The Earth's upper mesosphere and lower thermosphere serve as a critical interface where chemical and physical processes are influenced by energy transmission from both the lower atmosphere and near-Earth space (Hedin et al., 1989; Kelley, 2009). Neutral winds are essential for understanding the dynamics, transport, and energy budget in the mesopause region, making them a key parameter in atmospheric models (Christensen et al., 2009; Dhadly et al., 2023).

Over the past few decades, optical passive sounding techniques have become pivotal for precise measurements of neutral winds in the middle and upper atmosphere. Utilizing these techniques, Doppler winds are accurately determined by analyzing airglow emission lines from particles transported by atmospheric winds. Among these instruments, the Fabry-Perot Interferometer (FPI) is particularly noteworthy, providing high-precision Doppler shift measurements using a Fabry-Perot etalon. Equally important is the Michelson Interferometer (MI), which can be field-widened to obtain a high etendue, delivering outstanding signal-to-noise ratios and detection efficiency. After decades of refinement, both FPI and MI have achieved widespread adoption in ground-based and satellite-based wind measurement applications (Hays et al., 1981; Gault et al., 1996; Killen et al., 2006; McLandress et al., 1996). Despite the valuable observations provided by these instrument, neutral wind measurements remain sparse over long-term periods and across wide areas.

A notable variant of the Michelson Interferometer is one that utilizes a technology called Asymmetric Spatial Heterodyne Spectroscopy (ASHS), first proposed in 2007 (Englert et al., 2007). The ASHS is an inherent concept of spatial heterodyne spectroscopy (SHS) by extending the optical path of one interferometer arm to observe interferogram at high optical path differences (OPDs), which allows field-widening technology and interferometric measurements without moving parts while maintaining spectral resolution. Field-widening refers to modifying a conventional interferometer to enhance its ability to collect light over a wider field of view while maintaining its modulation efficiency. One application of this technology is the Redline DASH Demonstration Instrument (REDDI), which was built to make ground-based neutral wind observations by sounding atomic oxygen red line emission at 630 nm in thermosphere (Englert et al., 2010). Subsequently, a DASH interferometer, the Michelson Interferometer for Global High-Resolution Thermospheric Imaging (MIGHTI), was developed for the Ionospheric Connection Explorer (ICON) satellite to measure neutral winds by observing atomic oxygen green line emission at 557.7 nm and redline emission at 630.0 nm simultaneously in mesopause and thermosphere at an altitude of 90–300 km (Englert et al., 2017).

The cost-effective and precise ASHS instruments are ideally suited for establishing a ground-based network for thermospheric neutral wind observations. Compared with traditional MI, ASHS avoids complex scanning structures. Moreover, in comparison with MI and FPI, it can synchronously detect calibration spectral lines to mitigate the impact of instrumental drift. Wei et al. (2020) proposed a ground-based, thermally-insensitive monolithic ASHS interferometer for observing the atomic oxygen airglow at 630.0 nm. This configuration was verified in the laboratory and demonstrated excellent thermal stability (Dötzer, 2019; Liu et al., 2018; Wei et al., 2022). The methodology and design tools developed for that instrument are also adopted in the present work.

To the authors' knowledge, existing ground-based ASHS instruments are limited to observing the atomic oxygen red line in the middle and upper thermosphere. To extend wind measurements to a lower altitude region, it is advantageous to additionally observe the green line, which has its intensity peak just above the mesopause at around 96 km and spanning a layer approximately 10 km thick (Zhang and Shepherd, 2005). Detecting this faint emission from the ground is challenging due to the low layer's thinness. The scattering of artificial light sources in the lower atmosphere can reduce the sensitivity and precision of the detection.

Several ground-based instruments have been developed to measure neutral winds in mesopause region by observing the atomic oxygen green line. For instance, a three-channel FPI has been developed by Wu et al. (2004) to detect the green line in mesopause region. Another instrument is a variant of the Michelson Interferometer, named the E-Region Wind Interferometer (ERWIN II), which facilitates the detection of the green line in mesopause region (Kristoffersen et al., 2013). For a ground-based ASHS system, where constraints such as mass and volume are less critical, combining multiple channels within a single interferometer is less essential. Instead, separate instruments can be individually optimized and tailored to the geometry and radiative properties of their respective emission layers to improve observation quality and temporal resolution.

