Clouds and the Earth’s Radiant Energy System (CERES) FluxByCldTyp Edition 4 Data Product

The Clouds and the Earth’s Radiant Energy System (CERES) project has provided the climate community 20 years of globally observed top of the atmosphere (TOA) fluxes critical for climate and cloud feedback studies. The CERES Flux By Cloud Type (FBCT) product contains radiative fluxes by cloud-type, which can provide more stringent constraints when validating models and also reveal more insight into the interactions between clouds and climate. The FBCT product provides 1° regional daily and monthly shortwave (SW) and longwave (LW) cloud-type fluxes and cloud properties sorted by 7 pressure layers and 6 optical depth bins. Historically, cloud-type fluxes have been computed using radiative transfer models based on observed cloud properties. Instead of relying on radiative transfer models, the FBCT product utilizes Moderate Resolution Imaging Spectroradiometer (MODIS) radiances partitioned by cloud-type within a CERES footprint to estimate the cloud-type broadband fluxes. The MODIS multi-channel derived broadband fluxes were compared with the CERES observed footprint fluxes and were found to be within 1% and 2.5% for LW and SW, respectively, as well as being mostly free of cloud property dependencies. These biases are mitigated by constraining the cloud-type fluxes within each footprint with the CERES Single Scanner Footprint (SSF) observed flux. The FBCT all-sky and clear-sky monthly averaged fluxes were found to be consistent with the CERES SSF1deg product. Several examples of FBCT data are presented to highlight its utility for scientific applications.

Reliability Evaluation of the Joint Observation of Cloud Top Height by FY-4A and HIMAWARI-8

It is well known that the measurement of cloud top height (CTH) is important, and a geostationary satellite is an important measurement method. However, it is difficult for a single geostationary satellite to observe the global CTH, so joint observation by multiple satellites is imperative. We used both active and passive sensors to evaluate the reliability of joint observation of geostationary satellites, which includes consistency and accuracy. We analyzed the error of CTH of FY-4A and HIMAWARI-8 and the consistency between the two satellites and conducted research on the problem of missing measurement (Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) has CTH data, but FY-4A/HIMAWARI-8 does not) of the two satellites. The results show that FY-4A and HIMAWARI-8 have good consistency and can be jointly observed, but the measurement of CTH of FY-4A and HIMAWARI-8 has large errors, and the error of FY-4A is greater than that of HIMAWIRI-8. The error of CTH is affected by the CTH, cloud optical thickness (COT) and cloud type, and the consistency between the two satellites is mainly affected by the cloud type. FY-4A and HIMAWARI-8 have the problem of missing measurement. The missing rate of HIMAWARI-8 is greater than that of FY-4A, and the missing rate is not affected by the CTH, COT and surface type. Therefore, although FY-4A and HIMAWARI-8 have good consistency, the error of CTH and the problem of missing measurement still limit the reliability of their joint observation.

Cloud probability-based estimation of black-sky surface albedo from AVHRR data

Cloud cover constitutes a major challenge for the surface albedo estimation using Advanced Very High Resolution Radiometer AVHRR data for all possible conditions of cloud fraction and cloud type on any land cover type and solar zenith angle. Cloud masking has been the traditional way to estimate surface albedo from individual satellite images. Another approach to tackle cloudy conditions is presented in this study. Cloudy broadband albedo distributions were simulated first for theoretical cloud distributions and then using global cloud probability (CP) data of one month. A weighted mean approach based on the CP values was shown to produce very high accuracy black-sky surface albedo estimates for simulated data. The 90 % quantile for the error was 1.1 % (in absolute albedo percentage) and for the relative error it was 2.2 %. AVHRR based and in situ albedo distributions were in line with each other and also the monthly mean values were consistent. Comparison with binary cloud masking indicated that the developed method improves cloud contamination removal.

