Abstract. Accurate identification of clear- and cloudy-sky conditions is essential for reliable aerosol and cloud retrievals from ground-based remote sensing observations. This study investigates the spectral characteristics of zenith radiances measured by SKYNET sky radiometers and evaluates their potential for distinguishing clear- and cloudy-sky conditions based on wavelength dependence. Analyses of normalized spectral zenith radiances at multiple sites representing diverse atmospheric environments, including urban, maritime, tropical continental, and polar regions, revealed systematic differences between clear- and cloudy-sky conditions identified using quality-controlled MODIS cloud mask products. The spectral slope derived from the logarithmic contrast between zenith radiances at 0.400 and 0.675 µm showed strong sensitivity to cloudy-sky conditions but limited capability for detecting clear-sky conditions. In contrast, slopes based on the 0.500 and 0.675 µm and 0.400 and 0.675 µm wavelength pairs exhibited more balanced detection performance for both clear- and cloudy-sky conditions. Based on detailed analyses of the dependence of these slope values on solar geometry, aerosol properties, and cloud properties, together with classification scores evaluated against MODIS cloud mask data, this study proposes a simple, efficient, and easy-to-use sky-state detection criterion that classifies observations into clear, cloudy, and undetermined categories. Independent validation using collocated direct normal irradiance measurements confirmed the physical consistency of the slope-based classification. Further comparison with the standard cloud-screening procedure showed very good agreement in clear-sky aerosol data extracted using the newly developed criterion across diverse atmospheric environments. Overall, these results demonstrate that spectral zenith radiances provide a simple and physically interpretable basis for clear- and cloudy-sky detection. Incorporating such spectral diagnostics into ground-based radiometric networks can improve cloud screening and enhance the reliability of long-term aerosol and cloud climatological analyses.

Accurate and continuous measurements of atmospheric carbon dioxide (CO2) are essential for climate change research and monitoring of emission reduction efforts. NASA's Orbiting Carbon Observatory (OCO-2 and 3) satellites have been deployed to infer the column-averaged CO2 dry-air mixing ratio (XCO2) from passive spectroscopy, with a designed uncertainty of less than 1 ppm for the regional average. This accuracy is often not met in cloudy regions because clouds in the vicinity of a footprint introduce biases in the XCO2 retrievals. These arise from limitations in the one-dimensional (1D) forward radiative transfer (RT) model, which does not capture the spectral radiance perturbations introduced by clouds adjacent to a clear footprint. Our paper introduces a three-dimensional (3D) RT pipeline to explicitly account for these effects in real-world satellite observations. This is done by ingesting collocated imagery and reanalysis products to calculate the cloud-induced perturbations at the footprint level. To make that computationally feasible, a simple approximation for their spectral dependence is used. The calculated perturbations are then used to reverse (undo) the cloud vicinity effects at the radiance level, at which point the standard 1D OCO-2 retrieval code can be applied without modifications. For two cases over land, we demonstrate that this approach indeed reduces the XCO2 anomalies near clouds. We also characterize the dependence of the XCO2 footprint-level bias on the distance from clouds and other key scene parameters, such as surface reflectance. Although this dependence may be specific to cloud type, aerosols, and other factors, we illustrate how it could be parameterized to bypass our physics-based 3D-RT pipeline for use in an operational framework. In the future, we intend to explore this possibility by applying our tool to a variety of scenes over land and ocean.

Nishinoshima and Mukojima lie within the subtropical gyre, a region known for low-nutrient and low-chlorophyll (LNLC) conditions. These LNLC waters typically limit phytoplankton growth due to a lack of essential nutrients. The brightness temperature observed by Himawari-8/9 was correlated with volcanic activity at Nishinoshima, and we focus on the period from June to July 2020, when volcanic activity was most active. Data from the AQUA satellite's Moderate Resolution Imaging Spectroradiometer (MODIS) and Himawari-8 daytime observations showed that the chlorophyll-a concentration (Chl-a) around Mukojima also peaked, with Seasonal and Trend decomposition using Loess (STL) analysis residuals increasing by 4.43% during the eruption period. Particle simulations of backward trajectories of seawater around Mukojima suggest that the increase in Chl-a observed near Mukojima on July 4 was attributable to volcanic ash released on June 28. These results indicate that volcanic ash spreading northeastward from Nishinoshima was carried by ocean currents to areas around Mukojima and within 100 km of Nishinoshima, resulting in increased Chl-a and phytoplankton growth over approximately six days.