This work focuses specifically on the design, calibration, and verification of an ASHS instrument dedicated to observing the atomic oxygen green line and is organized as follows: Firstly, the concept of the ASHS interferometer and the principle of wind measurements are described in Sect. 2. In Sect. 3, the optical design principle is discussed. In the subsequent section, the ASHS instrument configuration and main parameters are presented. In order to evaluate the performance of the instrument, a thermal stability test and wind measurement calibration are carried out in Sect. 5. In Sect. 6, horizontal wind is retrieved from field observations and compared with co-located LiDAR measurements. Section 7 provides a summary of the study.

Figure 1 illustrates the configuration of a field-widening ASHS interferometer. The incident light is split into two beams by a beamsplitter, which then reflect off two gratings and recombine to form interference fringes, which can be modeled according to the characterization of wavefronts (Liu et al., 2018). The fringe intensity I can be expressed as 1I(x)∝1+VI(x)cos⁡2πLσ0-σLexp⁡-2π2σD2L2, where VI(x) is the visibility function of the interferometer, which is a function of sampling-position x on the detector, L=4tan⁡θx+2Δd is the OPD, θ is the Littrow angle corresponding to Littrow wavenumber σL, σ0 is the central wavenumber of the emission line, the term 2Δd represents the optical path difference between the nominal and extended arms of the interferometer at the Littrow wavelength, and σD is the Doppler broadening parameter, which is defined as 2σD=σ0kBTmc2 where, kB is the Boltzmann constant, T is the kinetic temperature, m is the mass of the emitter and c is the velocity of light. The cosine term is the phase of fringe in the interferogram. The exponential term damps the visibility of the interferogram due to Doppler broadening of the emission lines. Accordingly, the phase term of fringes in the interferogram can be described as a function of the sampling position x 3ϕ(x)=2π4tan⁡θx+2Δdσ0-σL. The phase difference between reference and Doppler-shifted interferograms can be expressed as: 4Δϕ(x)=ϕ′(x)-ϕ(x)=2π4tan⁡θx+2Δdσv-σ0, where σv is the Doppler-shifted wavenumber of the emission line. Then the Doppler velocity v can be determined from the phase difference Δϕ of two observations by 5v=Δϕ⋅c2πσ04tan⁡θx+2Δd.

The configuration of the ASHS instrument is depicted in Fig. 4, with component information provided in Table 2. Figure 5 shows photos of the interferometer module (a) and the assembled ASHS instrument (b), respectively. In addition to the ASHS, the complete instrument includes front optics that image the atmospheric scene onto the gratings and camera optics that image the gratings onto the focal plane array, which are designed separately. Within the front optics, a beamsplitter overlays the atmospheric scene with light from a krypton lamp, which is used to monitor instrument drifts. Furthermore, a mirror system is installed between the instrument and the atmosphere, allowing observations in the four cardinal directions and zenith to obtain zero wind measurements.

The M1, M2 mirrors are installed on a two-axis all-sky scanning system to collect airglow radiation in 45° elevation and zenith directions. The telescope L1 to L2 and the aperture stop are designed with a full field of view of 9° to ensure high luminous flux and signal-to-noise ratio of the instrument. The calibration line (λ = 557.029 nm) from a krypton lamp is attenuated by neutral density filter (ND) to an intensity that is comparable to the atomic oxygen green line. The collimated and holographically diffused calibration line is transmitted by a beamsplitter BS1 and enters the relay lens. The airglow and the calibration signal, collimated by relay lenses L4 to L5, both enter the interferometer and produce fringe images. F1 is a narrow band filter with a center wavenumber of 557.7 nm and a band width of 2 nm, so that background radiation and other spectral lines of the krypton lamp are suppressed.

The double telecentric system L6 to L7 with a magnification of 1.0 relays the fringe pattern from the localization plane of the interferometer to the focal plane of the detector. The 2048 × 2048 pixels weak signal detector with a pixel size of 13.5 µm acquires images in an area of 27.6 × 27.6 mm. To increase rigidity and reduce system deformation, the interferometer, telecentric system, and detector have been installed on a thick aluminum alloy plate. Additionally, a metal shell has been added to the plate. The components in dashed line in Fig. 4 are integrated into an insulation box made of insulation board filled with aluminum silicate fibers and covered with aluminum film so that the interferometer and optical systems are prevented from being impacted by rapidly changing ambient temperatures.