The impact of cloudiness and cloud type on the atmospheric heating rate of black and brown carbon in the Po Valley

We experimentally quantified the impact of cloud fraction and cloud type on the heating rate (HR) of black and brown carbon (HRBC and HRBrC). In particular, we examined in more detail the cloud effect on the HR detected in a previous study (Ferrero et al., 2018). High-time-resolution measurements of the aerosol absorption coefficient at multiple wavelengths were coupled with spectral measurements of the direct, diffuse and surface reflected irradiance and with lidar–ceilometer data during a field campaign in Milan, Po Valley (Italy). The experimental set-up allowed for a direct determination of the total HR (and its speciation: HRBC and HRBrC) in all-sky conditions (from clear-sky conditions to cloudy). The highest total HR values were found in the middle of winter (1.43 ± 0.05 K d−1), and the lowest were in spring (0.54 ± 0.02 K d−1). Overall, the HRBrC accounted for 13.7 ± 0.2 % of the total HR, with the BrC being characterized by an absorption Ångström exponent (AAE) of 3.49 ± 0.01. To investigate the role of clouds, sky conditions were classified in terms of cloudiness (fraction of the sky covered by clouds: oktas) and cloud type (stratus, St; cumulus, Cu; stratocumulus, Sc; altostratus, As; altocumulus, Ac; cirrus, Ci; and cirrocumulus–cirrostratus, Cc–Cs). During the campaign, clear-sky conditions were present 23 % of the time, with the remaining time (77 %) being characterized by cloudy conditions. The average cloudiness was 3.58 ± 0.04 oktas (highest in February at 4.56 ± 0.07 oktas and lowest in November at 2.91 ± 0.06 oktas). St clouds were mostly responsible for overcast conditions (7–8 oktas, frequency of 87 % and 96 %); Sc clouds dominated the intermediate cloudiness conditions (5–6 oktas, frequency of 47 % and 66 %); and the transition from Cc–Cs to Sc determined moderate cloudiness (3–4 oktas); finally, low cloudiness (1–2 oktas) was mostly dominated by Ci and Cu (frequency of 59 % and 40 %, respectively). HR measurements showed a constant decrease with increasing cloudiness of the atmosphere, enabling us to quantify for the first time the bias (in %) of the aerosol HR introduced by the simplified assumption of clear-sky conditions in radiative-transfer model calculations. Our results showed that the HR of light-absorbing aerosol was ∼ 20 %–30 % lower in low cloudiness (1–2 oktas) and up to 80 % lower in completely overcast conditions (i.e. 7–8 oktas) compared to clear-sky ones. This means that, in the simplified assumption of clear-sky conditions, the HR of light-absorbing aerosol can be largely overestimated (by 50 % in low cloudiness, 1–2 oktas, and up to 500 % in completely overcast conditions, 7–8 oktas). The impact of different cloud types on the HR was also investigated. Cirrus clouds were found to have a modest impact, decreasing the HRBC and HRBrC by −5 % at most. Cumulus clouds decreased the HRBC and HRBrC by −31 ± 12 % and −26 ± 7 %, respectively; cirrocumulus–cirrostratus clouds decreased the HRBC and HRBrC by −60 ± 8 % and −54 ± 4 %, which was comparable to the impact of altocumulus (−60 ± 6 % and −46 ± 4 %). A higher impact on the HRBC and HRBrC suppression was found for stratocumulus (−63 ± 6 % and −58 ± 4 %, respectively) and altostratus (−78 ± 5 % and −73 ± 4 %, respectively). The highest impact was associated with stratus, suppressing the HRBC and HRBrC by −85 ± 5 % and −83 ± 3 %, respectively. The presence of clouds caused a decrease of both the HRBC and HRBrC (normalized to the absorption coefficient of the respective species) of −11.8 ± 1.2 % and −12.6 ± 1.4 % per okta. This study highlights the need to take into account the role of both cloudiness and different cloud types when estimating the HR caused by both BC and BrC and in turn decrease the uncertainties associated with the quantification of their impact on the climate.