We introduce the Education and Research 3D Radiative Transfer Toolbox (EaR3T, pronounced []) for quantifying and mitigating artifacts in atmospheric radiation science algorithms due to spatially inhomogeneous clouds and surfaces and show the benefits of automated, realistic radiance and irradiance generation along extended satellite orbits, flight tracks from entire aircraft field missions, and synthetic data generation from model data. EaR3T is a modularized Python package that provides high-level interfaces to automate the process of 3D radiative transfer (3D-RT) calculations. After introducing the package, we present initial findings from four applications, which are intended as blueprints to future in-depth scientific studies. The first two applications use EaR3T as a satellite radiance simulator for the NASA Orbiting Carbon Observatory 2 (OCO-2) and Moderate Resolution Imaging Spectroradiometer (MODIS) missions, which generate synthetic satellite observations with 3D-RT on the basis of cloud field properties from imagery-based retrievals and other input data. In the case of inhomogeneous cloud fields, we show that the synthetic radiances are often inconsistent with the original radiance measurements. This lack of radiance consistency points to biases in heritage imagery cloud retrievals due to sub-pixel resolution clouds and 3D-RT effects. They come to light because the simulator's 3D-RT engine replicates processes in nature that conventional 1D-RT retrievals do not capture. We argue that 3D radiance consistency (closure) can serve as a metric for assessing the performance of a cloud retrieval in presence of spatial cloud inhomogeneity even with limited independent validation data. The other two applications show how airborne measured irradiance data can be used to independently validate imagery-derived cloud products via radiative closure in irradiance. This is accomplished by simulating downwelling irradiance from geostationary cloud retrievals of Advanced Himawari Imager (AHI) along all the below-cloud aircraft flight tracks of the Cloud, Aerosol and Monsoon Processes Philippines Experiment (CAMP2Ex, NASA 2019) and comparing the irradiances with the colocated airborne measurements. In contrast to case studies in the past, EaR3T facilitates the use of observations from entire field campaigns for the statistical validation of satellite-derived irradiance. From the CAMP2Ex mission, we find a low bias of 10 % in the satellite-derived cloud transmittance, which we are able to attribute to a combination of the coarse resolution of the geostationary imager and 3D-RT biases. Finally, we apply a recently developed context-aware Convolutional Neural Network (CNN) cloud retrieval framework to high-resolution airborne imagery from CAMP2Ex and show that the retrieved cloud optical thickness fields lead to better 3D radiance consistency than the heritage independent pixel algorithm, opening the door to future mitigation of 3D-RT cloud retrieval biases.

The diurnal variations of clouds and their relationships with the rainfall associated with the Baiu–Meiyu front in Eastern China and over the ocean in the south of Japan in June 2016–2019 are investigated with the cloud product retrieved from the infrared brightness temperature of the Himawari-8 satellite. Clouds are classified into three types: low (<3 km), middle (3–10 km), and high clouds (>10 km); high clouds are further divided into thin (0.1–1), moderately-thick (1–6), and thick (>6) clouds based on their cloud optical thickness. Prominent diurnal variations of high clouds were observed to have an evening-to-midnight maximum in Eastern China, an early-afternoon-to-evening maximum around the Ryukyu Islands, and an evening maximum around the Izu Islands. In Eastern China and around the Ryukyu Islands, there was a clear diurnal amplitude maximum of the thick cloud and the moderately-thick cloud. And the diurnal peak time of the thick cloud was 1–2 h later than the peak time of rainfall. Around the Izu Islands, a diurnal amplitude maximum of moderately-thick cloud was found. While for the thick cloud, its peak occurrence was found to be 4–6 h later than that of the rainfall. Our results suggest that the convective activity is highly related to the primary mode of the diurnal variation of rainfall and cloud in Eastern China and around the Ryukyu Islands. Meanwhile, it is less related to the diurnal variation of rainfall and cloud around the Izu Islands.

