wind speed.

Linear correlation and numerous analyses of wind velocity are shown in Fig. 7. The linear fits of scattering points fluctuate about the line $y=x$. In combination with the correlation coefficient ($R$: 0.974 of wind speed and 0.997 of wind direction), an obviously positive correlation between the wind dataset of the CDL and that of sounding balloon is displayed in Fig. 7. The typical discrepancy of wind speed and wind direction between the two systems is about $0.7\mathrm{m}/\mathrm{s}$ and 5.3 deg, respectively. Discarding the accidental error of sounding balloon of manual operation, the little bias of wind speed and direction is mainly affected by the intermediate frequency offset and due to north deflection of the CDL, respectively.

Fig. 7

Horizontal wind velocity correlation of (a) wind speed and (b) direction on August 2018.

To our knowledge, the routine sounding balloon system has an accuracy of horizontal wind speed $<1\mathrm{m}/\mathrm{s}$ and wind direction within 5 deg in the lower troposphere.^25,31 During the intercomparison experiments in Jingbian, China, as the reference with time period of 30 s, vertical resolution of sounding balloon is better than 150 m. Averaging time period shorter than 2 min cannot reach sufficiently smooth;²⁵ therefore, 30-s averaged wind velocity usually occurs natural turbulent fluctuations of wind field. By comparing the wind velocity measured by different principles, the good agreement of experimental results could testify the accuracy of the CDL with acceptable limits.

3.3.

Sample of Continuous Wind-Profile Measurements of the CDL

In August 2018, the field observation was launched and the continuous 3-D wind field datasets were obtained by the CDL. A measured continuous wind profiles were sampled and shown in Fig. 8. The maximum detection altitude on that day was 4400 m (local altitude: 1332 m), which provided sufficient detection range in the field experiments.

Fig. 8

Continuous wind profiles measurement by the ground-based CDL (local altitude: 1332 m) on August 1, 2018. (a) Horizontal wind speed profiles, (b) horizontal wind direction profiles, and (c) vertical wind velocity profiles.

According to the horizontal wind speed profiles [Fig. 8(a)], the horizontal wind speed was $>5\mathrm{m}/\mathrm{s}$ in most of the time. From 20:00 to 24:00, horizontal wind speed was intensive in range of 1500 to 2500 m. Meanwhile, the horizontal wind direction profiles [Fig. 8(b)] revealed that the horizontal wind direction was steady at that night.

In the vertical wind velocity profiles [Fig. 8(c)], sign setting on this ground-based system leads to positive relative speeds for downward winds or vice versa. Throughout the measurement period, vertical wind velocity is mostly positive corresponding to downdraft. Interestingly, the vertical wind velocity changes quickly and frequently from 16:00 to 19:00 and becomes relatively steady on that night.