The interferometer is integrated into an active temperature-controlled housing, which includes an aluminum container, a polystyrene insulating enclosure, a set of heating tapes, a set of temperature sensors, and a PID (proportional-integral-derivative) controller. Enclosure windows W1 and W2 are installed at the entrance and exit positions. Temperature sensing is completed by platinum resistors and thermistors and a target temperature at 30 °C is set and maintained by the PID controller. The whole temperature-control system keeps the temperature inside the aluminum container better than ±0.1 °C.

Neutral wind measurements were conducted at Mohe Station (53.5° N, 122.3° E) during nighttime on 14 January 2024, when the solar zenith angle exceeds 100°. The exposure time for each sample is 5 min, and a complete scanning cycle, which includes measurements in the northward, eastward, zenith, southward, and westward directions, takes approximately 26 min. The darkfield is taken every night before observation by closing the entrance aperture of the detector. Figure 10 is an example of a raw image containing information of atomic oxygen green line and calibration line. There are notches marked by laser on the top of one grating, which is used for calibrating the thermal drift outside the interferometer (Marr et al., 2020). After removing the dark field and replacing the hot pixels and spikes, the image is binned 2 × 8, 2 pixels in x direction and 8 pixels in y direction, resulting in a final resolution of 1024 × 256 elements. In order to avoid large distortions at the interferogram edges, only 800 × 150 elements are used for retrieving wind.

Figure 11 shows the observational results of the geomagnetic quiet night on 14 January 2024 at Mohe station. After removing the outliers, caused by light pollution or detector sampling errors, the average phase in the zenith direction throughout one night is taken as the zero-wind phase. In order to validate the ASHS instrument, the results are compared with measurements from atmosphere wind-temperature-metal-constituents LiDAR at Mohe station. The red line represents a weighted average result of LiDAR measurements. The weighting function is a Gaussian function defined as: 13y(z)=1σ2πexp⁡-(z-z0)22σ2, where z0 is the peak height, σ is the standard deviation, and the Full Width at Half Maximum (FWHM) equals 22ln⁡2σ. We referred to Liu et al. (2025)'s results, and chose the peak height of 95 km and the FWHM of 12.5 km.

The error bars of the ASHS represent a combination of the uncertainty associated with the zero wind reference and the retrieved phase. On the one hand, since the vertical wind is used as a zero wind reference, its uncertainty can be represented by the standard deviation of the vertical wind throughout the night. On the other hand, According to Eq. (3), there is a theoretical distribution for the phase of the interferogram, with its value being contingent upon the sampling position x of the detector. By conducting a linear regression analysis on the relationship between the retrieved phase and x, we can derive the fitted phase. The uncertainty of the retrieved phase can be obtained by calculating the root mean square of the fitted phase and the retrieved phase. This combination provides a comprehensive indication of the precision and reliability of the data.

Several primary factors contribute to the occurrence of large error bars in the ASHS instrument observations: Firstly, they manifest at the beginning and end of the observation period, primarily due to the degradation of the SNR of the airglow signal caused by the scattering of sunlight in the atmosphere during sunrise and sunset. Secondly, artificial light pollution in the observational environment or detector sampling errors can also lead to such large error bars. Thirdly, meteorological conditions, such as cloudy or rainy weather, play a role. Additionally, inherent variations in the intensity of the airglow signals themselves are another contributing factor.

During the observation periods, the sky was clear and geomagnetic activities were quiet. After sunset, the meridional wind was directed northward while the zonal wind blew eastward. At about 11:30 UT, the meridional wind abruptly shifted to the southward direction and maintained a velocity of approximately 25 m s⁻¹ until it steadily reverted to a northward direction from 18:00 to 20:00 UT. During the period from 14:00 to 16:00 UT, the zonal wind gradually transitioned from an eastward to a westward direction, subsequently reversing back to an eastward flow for the remainder of the observational period.

The meridional and zonal wind measurements from the ASHS and the LiDAR exhibit a satisfactory consistency in terms of both the temporal synchronization and the magnitude of wind variations. This congruency not only attests to the reliability of the ASHS in capturing wind dynamics but also underscores the comparability of the two distinct measurement systems. Additionally, the consistency of the east-west zonal wind and north-south meridional wind is confirmed by the measurements, as the ASHS instrument's observations in opposite directions corroborate each other. Furthermore, the measured zenith wind remained close to zero throughout the night, indicating strong instrument stability.