Different diurnal rainfall cycles occur over the offshore and inland regions of the South China coastal area (SCCA). Inspired by these findings, in this study, we investigated the diurnal evolution features of cloud systems and cloud properties inside such systems for both the SCCA offshore and inland regions, using cloud data retrieved from a recently developed deep neural network model. Rainy day data for June 2017 revealed that the ice cloud optical thickness and top height reached their peak intensities at noon (~12 local standard time (LST)) over the offshore region, approximately 2 h later than the rainfall peak. Over the inland region, cloud and rainfall peaks simultaneously appeared from ~18 to 20 LST. When further examining the cloud-amount variation of different ice-cloud types, we found a clear diurnal oscillation in the medium-thick cloud amount over the offshore region, while for the inland region, this cloud type had no obvious diurnal peak, showing a low cloud amount throughout the day. This phenomenon suggests different inner structures and intensities between offshore and inland convections. To better elucidate the convection features over different regions, a tracking algorithm was applied to obtain various parameters, such as size, number, and duration of mesoscale convective systems. The strongest convections, which lasted over 12 h, tended to be abundant over the offshore region from ~03 to 12 LST, and an inland to offshore migration at ~03 LST was facilitated by the beneficial meteorological conditions observed at 113–116˚E, 20.5–22.5˚N.

Clouds constitute a key component of weather and climate systems, whereas the uniform retrieval of cloud properties, such as cloud top height (CTH) and cloud optical thickness (COT), requires accuracy and computational efficiency improvements. In this study, an image-based deep neural network (DNN) model for cloud identification and simultaneous retrieval of CTH and ice-COT is developed for Himawari-8 satellite infrared measurements. The DNN model is trained with brightness temperature data from four months in 2016 as the input, and cloud properties of an active remote sensing product from CloudSat and Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) as the target truth. Supplementary variables, including the vertical temperature profile, the surface elevation, and the geometrical parameters, are added as the input data. DNN model performance is first tested with an independent dataset, and then cases over a CloudSat track and a Himawari-8 granule (85?E-205?E, 60?S-60?N) are selected for further validation of the model by comparing its results with those from two physics-based models. For both the water-and ice-CTH estimates, the DNN model shows high consistency with the target values, with an overall CTH correlation coefficient of 0.90 for high ice clouds with COT >= 0.3. Notably, as an infrared method in nature, the DNN extends the predictable ice-COT to similar to 200, with relative biases of similar to 20% for high ice clouds with COT > 1. The strong accuracy of the DNN model is primarily derived from its ability to learn from the spatial features imprinted on the input brightness temperature image, and its integration of information from neighboring pixels in a three-dimensional space. A single full disk estimation with the DNN model takes about 20 min using one processor; therefore, near-real-time cloud property retrieval that is uniformly available over 24 h can be obtained for severe weather monitoring and mesoscale cloud-system studies.

We introduce a new machine learning approach to retrieve cloud optical thickness (COT) fields from visible passive imagery. In contrast to the heritage independent pixel approximation (IPA), our convolutional neural network (CNN) retrieval takes the spatial context of a pixel into account and thereby reduces artifacts arising from net horizontal photon transfer, which is commonly known as independent pixel (IP) bias. The CNN maps radiance fields acquired by imaging radiometers at a single wavelength channel to COT fields. It is trained with a low-complexity and therefore fast U-Net architecture with which the mapping is implemented as a segmentation problem with 36 COT classes. As a training data set, we use a single radiance channel (600 nm) generated from a 3D radiative transfer model using large eddy simulations (LESs) from the Sulu Sea. We study the CNN model under various conditions based on different permutations of cloud aspect ratio and morphology, and we use appropriate cloud morphology metrics to measure the performance of the retrievals. Additionally, we test the general applicability of the CNN on a new geographic location with LES data from the equatorial Atlantic. Results indicate that the CNN is broadly successful in overcoming the IP bias and outperforms IPA retrievals across all morphologies. Over the Atlantic, the CNN tends to overestimate the COT but shows promise in regions with high cloud fractions and high optical thicknesses, despite being outside the general training envelope. This work is intended to be used as a baseline for future implementations of the CNN that can enable generalization to different regions, scales, wavelengths, and sun-sensor geometries with limited training.

