Clear Nights Research Articles

Diurnal changes in cloud cover in eastern Gabon and their impacts on energy balance, light availability and water demand: a case study of the 2022 dry season

Abstract Western Central Africa is atypical of the equatorial domain as the main dry season is cloudier than the rainy seasons. To understand this cloud cover's diurnal evolution, we set-up an infrared camera and acquired measurements of the total cloud cover fraction (TCF) and cloud optical depth at Bambidie, Gabon (0°44’30.5” S,12°58’12.4” O) from May to October 2022. Diurnal variations in TCF can be summarized into four types, mostly discretized through the timing and duration of clouds clearing in the afternoon (Early afternoon Clearing: EaC, Late afternoon Clearing: LaC and Clear Night: CNi) while one type (No Clearing: NoC) shows overcast conditions all day long. Meteorological measurements show that NoC days record 50W/m2 less shortwave incoming surface radiation resulting in daytime temperatures 1°C lower than the seasonal norm, but 20% more diffuse light and 0.5mm/day less ETo. Conversely, EaC days record 50W/m2 more shortwave incoming surface radiation leading to temperatures 1.5°C higher than the seasonal norm, but 40% more direct light. The larger water demand (0.5mm/day more ETo) is partly compensated by more frequent rainfall at night-time. The SAFNWC satellite estimates well capture the TCF variations for most of the 4 types. They confirm that TCF is dominated by very low and low clouds whose dissipation in the afternoon and evolution into fractional and cumuliform convective clouds explains the clearings on EaC and LaC days. Satellite estimates also show that the 4 types of days extracted at Bambidie are representative of a larger-scale cloud cover evolution in Western Central Africa, with a W-E gradient in the timing of afternoon cloud dissipation.

Estimating Visual Acuity Without a Visual Acuity Chart.

This study explored whether visual acuity (VA) can be inferred from self-reported ability to recognize everyday objects using a set of yes/no questions. Participants answered 100 yes/no questions designed to assess their ability to recognize familiar objects at typical viewing distances, such as distinguishing between a full moon and a half moon on a clear night. The questions demanded VA ranging from normal to severe vision impairment. Responses were analyzed using item response theory, and the results were compared with participants' VA values. We recruited 385 participants from 4 sites in the United States. Participants had a mean age of 56.7years with VA ranging from -0.3 to 2.0 logarithm of the minimum angle of resolution (logMAR) (mean = 0.58). A strong relationship was observed between participants' estimated vision ability and their VA (r = -0.72). The linear relationship can be used to predict each participant's VA based on their estimated vision ability. The average signed and unsigned prediction errors were 0 and 0.24 logMAR, respectively, with a coefficient of repeatability of 0.59 logMAR between the estimated VA and measured VA. The same linear function was used to determine the VA limit required for each question. For instance, the VA limit for the moon question was 1.0 logMAR. Yes/no questions about everyday visual activities have the potential to estimate an individual's VA. Future refinements may enhance reliability. The survey provides insights into the real-world visual capabilities of people with low vision, making it potentially useful for telehealth applications.

Accuracy of daily extreme air temperatures under natural variations in thermometer screen ventilation

AbstractAccurate air temperatures underpin environmental research. Most professional meteorological air temperature measurements still expose thermometers within traditional, naturally ventilated screens. Their representation of true air temperature depends on screen airflow, and therefore local winds. Accuracies of daily maximum (Tmax) and minimum (Tmin) air temperatures are assessed by comparison between a naturally ventilated large conventional screen and a co‐located aspirated reference screen. In over 1200 days' data, the naturally ventilated Tmin and Tmax both showed small (median < 0.06°C) cold bias, but, in 1% of cases, warm Tmax bias and cold Tmin bias >|1°C|. The Tmin cold bias is associated with calm clear nights, and the Tmax warm bias events with calm winter days at low sun angles, allowing solar heating of the screen. The prevalence of poor natural ventilation, potentially affecting Tmin and Tmax, is estimated across European sites. Poor ventilation occurred at Tmin for 12% of values, and at Tmax for 4%. Climatological averaging will reduce these effects, but, without corroborating wind data, statistical changes in Tmin or Tmax, including identifying “Tropical Nights” (Tmin > 20°C) or occurrences of winter extremes, may have limited value. Wider adoption of aspirated thermometer screens, with an initial overlap period, will largely eliminate these effects.