To facilitate a comparative analysis of the correlation between data obtained from two distinct systems, it is imperative to standardize the observational outcomes. For this purpose, data from both ASHS and LiDAR systems, collected during overlapping observation periods spanning 5 d in January 2024. Given the disparate observation frequencies of these systems, piecewise linear interpolation was employed to synchronize their time series. Subsequently, the average horizontal wind component aligned with the north-south axis was designated as the meridional wind, while that aligned with the east-west axis was identified as the zonal wind. The processed time series results, with a five-minute interval, are presented in Fig. 12. The upper and lower panels of the figure illustrate the meridional and zonal wind results, respectively. Within the figure, red dots signify LiDAR observations, whereas blue dots represent observations from the ASHS interferometer. Notably, the observational data from both instruments exhibited good agreement on the nights of the 14 and 15 January. However, from 17 to 31 January, a certain degree of deviation was observed, which may primarily be attributed to inaccuracies in the weighted average calculation of LiDAR detection results across various altitudes. The spatial and temporal variability in the height distribution of oxygen atom OI 557.7 nm airglow necessitates a more precise weighted average computation, based on the specific airglow emissivity height distribution prevalent in the region on the respective day.

Figure 13 demonstrates the correlation between the observational outcomes of the ASHS interferometer and LiDAR. The vertical coordinates of the blue dots in the figure represent the observational data from the ASHS interferometer, while the horizontal coordinates correspond to the weighted average of the measurement results obtained from LiDAR. The red line depicted in the figure results from a linear regression fit utilizing the least squares method. The upper left corner of the figure indicates the linear fitting equation (y) with regression slopes (k = 0.880 ± 0.023 meridional; k = 0.812 ± 0.027 zonal) and the Pearson correlation coefficient (R = 0.845 meridional; R = 0.780 zonal), both of which highlight a significant positive correlation between the observational results of the two systems. In the linear fitting equation, the slope is observed to be less than 1. As found by Kristoffersen et al. (2024) and Liu et al. (2025) winds measured by different type of optical interferometers are smaller than weighted winds derived from lidars and meteor radars. It may result from that the assumed weighting function has a bias by comparing with the real airglow layer. Noting that in Fig. 13, the slopes of the fitted lines for meridional wind and zonal wind are inconsistent, which may stem from two factors. The field of view and detection scope exhibit differences between ASHS and lidar systems. ASHS detects winds in the four cardinal directions separately and then combines these measurements to derive the meridional and zonal winds. LiDAR measures winds in two orthogonal directions to obtain the meridional and zonal wind components. Under the assumption of a uniformly distributed wind field near the observation region, the slopes of the meridional and zonal winds in Fig. 13 should theoretically be identical. However, during the five-night observation period presented, fluctuations in the wind field distribution near the observation region may have occurred, potentially influenced by small-scale gravity waves, for instance. Another contributing factor could be the assumption of homogeneous airglow volume emissivity distribution in the area during the weighted – average calculation of LiDAR data. As the actual distribution is asymmetric, it causes a greater discrepancy between estimated and actual values in one direction. Consequently, this may result in the zonal fitting slope being lower than the meridional one.

We present a newly developed ground-based ASHS instrument for observing the atomic oxygen green line to derive neutral winds in mesopause region. The temperature of the interference module in the field observatory is kept within ±0.1 °C of the target temperature by utilizing a combination of active and passive temperature control systems. The thermal compensation design results in a spatial frequency thermal sensitivity of 0.038 cm⁻¹ °C⁻¹ and a phase offset thermal sensitivity of 0.624 rad °C⁻¹. In laboratory measurements, the sensitivity of the DASH instrument on Doppler velocity is calibrated, indicating an uncertainty of less than 2 m s⁻¹ due to the thermal sensitivity of the instrument. First field measurements taken at Mohe Station were presented, showing a well consistent picture of wind fields with the LiDAR measurements. The regression slopes less than 1 could potentially arise from a systematic underestimation inherent in the ASHS measurement technique or from biases introduced during the process of weighted averaging of wind data across various altitudes. Considering that the present comparative analysis is confined to data collected over a mere five-day period, a comprehensive and in-depth examination of the root causes necessitates the acquisition of extended-duration datasets.