Advances in Internet of Things (IoT) and cloud/edge computing systems could precisely monitor the meteorological elements and environmental conditions. Remote automated observation system (RAOS) makes the full use of IoT to communicate with other sensors, enabling the active responses from passive devices for smart weather. Cloud observation and classification have been regarded as a successful application that could automatically perform emergency tasks in RAOS. However, with the increasing growth of resource exploitation, the performance of communications among the automatic observation platforms, and the efficiency of task allocation among them has become a critical challenge. In this article, an ensemble learning method and resource allocation scheme are proposed to realize the cloud observation and classification with the help of reliable and controllable infrastructures. On the one hand, several ensemble methods, like Bagging, AdaBoost, and Snapshot are selected as a base classifier to capture the cross-semantic and structure features of cloud, while applying them to the ensemble using convolutional neural networks with different base learners and residual neural networks with different depths. on the other hand, a particular cloud-edge distributed framework is proposed for cloud classification approach based on the intelligent network, to overcome the difficulty in the massive data transmission. The experimental results verify that the proposed ensemble approach achieves high accuracy of cloud classification, and effectively improves the number of allocated tasks. Ensemble methods can generate a more accurate prediction than any single classifier or the majority algorithms. It consistently yields lower error rates than single state-of-the-art models at no additional training cost.

Extensive observational and numerical investigations have been performed to better characterize cloud properties. However, due to the large variations in cloud spatiotemporal distributions and physical properties, quantitative depictions of clouds in different atmospheric reanalysis datasets are still highly uncertain. A radiance-based evaluation approach is introduced and performed to evaluate the quality of cloud properties from reanalysis datasets. The China Meteorological Administration reanalysis (CRA); the ECMWF fifth-generation reanalysis (ERA5); and the Modern-Era Retrospective analysis for Applications, Version 2 (MERRA-2), i.e., those reanalyses providing sufficient cloud information, are considered. To avoid the influence of assumptions and uncertainties on satellite retrieval algorithms, forward radiative transfer simulations are used as a bridge to translate the reanalyses to corresponding radiances that are expected to be observed by satellites. The simulated reflectances and brightness temperatures (BTs) are directly compared with observations from the Advanced Himawari Imager onboard the Himawari-8 satellite in the East Asia region. We find that the simulated reflectances and BTs based on CRA and ERA5 are close to each other. CRA represents the total and midlayer cloud cover better than the other two datasets, and ERA5 depicts deep-convection structures more closely than CRA does. Comparisons of the simulated and observed BT differences suggest that water clouds are generally overestimated in ERA5 and MERRA-2, and MERRA-2 also overestimates the ice clouds over cyclone centers. Overall, clouds from CRA, ERA5, and MERRA-2 show their own advantages in different aspects. The ERA5 reanalysis has the best capability to represent the cloudy atmospheres over East Asia, and the CRA representations are close to those in ERA5.

