We present a new model based on a convolutional neural network (CNN) to predict daytime cloud cover (CC) from sky images captured by all‐sky cameras, which is called CNN‐CC. A total of 49,016 daytime sky images, recorded at different Spanish locations (Valladolid, La Palma, and Izaña) from two different all‐sky camera types, are manually classified into different CC (oktas) values by trained researchers. Subsequently, the images are randomly split into a training set and a test set to validate the model. The CC values predicted by the CNN‐CC model are compared with the observations made by trained people on the test set, which serve as reference. The predicted CC values closely match the reference values within 1 oktas in 99% of the cloud‐free and overcast cases. Moreover, this percentage is above 93% for the rest of partially cloudy cases. The mean bias error (MBE) and standard deviation (SD) of the differences between the predicted and reference CC values are calculated, resulting in oktas and oktas. The MBE and SD are also represented for different intervals of measured aerosol optical depth and Ångström exponent values, revealing that the performance of the CNN‐CC model does not depend on aerosol load or size. Once the model is validated, the CC obtained from a set of images captured every 5 min, from January 2018 to March 2022, at the Antarctic station of Marambio (Argentina) is compared against direct field observations of CC (not from images) taken at this location, which is not used in the training process. As a result, the model slightly underestimates the observations with an MBE of 0.3 oktas. The retrieved data are analyzed in detail. The monthly and annual CC values are calculated. Overcast conditions are the most frequent, accounting for 46.5% of all observations throughout the year, rising to 64.5% in January. The annual mean CC value at this location is 5.5 oktas, with a standard deviation of approximately 3.1 oktas. A similar analysis is conducted, separating data by hours, but no significant diurnal cycles are observed except for some isolated months.

Aerosol optical depth (AOD) at night-time has become a hot topic in recent years due to the development of new instruments recording accurate ground-based lunar irradiance measurements, and the development of calibration methods and extraterrestrial irradiance models adapted to lunar photometry. This study uses all daytime and night-time AOD data available at Valladolid (Spain) from October 2016 to March 2022 in order to analyze its behavior and the added contribution of night data. The annual, monthly and daily AOD evolution is studied comparing daytime and night-time values and checking the correlation between them. For this purpose, the daily averages are computed, showing an annual pattern, with low AOD values throughout the year (mean value of AOD at 440 nm: 0.122), where winter months have the lower and summer the higher values, as observed in previous studies. All these AOD values are modulated by frequent desert dust events over the Iberian Peninsula, with a strong influence on daily and monthly mean values in February and March, where the strongest desert outbreaks occurred. The added new data confirm these results and the good correlation between daytime and night-time data. Also, a complete daily evolution is shown, observing that AOD and Ångström exponent (AE) mean values vary by only ±0.02 in 24 h, with a maximum value at 06:00 UTC and minimum at 18:00 UTC for both parameters.

The lack of aerosol optical depth (AOD) data at night can be partially addressed through Moon photometer measurements or fully covered with star photometer observations. However, the limited availability and complexity of star photometers have motivated this study to use all-sky cameras to extract starlight signals and derive AOD at night using star photometry. For this purpose, eight all-sky cameras were configured and deployed in nine different locations to capture raw images with varying exposure times every 2 min during the night. This work proposes a novel methodology to extract the starlight signal from the raw data from all-sky cameras and convert it into AOD values. This process consists of the following steps: removing the background image, selecting the pixels, and extracting the signal for each star from a predefined list of 56 stars; performing in situ Langley calibration of the instruments and retrieving the total optical depth (TOD); calculating the effective wavelength for each camera channel; deriving the AOD by subtracting the gas contribution to TOD; and averaging, cloud-screening, and quality-assuring the AOD time series. The AOD time series obtained through this methodology are compared with independent AOD measurements from collocated Moon photometers in the nine locations. The obtained results show that the AOD values derived with the proposed method generally correlate with reference values, often achieving correlation coefficients (r) above 0.90. The AOD values retrieved using the cameras tend to overestimate the reference values by approximately 0.02 and exhibit a precision of around 0.03–0.04. The agreement between the two datasets varies with wavelength and decreases at high-latitude locations, likely due to the poorer performance of Langley calibration in these regions. AOD values align well with day-to-night transitions obtained by solar photometers, demonstrating their reliability. Despite the slight overestimation, the AOD values derived by this new method approximate the real values and provide coverage throughout the entire night, without requiring the presence of the Moon. Therefore, they serve to study and monitor the nocturnal evolution of AOD.