The effect of taillight shapes and vehicle distance on rearward drivers’ hazard perception

Rear-end collisions are prevalent during night driving. Taillights are vital in conveying hazards between leading and following vehicles at night. Taillights highlight a vehicle’s presence, particularly for vehicles following from behind. Therefore, it is essential to design taillights that maximize the perception of hazards for drivers behind them. This study explored how different taillight shapes affect rear drivers’ hazard perceptions during nighttime driving. Two experiments were conducted using ERP measurements to study the effects of taillight shapes (three square types in Experiment 1: solid, array, and contour; two linear types in Experiment 2: through type and non-through type) and the distance to the leading vehicle on participants’ hazard perceptions. Participants responded to images of nighttime driving scenarios during the experiments while their ERP and behavioral data were recorded. The stimulus images showed the view from a driver looking at the vehicle ahead on a clear night. The neural process of hazard perception consists of two stages: automatic detection (early) and evaluation (later). In this study, P2 indicated attention bias to the stimulus during the detection stage, and LPP indicated the negative emotion triggered by the stimulus during the evaluation stage. Experiment 1 showed that solid-shaped taillights were perceived as more hazardous and processed faster than other square taillight shapes. Experiment 2 found that non-through-type and through-type linear taillights were perceived as more hazardous during the automatic detection and subjective evaluation stages, respectively. However, the behavioral data showed that through-type taillights were considered more hazardous and were associated with shorter response times. Thus, solid square taillights and through-type linear taillights can be optimal ergonomic solutions. These findings can serve as a reference for taillight designers, manufacturers, and potential car buyers regarding safety considerations.

Intermediate complexity atmospheric modeling in complex terrain: is it right?

Dynamic downscaling of atmospheric forcing data to the hectometer resolution has shown increases in accuracy for landsurface models, but at great computational cost. Here we present a validation of a novel intermediate complexity atmospheric model, HICAR, developed for hectometer scale applications. HICAR can run more than 500x faster than conventional atmospheric models, while containing many of the same physics parameterizations. Station measurements of air temperature, wind speed, and radiation, in combination with data from a scanning Doppler wind LiDAR, are compared to 50 m resolution HICAR output during late spring. We examine the model’s performance over bare ground and melting snow. The model shows a smaller root mean squared error in 2 m air temperature than the driving model, and approximates the 3D flow features present around ridges and along slopes. Timing and magnitude of changes in shortwave and longwave radiation also show agreement with measurements. Nocturnal cooling during clear nights is overestimated at the snow covered site. Additionally, the thermal wind parameterization employed by the model typically produces excessively strong surface winds, driven in part by this excessive nocturnal cooling over snow. These findings highlight the utility of HICAR as a tool for dynamically downscaling forcing datasets, and expose the need for improvements to the snow model used in HICAR.

Analysis of night behavior and negative running for PVT system

High absorption of PV benefits heat collection in the daytime, but the heat loss to the space/environment because of its high emissivity also should be noticed. Although there are many papers investigating PVT, few of them cover the night behavior and negative running during late afternoon. This paper supplements detailed studies on related content to draw people's attention to the night and late-afternoon operation that helps better strategy establishment and less heat loss. From experiments, the clouds help hinder the PV radiant heat exchange with space. Under clear night, the PV bottom temperature can be 6.88 °C lower than air temperature and the water tank's bottom temperature drops 2.8 °C. Because of the fluid viscosity, no apparent temperature stratification and flow occur in the PVT water pipe, causing even PV temperature. The higher the water temperature, the easier the negative running occurs during late afternoon, although the irradiation is still high at 521W/m2. Heat loss power and energy are −199W and 1791 KJ from 15:10 to 17:40. Based on different regressed linear correlations of different PVT systems, the effect of water temperature, irradiation, and ambient temperature on the negative performance is also comparatively studied.

Contrastive Learning for Low-Light Raw Denoising (Student Abstract)

Image/video denoising in low-light scenes is an extremely challenging problem due to limited photon count and high noise. In this paper, we propose a novel approach with contrastive learning to address this issue. Inspired by the success of contrastive learning used in some high-level computer vision tasks, we bring in this idea to the low-level denoising task. In order to achieve this goal, we introduce a new denoising contrastive regularization (DCR) to exploit the information of noisy images and clean images. In the feature space, DCR makes the denoised image closer to the clean image and far away from the noisy image. In addition, we build a new feature embedding network called Wnet, which is more effective to extract high-frequency information. We conduct the experiments on a real low-light dataset that captures still images taken on a moonless clear night in 0.6 millilux and videos under starlight (no moon present). The results show that our method can achieve a higher PSNR and better visual quality compared with existing methods.