Applying self-supervised learning for semantic cloud segmentation of all-sky images

Semantic segmentation of ground-based all-sky images (ASIs) can provide high-resolution cloud coverage information of distinct cloud types, applicable for meteorology, climatology and solar energy-related applications. Since the shape and appearance of clouds is variable and there is high similarity between cloud types, a clear classification is difficult. Therefore, most state-of-the-art methods focus on the distinction between cloudy- and cloudfree-pixels, without taking into account the cloud type. On the other hand, cloud classification is typically determined separately on image-level, neglecting the cloud's position and only considering the prevailing cloud type. Deep neural networks have proven to be very effective and robust for segmentation tasks, however they require large training datasets to learn complex visual features. In this work, we present a self-supervised learning approach to exploit much more data than in purely supervised training and thus increase the model's performance. In the first step, we use about 300,000 ASIs in two different pretext tasks for pretraining. One of them pursues an image reconstruction approach. The other one is based on the DeepCluster model, an iterative procedure of clustering and classifying the neural network output. In the second step, our model is fine-tuned on a small labeled dataset of 770 ASIs, of which 616 are used for training and 154 for validation. For each of them, a ground truth mask was created that classifies each pixel into clear sky, low-layer, mid-layer or high-layer cloud. To analyze the effectiveness of self-supervised pretraining, we compare our approach to randomly initialized and pretrained ImageNet weights, using the same training and validation sets. Achieving 85.8 % pixel-accuracy on average, our best self-supervised model outperforms the conventional approaches of random (78.3 %) and pretrained ImageNet initialization (82.1 %). The benefits become even more evident when regarding precision, recall and intersection over union (IoU) on the respective cloud classes, where the improvement is between 5 and 20 % points. Furthermore, we compare the performance of our best model on binary segmentation with a clear-sky library (CSL) from the literature. Our model outperforms the CSL by over 7 % points, reaching a pixel-accuracy of 95 %.

Development of a random forest cloud regime classification model based on surface radiation and cloud products

Various methods have been developed to characterize cloud type, otherwise referred to as cloud regime. These include manual sky observations, combining radiative and cloud vertical properties observed from satellite, surface-based remote sensing, and digital processing of sky imagers. While each methodology has inherent advantages and disadvantages, none of these cloud typing methods actually include measurements of surface shortwave or longwave radiative fluxes. Here, a methodology that relies upon detailed, surface-based radiation and cloud measurements and derived data products to train a random forest machine learning cloud classification model is introduced. Measurements from five years of data from the ARM Southern Great Plains site were compiled to train and independently evaluate the model classification performance. A cloud type accuracy of approximately 80% using the random forest classifier reveals the model is well suited to predict climatological cloud properties. Furthermore, an analysis of the cloud type misclassifications is performed. While physical cloud types may be misreported, the shortwave radiative signatures are similar between misclassified cloud types. From this, we assert the cloud regime model has the capacity to successfully differentiate clouds with comparable cloud-radiative interactions. Therefore, we conclude the model can provide useful cloud property information for fundamental cloud studies, inform renewable energy studies, a tool for numerical model evaluation and parameterization improvement, among many other applications.

Improving cloud type classification of ground-based images using region covariance descriptors

The distribution and frequency of occurrence of different cloud types affect the energy balance of the Earth. Automatic cloud type classification of images continuously observed by the ground-based imagers could help climate researchers find the relationship between cloud type variations and climate change. However, by far it is still a huge challenge to design a powerful discriminative classifier for cloud categorization. To tackle this difficulty, in this paper, we present an improved method with region covariance descriptors (RCovDs) and the Riemannian bag-of-feature (BoF) method. RCovDs model the correlations among different dimensional features, which allows for a more discriminative representation. BoF is extended from Euclidean space to Riemannian manifold by k-means clustering, in which Stein divergence is adopted as a similarity metric. The histogram feature is extracted by encoding RCovDs of the cloud image blocks with a BoF-based codebook. The multiclass support vector machine (SVM) is utilized for the recognition of cloud types. The experiments on the ground-based cloud image datasets show that a very high prediction accuracy (more than 98 % on two datasets) can be obtained with a small number of training samples, which validate the proposed method and exhibit the competitive performance against state-of-the-art methods.