An efficient radiative transfer model (ERTM) is developed to simulate thermal infrared brightness temperatures observed by the Advanced Himawari Imager (AHI) in this study. The ERTM contains an alternate mapping correlated k-distribution (AMCKD) scheme, a parameterization for cloud optical property, and a rapid infrared radiative transfer scheme. The AMCKD is employed to calculate the gaseous absorption in the inhomogeneous thermodynamic atmosphere. The optical properties of clouds are parameterized by the effective length for ice clouds based on the Voronoi model, and by the effective radius for water clouds based on the Lorenz-Mie theory. The adding method of four-stream discrete ordinates method (4DDA) is extended to be able to calculate the thermal infrared radiative intensity varying with the zenith angle in ERTM. The efficiency and accuracy of ERTM are evaluated by comparing with the benchmark model which is composed of discrete ordinate radiative transfer (DISORT) and line-by-line radiative transfer model (LBLRTM). Under the standard atmospheric profiles, the root mean square error (RMSE) of simulated brightness temperatures reaches a maximum of 0.21K at the B16 (13.28 µm) channel of AHI. The computational efficiency of ERTM is approximately five orders of magnitude higher than that of the benchmark model. Moreover, the simulated brightness temperatures by ERTM are highly consistent with the rigorous results and AHI observations in the application to the Typhoon Mujigae case.

Aerosol effects on deep convective cloud (DCC) have been recognized as one of the complex subjects in climatic studies because of the difficulty in quantifying the sole effect of aerosols on DCC. The complexity further arises if the atmosphere has very strong temporal and spatial variations such as that of Indo-Pacific Warm Pool (IPWP) region. Considering the strong influence of IPWP region on global climate change and water circulation, we investigated aerosol effects on DCC over this region by using data of 2015-2016 El Nino and the 2017-2018 La Nina events. We developed a spectral analysis based framework to identify and decouple the influences of major external factors on aerosol-DCC relationship. We found that temporal variations of aerosols, clouds, and meteorology longer than 2 days' time scale can have larger influences than their diurnal and spatial variations on aerosol-DCC relationship. By removing the effects of those spatial and temporal variations of different scales, the study suggests that aerosols of IPWP region can affect DCC properties with time lags less than similar to 5 hours and by increasing cloud-top height, cloud coverage, and DCC number concentration with the increase of aerosols.

Deep convective systems (DCSs) are associated with severe weather events and can affect regional and global climate. To study the semi-diurnal variation of DCSs over Eastern China and its surrounding seas in summer, we modified the Tracking of Organized Convection Algorithm through a 3-D segmentatioN (TOOCAN) by employing Himawari-8 operational cloud property (CLP) products instead of original infrared images, and renamed the algorithm as TOOCAN-CLP. The DCSs detected over land and sea are divided into small-, medium-, and large-sized classes based on the convective core equivalent radius. The small and medium-sized DCSs over land exhibit a maximum occurrence in the afternoon, which is associated with local thermal instability and sea breeze circulation. The occurrence of small DCSs over the tropical sea areas varies analogously to that of small continental DCSs but with a smaller amplitude. However, medium-sized DCSs over the sea, which account for the majority of DCSs over the sea, exhibit weak semi-diurnal variability. Large DCSs over inland China and its surrounding seas tend to initiate at night and decay in the daytime. The generation of large DCSs over inland China at night is mainly due to the enhanced transport of warm and moist air by strong large-scale prevailing southerly or southwesterly winds, while the large offshore DCSs accompanied by heavy rainfall is closely associated with the interaction between local offshore breeze and large-scale monsoon flows, as well as gravity waves.

The impact of radiative transfer scheme on global climate model (GCM) simulation is presented in this paper by comparing the difference between delta-two-stream adding method (delta-2DDA) and adding algorithm of the delta-four-stream discrete ordinates method (delta-4DDA) radiation schemes in the Atmospheric General Circulation Model of the Beijing Climate Center (BCC_AGCM). Only consider the effects of the calculation method itself, the delta-4DDA reduces the negative shortwave cloud radiative effect (CRE) in the areas with a significant fraction of low cloud, while enhances the negative shortwave CRE in the areas with the large fraction of high cloud. For the longwave CRE, the delta-4DDA enhances the longwave CRE drastically in the regions with a significant fraction of the high cloud. The feedback of clouds results in more interesting results. The delta-4DDA produces more accurate shortwave CRE in the region over the land and ocean in the middle and high latitude areas. The longwave CRE simulated by delta-4DDA is better than that affected by delta-2DDA over the ground in Africa, South America, and Atlantic. The change of radiation scheme affects the simulation of other meteorological variables. The simulation of global humidity by delta-4DDA is improved obviously. The delta-4DDA simulates more accurate temperature in continents of the northern hemisphere and precipitation in North America, Africa, northern Indian Ocean and western Pacific. Although the improvement of every physical process is required to develop the models, implementing delta-4DDA scheme into GCM and evaluating the effect of it are necessary and meaningful. (C) 2019 Published by Elsevier Ltd.