NightRain: Nighttime Video Deraining via Adaptive-Rain-Removal and Adaptive-Correction

Existing deep-learning-based methods for nighttime video deraining rely on synthetic data due to the absence of real-world paired data. However, the intricacies of the real world, particularly with the presence of light effects and low-light regions affected by noise, create significant domain gaps, hampering synthetic-trained models in removing rain streaks properly and leading to over-saturation and color shifts. Motivated by this, we introduce NightRain, a novel nighttime video deraining method with adaptive-rain-removal and adaptive-correction. Our adaptive-rain-removal uses unlabeled rain videos to enable our model to derain real-world rain videos, particularly in regions affected by complex light effects. The idea is to allow our model to obtain rain-free regions based on the confidence scores. Once rain-free regions and the corresponding regions from our input are obtained, we can have region-based paired real data. These paired data are used to train our model using a teacher-student framework, allowing the model to iteratively learn from less challenging regions to more challenging regions. Our adaptive-correction aims to rectify errors in our model's predictions, such as over-saturation and color shifts. The idea is to learn from clear night input training videos based on the differences or distance between those input videos and their corresponding predictions. Our model learns from these differences, compelling our model to correct the errors. From extensive experiments, our method demonstrates state-of-the-art performance. It achieves a PSNR of 26.73dB, surpassing existing nighttime video deraining methods by a substantial margin of 13.7%.

Near-Surface Wind Profiling in a Utility-Scale Onshore Wind Farm Using Scanning Doppler Lidar: Quality Control and Validation

Wind profiling within operating wind farms is important for both wind resource assessment and wind power prediction. With increasing wind turbine size, it is getting difficult to obtain wind profiles covering the turbine-affecting area due to the limited height of wind towers. In this study, a stepwise quality control and optimizing process for deriving high-quality near-surface wind profiles within wind farms is proposed. The method is based on the radial wind speed obtained by the Doppler Wind Lidar velocity-azimuth display (VAD) technique. The method is used to obtain the whole wind profile from ground level to the height affected by wind turbines within a utility-scale onshore wind farm, in northern China. Compared with the traditional carrier-to-noise ratio (CNR) filter-based quality control method, the proposed data processing method can significantly improve the accuracy of the derived wind. For a 10 m wind speed, an increase in coefficient of determination (R2) from 0.826 to 0.932, and a decrease in mean absolute error (MAE) from 1.231% to 0.927% are obtained; while for 70 m wind speed, R2 increased from 0.926 to 0.958, and MAE decreased from 1.023% to 0.771%. For wind direction, R2 increased from 0.978 to 0.992 at 10 m, and increased from 0.983 to 0.995 at 70 m. The optimized method also presents advantages in improving the accuracy of derived wind under complex wind environments, e.g., inside a wind farm, and increasing the data availability during clear nights. The proposed method could be used to derive wind profiles from below the minimum range of a vertically operating scanning Doppler Lidar to a height affected by wind turbines. Combined with Doppler beam-swinging (DBS) scanning data, the method could be used to obtain the complete wind profile in the boundary layer. These wind profiles could be further used to predict wind power and evaluate the climate and environmental effects of wind farms.

Night sky brightness trends above Zagreb, 2012–2022

The night sky brightness at the RGN site (near the centre of Zagreb, Croatia) was monitored from January 2012 to August 2022. The first data set, covering the period 2012 to 2017, has already been analysed and the conclusions were published in Andreić (2018). The main conclusions are repeated here for comparison and completeness. The results of the analysis of the second dataset, covering the period from January 2018 to July 2022, are reported here. The first data analysis showed that the average night sky brightness did not change significantly from 2012 to 2017, excluding differences due to annual variations in meteorological parameters. The second data set (2018 to 2022) showed a slight decrease in night sky brightness compared to the first data set. The difference is small, about 0.2 mag/arcsec2 (mean values). In the second data set, there is a trend toward darker nights in later years that began three or four years ago (around 2019 or 2020). It is attributed to the modernisation of the public lighting network, where old, often very poor lighting fixtures are slowly being replaced by modern LED lighting fixtures, and to the effects of climatic changes leading to warmer winters with more clear nights.

