New Algorithm Speeds Up Mesospheric Temperature Retrieval

Researchers from Wuhan University and the Nanjing University of Information Science and Technology have developed a fast spectral line-ratio algorithm, designated FASLR, to retrieve rotational temperatures in the mesopause region from airglow spectra. Published in Atmospheric Measurement Techniques, the study presents a non-iterative method that significantly reduces the computational time required to process long-duration, high-temporal-resolution observations from the Mesospheric Airglow Spectrum Photometer (MASP). By establishing ratio-temperature relationships using forward-modeled synthetic spectra, the algorithm enables efficient analysis of atmospheric wave activity and nocturnal temperature variations without the heavy computational burden of previous iterative inversion techniques.

The mesosphere and lower thermosphere, spanning approximately 50 to 120 kilometers above Earth, constitute a critical transition zone between the lower atmosphere and the thermosphere-ionosphere system. Within this region, the mesopause is of particular importance because its temperature serves as a key parameter for characterizing multiscale dynamical processes and associated energy transfer. However, the thermal structure of the mesopause exhibits pronounced spatiotemporal variability, making single-point observations often insufficient to capture both regional structure and temporal evolution. Passive optical remote sensing, specifically the analysis of nocturnal airglow emissions from oxygen and hydroxyl, is well-suited for these temperature observations because the rotational spectral features contain line intensities that vary systematically with temperature.

Previous methods for retrieving rotational temperature, such as the multi-parameter iterative inversion algorithm (MPII), required iterative adjustment of instrumental parameters and repeated forward-model evaluations for each observed spectrum. While accurate, this approach is computationally intensive, which constrains its use for processing continuous, high-temporal-resolution data. The new FASLR algorithm addresses this limitation by using selected rotational lines in the oxygen (0-1) band to construct line-intensity ratios. These ratios are converted to temperature estimates using fitted relationships derived from forward-modeled synthetic spectra, allowing for a direct, non-iterative retrieval process that bypasses the need for per-spectrum optimization.

To account for instrumental perturbations, FASLR employs a sensitivity-based weighting scheme. The algorithm evaluates the response of three specific line-intensity ratios to variations in filter temperature and optical-system focal length. Ratios that are less affected by these instrumental perturbations are assigned larger weights in the final temperature calculation. This weighting is determined by the relative sensitivity of each ratio to the instrumental parameters, ensuring that the most stable measurements contribute most significantly to the final rotational temperature estimate. The final temperature is calculated as a weighted average of the individual ratio-based estimates, with weights evaluated at a common reference temperature to maintain consistency.

The computational efficiency of FASLR was demonstrated through comparisons with MPII using data from two observation nights in 2023. For a night containing 132 observations, FASLR completed the processing in 19.0 seconds, whereas MPII required 296.7 seconds. On another night with 120 observations, FASLR took 16.8 seconds compared to 350.0 seconds for MPII. This represents a reduction in processing time by a factor of approximately 15.6 to 20.8 times. Such efficiency is crucial for enabling real-time or near-real-time analysis of continuous airglow observations, which are essential for studying short-term temperature variability and wave dynamics in the upper atmosphere.

Validation of the FASLR algorithm involved comparisons with independent measurements from a sodium fluorescence Doppler lidar at the National Space Science Center in Beijing. On two nights in 2023, the FASLR-derived temperatures showed generally similar major nocturnal variations to the lidar measurements, including cooling before midnight and subsequent warming. However, differences in absolute temperature levels were observed, with FASLR temperatures generally higher than the lidar temperatures by mean differences of 8.65 K and 5.38 K on the two nights. The study notes that while the temporal evolution is well-captured, a broader assessment of absolute temperature accuracy would benefit from additional simultaneous comparisons over more observation periods.

The algorithm was further applied to six nights of dual-channel MASP observations at the Tulihe Meteorological Observatory in Inner Mongolia, China, from 2 to 7 February 2024. These observations included both oxygen (0-1) and hydroxyl (6-2) emissions. The retrieved oxygen rotational temperatures were mainly distributed within 182 to 212 K, while hydroxyl temperatures fell within 197 to 227 K. Comparisons with satellite observations from SABER and Aura/MLS showed that the available satellite temperatures generally fell within or close to the range of the FASLR retrievals. Differences in absolute levels were attributed to the different sampling characteristics and spatial representativeness of the satellite and ground-based observations.

Model comparisons with SD-WACCMX and NRLMSIS 2.1 revealed distinct relationships for the two emission layers. For oxygen, the FASLR temperatures were generally closer to NRLMSIS 2.1 in overall temperature level, although their nocturnal variations differed. In contrast, SD-WACCMX generally gave lower temperatures and weaker nocturnal variability. For hydroxyl, the FASLR temperatures were higher than NRLMSIS 2.1 early in the night but became closer later. The study suggests that these differences may partly result from variations in the effective emission height and layer thickness, which can affect comparisons based on fixed altitude ranges.

Case studies using the FASLR-derived temperatures illustrated the algorithm’s capability for wave diagnostics. On 2 February 2024, spectral analysis identified a common dominant period of approximately 6.8 to 6.9 hours in both temperature and intensity perturbations for the oxygen and hydroxyl layers. The temperature variation led the intensity variation by about 15.4 degrees in the oxygen layer and 17.4 degrees in the hydroxyl layer. On 7 February 2024, wavelet analysis detected concurrent short-period power enhancements in both layers, with dominant periods between 1.5 and 2.2 hours. Simultaneous observations from the Mesospheric Airglow Wide-angle Imager (MAWI) independently supported the presence of a short-period gravity-wave disturbance during this interval, showing coherent banded structures in the imager frames.