An optimal estimation algorithm to retrieve the cloud optical depth (COD) and cloud particle effective radius (CER) from spectral zenith radiances observed by narrow field-of-view (FOV) ground-based sky radiometers was developed. To further address the filter degradation problem while analyzing long-term observation data, an on-site calibration procedure is proposed, which has good accuracy compared with the standard calibration transfer method. An error evaluation study conducted by assuming errors in observed transmittances and ancillary data for water vapor concentration and surface albedo suggests that the errors in input data affect retrieved CER more than COD. Except for some narrow domains that fall within a COD of < 15, the retrieval errors are small for both COD and CER. The retrieved cloud properties reproduce the broadband radiances observed by a narrow FOV radiometer more precisely than broadband irradiances observed by a wide-FOV pyranometer, justifying the quality of the retrieved product (at least of COD) and indicating the important effect of the instrument FOV in cloud remote sensing. Furthermore, CODs (CERs) from sky radiometer and satellite observations show good (poor) agreement.

We have developed a rapid simplified algorithm to retrieve cloud optical thickness and cloud-top height from measurements of the infrared split-window bands of Himawari-8. The method is based on a rapid calculation model for clear-sky brightness temperatures and empirical equations for cloud, for which the coefficients are determined by a fit to a more rigorous radiative-transfer model. This method can be applied regardless of regions excluding the polar regions and season by taking into account the temperature, humidity, sea surface temperature, and surface emissivity. In this study, we have demonstrated that this method captures well the diurnal cycle of cloud amounts of different cloud types in the warm-pool region around Indonesia. With an accelerated retrieval process by a factor of around 1,000 compared with the the physics-based retrieval, our rapid cloud retrieval algorithm yielded cloud amounts that agree quantitatively with those from a more rigorous, physics-based cloud retrieval method.

Abstract The responses of cloud and diagnostic fields to the poleward shift of the local Hadley cell subsiding edge are examined using observational and reanalysis data from December 1982 to February 2016. Over the western North Pacific, the interannual variability of the local Hadley cell subsiding edge is marked by an anomalous rising motion, decreased large‐scale static stability, and corresponding increases of the midlevel cloud fraction over the climatological sinking zone around 35°N. The most sensitive cloud type is identified to be cumulus congestus, with cloud top pressures in the range of 440–680 hPa and optical thicknesses in the range of 23–60. This kind of cloud is estimated to be the main contributor to the net negative top‐of‐atmosphere radiation anomaly and to constrain the distribution of the anomalous precipitation associated with the local HC expansion.

Observation of the spatial distribution of cloud optical thickness (COT) is useful for the prediction and diagnosis of photovoltaic power generation. However, there is not a one-to-one relationship between transmitted radiance and COT (so-called COT ambiguity), and it is difficult to estimate COT because of three-dimensional (3D) radiative transfer effects. We propose a method to train a convolutional neural network (CNN) based on a 3D radiative transfer model, which enables the quick estimation of the slant-column COT (SCOT) distribution from the image of a ground-mounted radiometrically calibrated digital camera. The CNN retrieves the SCOT spatial distribution using spectral features and spatial contexts. An evaluation of the method using synthetic data shows a high accuracy with a mean absolute percentage error of 18% in the SCOT range of 1–100, greatly reducing the influence of the 3D radiative effect. As an initial analysis result, COT is estimated from a sky image taken by a digital camera, and a high correlation is shown with the effective COT estimated using a pyranometer. The discrepancy between the two is reasonable, considering the difference in the size of the field of view, the space–time averaging method, and the 3D radiative effect.