Nocturnal downward transport of NO3 radical from the residual layer to a surface site by the mountain breeze in Seoul, South Korea

Using the long-path differential optical absorption spectroscopy (LP-DOAS) technique, the nocturnal mixing ratios of NO3 radicals were measured at an urban site in Seoul, South Korea, from May 13 to June 12, 2022. The primary objectives of the study were to estimate the NO3 radical concentration in the nighttime boundary layer and to examine the variations in their generation and loss over time, including an evaluation of their sources and sinks. The observed mean NO3 mixing ratios and lifetimes were 16 pptv and 73 s, respectively. NO3 concentrations and lifetimes were closely associated with their production, involving O3 and NO2, as well as the loss processes regulated by NO, VOCs, and aerosol surface area. Applying the steady-state assumption for NO3 and N2O5, the mean N2O5 uptake coefficient on the aerosol surface for the entire period was 0.031. When foggy nights were excluded from the analysis, the uptake coefficient decreased to 0.025. These values are within the higher range of the results reported in other studies. High relative humidity, which averaged 75% at night throughout the collection period, favored high N2O5 uptake coefficients. NO3 concentrations generally peaked shortly after sunset and gradually declined until sunrise. However, we frequently observed distinct increases in NO3 throughout the early morning hours, especially during clear nights, when local circulation with the westerly mountain breeze prevailed at the research site. The mountain breeze transported residual air downward, resulting in 110% and 38% higher NO3 and O3 levels at the sampling locations, respectively, compared to normal conditions. This additional source of nocturnal radicals and O3 has consequences for increased secondary aerosols and O3 production, particularly at urban surface levels where they are readily depleted in the early morning hours. To this end, a more thorough examination is required to measure N2O5 and a larger spectrum of VOCs species, as these parameters are crucial for nighttime radical budgets, secondary aerosols, and O3 generation in cities like Seoul.

Nighttime bionic compass based on a short-wave infrared polarization sensing system.

Sky-bionic polar co-ordinate navigation is an effective means of providing navigational information in the absence of a priori information. Polar co-ordinate navigation during clear daytime conditions has been studied, but there has been a lack of research of it at night due to problems with noise. Therefore, in this paper, a short-wave infrared polarimetric sensor system is designed, which is capable of acquiring atmospheric polarimetric information in low illumination environments at night, compared with traditional visible band sensors. Additionally, based on the statistics of polarization angle information, an algorithm for removing noise and starlight is proposed to solve the influence of starlight and noise on the polarization information at night. After many outdoor experiments, we found that the method can output the heading angle stably and accurately, and its standard deviation is controlled to be 0.42° in a clear night.

Enhancing solar efficiency around the clock through simultaneous solar energy harvesting and radiative cooling

Despite global urbanization, a significant part of the world's population lacks access to the electrical grid. Photovoltaic (PV) cells offer off-grid electricity but face efficiency challenges and shortened lifespans due to prolonged exposure to high temperatures. While PV cells traditionally generate electricity only during daylight, our innovative system integrates radiative cooling (RC), thermoelectric generators (TE), and phase change materials (PCM). In comprehensive tests, our system effectively lowered the PV module temperature by 15 °C during daylight at 800 W/m2 irradiance. Under a clear night sky, it achieved a remarkable 20 % improvement in power generation (120 mV) compared to previous studies. This advancement positions our system as a continuous power source for sustained electricity generation in off-grid settings, contributing to temperature moderation and extending Photovoltaic module lifespan without compromising efficiency. To optimize performance, we used COMSOL software, analyzing factors like wind speed, ambient temperature, irradiation intensity, and the photovoltaic cell to thermoelectric generator area ratio. Insights from this analysis form the basis for future enhancements and upgrades.