The development of FASLR provides a robust and efficient tool for analyzing mesospheric airglow observations. By reducing computational costs while maintaining the ability to capture major temporal variations and wave signatures, the algorithm supports investigations of multiscale wave processes in the mesopause region. The study highlights the importance of efficient data processing for making full use of continuous, high-temporal-resolution measurements in atmospheric research. Future efforts may explore the development of frameworks for retrieving two-dimensional temperature-perturbation fields by combining FASLR temperatures with wave-structure information from imager sequences, further enhancing the understanding of atmospheric dynamics in the upper atmosphere.

The FASLR algorithm specifically targets the O2 (0–1) airglow emission, which peaks near 94 km altitude, and the OH (6–2) emission, which peaks near 87 km. In the dual-channel configuration of the Mesospheric Airglow Spectrum Photometer, the instrument alternates between measuring these two emissions. The exposure time is set to 3 minutes for the oxygen channel and 2 minutes for the hydroxyl channel. Because the measurements alternate, successive observations of the same emission band are separated by approximately 5.3 minutes. This effective temporal resolution allows the retrieval of temperature series that capture short-term atmospheric fluctuations while maintaining the signal-to-noise ratio necessary for accurate spectral line analysis. The instrument uses interference filters with center wavelengths of 867.1 nm for oxygen and 837.17 nm for hydroxyl, with bandwidths of 0.24 nm and 0.26 nm, respectively. These specific optical settings are critical for isolating the rotational spectral features used in the temperature retrieval process.

The physical basis for the retrieval relies on the assumption of near-local thermodynamic equilibrium for the lower rotational levels of the O2 (0–1) emission. Under this condition, the relative intensities of rotational lines within a vibrational band vary systematically with rotational temperature. The algorithm selects four specific rotational lines, designated P1(1) through P1(4), from the oxygen band. Among these, the P1(3) line is chosen as the reference because it exhibits the weakest temperature dependence over the considered range. Three line-intensity ratios are constructed: P1(1) / P1(3), P1(2) / P1(3), and P1(3) / P1(4). These ratios are fitted as functions of temperature over a range of 130 to 280 K using forward-modeled synthetic spectra. By using ratios rather than absolute intensities, the method reduces the influence of common instrumental effects, although it cannot fully remove line-dependent instrumental responses or the unequal sensitivities of different ratios to instrumental-parameter variations.

The sensitivity-based weighting scheme is designed to mitigate the impact of instrumental perturbations on the final temperature estimate. The algorithm evaluates how the three selected line-intensity ratios respond to variations in filter temperature and optical-system focal length. Filter temperature is varied over a range of 18 to 27 degrees Celsius, while focal length is varied over 53.0 to 53.9 mm. The standard deviations of the ratios over these perturbation ranges are calculated to quantify their sensitivity. Ratios that show less variation under these perturbations are assigned larger weights in the final calculation. This ensures that the most stable measurements contribute most significantly to the final rotational temperature estimate. The weights are temperature-dependent and are evaluated at a common reference temperature, defined as the median of the three individual ratio-based estimates, to maintain consistency in the weighted average.

The study acknowledges several limitations in the validation of the FASLR algorithm. While the temporal variations retrieved by FASLR are generally similar to those from independent lidar measurements, differences in absolute temperature levels remain. The mean temperature differences between FASLR and lidar were 8.65 K and 5.38 K on the two comparison nights. The authors note that a broader assessment of absolute temperature accuracy would benefit from additional simultaneous comparisons with independent temperature measurements over more observation periods. Furthermore, the comparison with model temperatures from SD-WACCMX and NRLMSIS 2.1 focuses on the background temperature level and slowly varying nocturnal evolution rather than short-term fluctuations, due to the coarser spatiotemporal resolution of the model fields. The differences between ground-based OH rotational temperatures and model temperatures may partly result from variations in the effective emission height and layer thickness, which can affect comparisons based on fixed altitude ranges.

The data supporting this study are archived in Zenodo, including the raw dual-channel MASP observations from the six nights in February 2024 and the processed temperature comparison data from the 2023 Beijing observations. The code used for spectral processing and rotational temperature retrieval is not publicly archived but is available from the corresponding author upon reasonable request. The study highlights the potential of FASLR for broader deployment in observational networks, where efficient and reliable data processing is essential for making full use of continuous, high-temporal-resolution measurements. By reducing the computational cost of temperature retrieval, the algorithm supports investigations of multiscale wave processes and long-term changes in the upper atmosphere, contributing to a better understanding of the mesopause region’s role in atmospheric dynamics.

Subject of Research: Development and validation of a fast spectral line-ratio algorithm for mesospheric airglow rotational temperature retrieval

Article Title: A fast spectral line-ratio algorithm for mesospheric airglow rotational temperature retrieval and validation

Article References: Niu, S., Gu, S., Gao, H., Gong, D., Wei, Y., & Qin, Y. (2026). A fast spectral line-ratio algorithm for mesospheric airglow rotational temperature retrieval and validation. Atmospheric Measurement Techniques, 19(21), 7113-7129. https://doi.org/10.5194/amt-19-7113-2026

Image Credits: AI Generated

DOI: 10.5194/amt-19-7113-2026

Keywords: mesosphere, airglow, temperature retrieval, atmospheric waves, remote sensing, spectroscopy, FASLR, MASP, fast, spectral, line-ratio, algorithm