The present work proposes a new model based on a convolutional neural network (CNN) to retrieve solar shortwave (SW) irradiance via the estimation of the cloud modification factor (CMF) from daytime sky images captured by all-sky cameras; this model is named CNN-CMF. To this end, a total of 237,669 sky images paired with SW irradiance measurements obtained by using pyranometers were selected at the following three sites: Valladolid and Izaña, Spain, and Lindenberg, Germany. This dataset was randomly split into training and testing sets, with the latter excluded from the training model in order to validate it using the same locations. Subsequently, the test dataset was compared with the corresponding SW irradiance measurements obtained by the pyranometers in scatter density plots. The linear fit shows a high determination coefficient ($R^2$) of $0.99$. Statistical analyses based on the mean bias error (MBE) values and the standard deviation (SD) of the SW irradiance differences yield results close to $-2\%$ and $9\%$, respectively. The MBE indicates a slight underestimation of the CNN-CMF model compared to the measurement values. After its validation, model performance was evaluated at the Antarctic station of Marambio (Argentina), a location not used in the training process. A similar comparison between the model-predicted SW irradiance and pyranometer measurements yielded $R^2=0.95$, with an MBE of around $2\%$ and an SD of approximately $26\%$. Although the precision provided by the SD at the Marambio station is lower, the MBE shows that the model’s accuracy is similar to previous results but with a slight overestimation of the SW irradiance. Finally, the determination coefficient improved to $0.99$, and the MBE and SD are about $3\%$ and $11\%$, respectively, when the CNN-CMF model is used to estimate daily SW irradiation values.

This study explores the potential to retrieve aerosol properties with the GRASP algorithm (Generalized Retrieval of Atmosphere and Surface Properties) using as input measurements of zenith sky radiance (ZSR), which are sky radiance values measured in the zenith direction, recorded at four wavelengths by a ZEN-R52 radiometer. To this end, the ZSR measured at 440, 500, 675 and 870 nm by a ZEN-R52 (ZSR_ZEN), installed in Valladolid (Spain), is employed. This instrument is calibrated by intercomparing the signal of each channel with coincident ZSR values simulated (ZSR_SIM) at the same wavelengths with a radiative transfer model (RTM). These simulations are carried out using the GRASP forward module as RTM and the aerosol information from a co-located CE318 photometer belonging to AERONET (AErosol RObotic NETwork) as input. The dark signal and the signal dependence on temperature are characterized and included in the calibration process. The uncertainties for each channel are quantified by an intercomparison with a co-located CE318 photometer, obtaining lower values for shorter wavelengths; they are between 3 % for 440 nm and 21 % for 870 nm. The proposed inversion strategy for the aerosol retrieval using the ZSR_ZEN measurements as input, i.e. so-called GRASP-ZEN, assumes the aerosol as an external mixture of five pre-calculated aerosol types. A sensitivity analysis is conducted using synthetic ZSR_ZEN measurements, pointing out that these measurements are sensitive to aerosol load and type. It also assesses that the retrieved aerosol optical depth (AOD) values in general overestimate the reference ones by 0.03, 0.02, 0.02 and 0.01 for 440, 500, 675 and 870 nm, respectively. The calibrated ZSR_ZEN measurements, recorded during 2.5 years at Valladolid, are inverted by the GRASP-ZEN strategy to retrieve some aerosol properties like AOD. The retrieved AOD shows a high correlation with respect to independent values obtained from a co-located AERONET CE318 photometer, with determination coefficients (r²) of 0.86, 0.85, 0.79 and 0.72 for 440, 500, 675 and 870 nm, respectively, and finding uncertainties between 0.02 and 0.03 with respect to the AERONET values. Finally, the retrieval of other aerosol properties, like aerosol volume concentration for total, fine and coarse modes (VCT, VCF and VCC, respectively), is also explored. The comparison against independent values from AERONET presents r² values of 0.57, 0.56 and 0.66 and uncertainties of 0.009, 0.016 and 0.02 µm³ µm⁻² for VCT, VCF and VCC, respectively.