Using an Absolute Cavity Pyrgeometer to Calibrate Pyrgeometers Outdoors with Respect to the International System of Units

Accurate measurement of the atmospheric longwave irradiance is important for renewable energy and atmospheric science applications. Pyrgeometers are deployed outdoors all over the world to measure the atmospheric longwave irradiance and presently are calibrated with traceability to the interim standards for atmospheric longwave radiation measurement, the standards are based on four pyrgeometers and their average irradiance is the World InfraRed Standard Group (WISG) which is developed and maintained by The Physikalisch-Meteorologisches Observatorium Davos/World Radiation Center (PMOD/WRC). Since 2013 the InfraRed Integrating Sphere (IRIS) developed by PMOD/WRC and the Absolute Cavity Pyrgeometer (ACP) developed by the National Renewable Energy Laboratory (NREL) have been compared outdoors six times at different locations and the difference between the measured atmospheric longwave irradiance by ACP and IRIS was less than 2 W/m<sup>2</sup> with traceability to the International System of Units (SI). During the six comparisons the irradiance measured by the interim WISG was 5 W/m<sup>2</sup> lower than the irradiance measured by the average irradiance measured by the ACP and IRIS [1]. Based on this discrepancy, the World Meteorological Organization’s Commission for Instruments and Methods of Observation (CIMO) recommended that the interim WISG should be adjusted to be traceable to SI units [2]. In anticipation of CIMO’s expert team agreement on establishing the world reference using the average irradiance measured by ACP and IRIS in this article we describe a procedure to calibrate pyrgeometers with traceability to SI. One Absolute Cavity Pyrgeometer (ACP95F3) was used to calibrate four pyrgeometers traceable to SI units. Three Eppley PIRs and one Kipp&Zonen CG4 were originally calibrated with traceability to the interim WISG. Using the described procedure below, the responsivity of each pyrgeometer was then adjusted to match the irradiance measured by ACP. Outdoor data was collected during one clear sky nights monitored by the output thermopile voltage of ACP95F3. The irradiance measured by the PIRs was calculated using NREL equation and the CG4 using NREL equation and PMOD/WRC equation. Using the NREL equation, the calculated uncertainty (<i>U</i><sup>95</sup>) of the PIRs varied from 2.43 W/m<sup>2</sup> to 2.67 W/m<sup>2</sup>, and for the CG4 using the NREL equation <i>U</i><sub>95</sub> equals 1.97 W/m<sup>2</sup>, and using the PMOD equation <i>U</i><sub>95</sub> equals 2.88 W/m<sup>2</sup> with respect to SI.

Lunar crater names. The invisible face of epistemic violence shining upon us every night


 
 
 
 Lunar craters are the only and most significant surface features of a planetary body in Outer Space that are visible to the naked eye. While we can only look at the surface of other planets and their moons through probes, robots and telescopes, the moon requires no instruments or technological tools to be seen. Despite this significant difference – that the Moon’s surface features are visible to any human on a clear night, the history and current process of naming Lunar craters does not recognize that dif- ference. Rather, Lunar craters are subject to the same type of seemingly scientific naming process as that ascribed to the remaining 40 planetary bodies in our Solar System that currently have a naming protocol for surface features. In this paper, I argue that both this lack of distinction and the resistance (under the umbrella of both historical precedent and scientific selenographic naming conventions) to name changes of prominent Lunar craters has resulted in a unique form of post-terrestrial epistemic violence – a violence evident in global social pedagogies of outer space that remain silent on this most important issue of who has the right to name the heavens.
 
 
 

Canopy Structure and Air Temperature Inversions Impact Simulation of Sub‐Canopy Longwave Radiation in Snow‐Covered Boreal Forests

AbstractLongwave radiation is often the dominant source of energy for snowmelt in forests. Measurements at forest sites of varying density in Sweden and Finland show that downwelling longwave radiation is enhanced under forest canopies, even for sparse canopies and particularly for clear skies. Canopy density must be estimated accurately to predict this enhancement. Linear regression with above‐canopy longwave radiation and air temperature as predictors of sub‐canopy radiation gives good predictions of sub‐canopy longwave radiation with weightings for transmission through canopy gaps that are close to measured sky view fractions. Air temperature serves here as a proxy for effective canopy radiative temperature. Adding above‐canopy shortwave radiation as a predictor gives little improvement in the predictions, suggesting that daytime heating of trunks above the air temperature (“hot trees”) has limited influence on longwave radiation under these continuous canopies. The influence of canopy temperatures falling below the above‐canopy air temperature (“cold trees”) on calm, clear nights, however, is apparent. Decoupling of canopy and above‐canopy air temperatures in an energy balance model of the type used in large‐scale land surface modeling allows this cooling.

Thermally stable near-infrared emission from a double perovskite K2LiScF6:Cr3+ phosphor for light-emitting diodes

The exploration of Cr3+-doped near-infrared (NIR) phosphors for light-emitting diodes has attracted widespread attention. In this work, we present a fluoride double perovskite K2LiScF6:Cr3+ NIR-emitting phosphor through a co-precipitation method. It was found that the Cr3+ activators undergo a weak crystal field and present a NIR emission band peaking at 768 nm under blue light excitation (430 nm). Meanwhile, the K2LiScF6:Cr3+ exhibits a high activation energy (0.537 eV), a weak electron-phonon coupling (EPC) effect and a remarkable photoluminescence thermal stability at 423 K, with an intensity of 95.1% of that at room temperature. Profiting from the luminescent properties, clear night vision images and human palm vein photographs were realized by coating the K2LiScF6:Cr3+ phosphor on a blue InGaN chip and using it as a lighting source, suggesting its applicability in light-emitting diodes for NIR spectroscopy applications.

Transmuting Trauma and ‘Alchemiz[ing] Inherited Wounds’: iele paloumpis’s disabled, diasporic choreographies of care

In the programme notes for the premiere of In place of catastrophe, a clear night sky at Danspace Project in May 2022, choreographer iele paloumpis asks, ‘What vitality allowed our ancestors to survive generations of trauma, and what wisdoms have been passed down to us?’ Moving with and through grief surrounding various connected histories of violence, paloumpis and collaborators simultaneously find modes of tuning into various connected histories of resilience. This essay explores how paloumpis and collaborators collectively tap into disability and diaspora to transform intergenerational trauma into knowledge, power and more livable forms of relationality. Through overlapping practices of textile weaving, interdisciplinary somatic practices, audio description and sculpture, the piece enacts multisensorial choreographies of care that allow for alchemical grief processes to unfold between human performers themselves, between performers and non-human materials, and between performers and audience members. Approaching choreographic practices as care practices and care practices as choreographic practices, I use the phrase ‘choreographies of care' to refer to disabled, diasporic relationalities in performance. My engagement with these relationalities takes cues from Leah Lakshmi Piepzna-Samarasinha's conceptions of ‘care webs' (2018), Édouard Glissant's theories of ‘Relation' and ‘opacity' (1997) and Alice Sheppard's discussions of care as an ‘anchor point' of disability culture in performance contexts (2021). A certified end-of-life doula and grief worker, paloumpis carves out room for shared mourning on their own terms through a process of what they call ‘alchemiz[ing] inherited wounds into collective care’ (2022b). In this article, I will attend to some of the ways in which paloumpis and collaborators move through this alchemical transmutation of trauma by weaving across differing but connected experiences of disability, chronic illness and diaspora.

Site-testing at the Muztagh-ata Site V. Nighttime Cloud Amount during the Last Five Years

The clarity of nights is the major factor that should be carefully considered for optical/infrared astronomical observatories in site-testing campaigns. Cloud coverage is directly related to the amount of time available for scientific observations at observatories. In this article, we report on the results of detailed night-time cloud statistics and continuous observing time derived from ground-based all-sky cameras (ASCs) at the Muztagh-ata site from 2017 to 2021. Results obtained from acquisition data show that the proportion of the annual observing time at the Muztagh-ata site is 65%, and the best period with the least cloud coverage and longer continuous observing time is from September to February. We made a comparison of the monthly mean observing nights obtained from our ASC and CLARA data set, and results show that the discrepancy between them may depend on the cloud top heights. On average, this site can provide 175 clear nights and 169 nights with at least 4 hr of continuous observing time per year.

Analysis of Airglow Image Classification Based on Feature Map Visualization

All-sky airglow imagers (ASAIs) are used in the Meridian Project to observe the airglow in the middle and upper atmosphere to study the atmospheric perturbation. However, the ripples of airglow caused by the perturbation are only visible in the airglow images taken on a clear night. It is a problem to effectively select images suitable for scientific analysis from the enormous amount of airglow images captured under various environments due to the low efficiency and subjectivity of traditional manual classification. We trained a classification model based on convolutional neural network to distinguish between airglow images from clear nights and unclear nights. The data base contains 1688 images selected from the airglow images captured at Xinglong station (40.4° N, 30.5° E). The entire training process was tracked by feature maps which visualized every resulting classification model. The classification models with the clearest feature maps were saved for future use. We cropped the central part of the airglow images to avoid disturbance from the artificial lights at the edge of the vision field according to the feature maps of our first training. The accuracy of the saved model is 99%. The feature maps of five categories also indicate the reliability of the classification model.