Using Convolutional Neural Networks for Cloud Detection on VENμS Images over Multiple Land-Cover Types

Moderate resolution imaging spectrometer (MODIS) images are widely used in land, ocean, and atmospheric monitoring, due to their wide spectral coverage, high temporal resolution, and convenient data acquisition. Accurate cloud detection is critical to the fine processing and application of MODIS images. Owing to spatial resolution limitations and the influence of mixed pixels, most MODIS cloud detection algorithms struggle to effectively recognize of clouds and ground objects. Here, we propose a novel cloud detection method based on land surface reflectance and a multiscale feature convolutional neural network to achieve high-precision cloud detection, particularly for thin clouds and clouds over bright surface. A monthly surface reflectance dataset was constructed by MODIS products (MOD09A1) and employed to provide background information for cloud detection. Difference-based samples were obtained using surface reflectance as well MODIS images of different phases based on difference operations. The multiscale feature network (MFCD-Net) using an atrous spatial pyramid pooling and a channel and spatial attention module integrated low-level spatial features and high-level semantic information to capture multiscale features and generate a high-precision cloud mask. For cloud detection experiments and quantitative analysis, 61 MODIS images acquired at different times on various underlying surface types were used. Cloud detection results were compared to those of UNet, Deeplabv3+, UNet++, PSPNet, and top of atmosphere-based (MFCD-TOA) methods. The proposed method performed well, with the highest overall accuracy (96.55%), precision (92.13%), and recall (88.90%). It improved cloud detection accuracy in various scenarios, reducing thin cloud omission and bright surface misidentification.

Accurate cloud detection is critical for advancing atmospheric monitoring and meteorological forecasting. This paper presents the Cloud Detection Challenge, an initiative aimed at enhancing cloud detection through innovative solutions using lidar-based ceilometer data. This initiative was hosted by IEEE MetroXRAINE 2024, and 11 teams participated in this initiative. Participants were provided with a novel dataset of backscatter profiles converted into time-height plots, offering unique insights into atmospheric conditions beyond conventional imagery. Data collection employed a Lufft CHM 15k ceilometer, capturing cloud dynamics every 15 seconds located near Mt. Etna, an active volcano in Italy. The dataset includes 1568 hourly labeled backscatter profiles, serving as a benchmark for state-of-the-art deep learning models. The challenge sets a baseline performance of 89.57% accuracy, 92.73% F1-score, 89.82% precision, and 95.84% recall, inviting participants to develop models to exceed these results. Submissions proposed a wide-range of AI-based approaches, including Transformer and Convolutional Neural Network architectures, showcasing the potential of advanced image analysis techniques in lidar-based cloud detection. This paper details the challenge framework, as well as the methodologies proposed by top-performing teams, offering a comparative evaluation of their effectiveness. Our initiative advances cloud detection technologies and underscores their broader implications for environmental monitoring, agriculture, and satellite imaging. The insights and dataset presented herein lay the groundwork for future advancements in leveraging lidar data for atmospheric analysis.

High-precision cloud detection for remote sensing image is of great significance for applications in agriculture, environment, meteorology and other fields. Cloud detection in bright surface environments and thin clouds identification have always been a challenge in cloud detection research. Aiming at this problem, a cloud detection method for remote sensing images based on difference features and semantic segmentation network is proposed in this paper. Cloudy and cloudless images of the same area are used to obtain difference features, and cloudless images are used as the surface reference. The multi-band cloud detection network based on U-network (U-net) fully learns the difference feature between clouds and the surface, and combines the feature information of the shallow and deep layers to accurately identify the cloud and the surface. Experiments were conducted on the Landsat 8 validation dataset. The results show that the proposed method achieves good performance. It improves the detection accuracy and reduces the misclassification over the bright surface underlying compared with the method based on top of atmosphere reflectance (TOA).

An in-depth analysis of Convolutional Neural Network architectures with transfer learning for skin disease diagnosis

Land use assessment and land cover transition need remote sensing (RS) and geographic information systems (GIS). Land use/land cover changes of Ado-Ekiti Local Government Area, Ekiti State, Nigeria, were examined in this research. Landsat 5 TM, Landsat 7 ETM+ and Landsat 8 OLI were acquired for 1985, 2000, and 2015 respectively. Image scene with path 190 and row 055 was used for the three Landsat Images. A supervised digital image classification approach was used in the study, which was carried out using the ArcMap 10.4 Software. Five land use/land cover categories were recognised and recorded as polygons, including Built-up Areas, Bare surface, water body, Dense Vegetation and Sparse Vegetation. The variations in the area covered by the various polygons were measured in hectares. This study revealed that between 1985 and 2015, there was a significant change in Built-up areas from 1694 hectares to 5656 hectares. However, there was a reduction in water body from 25 hectares in 1985 to 19 hectares in 2015; there was a severe reduction in the bare surface from 4641 hectares in 1985 to 2237 hectares in 2015. Generally, the findings show that the number of people building houses in the study area has grown over time, as many people reside in the outskirts of the Local Government Area, resulting in a decrease in the vegetation and bare surfaces. The maps created in this research will be useful to the Ekiti State Ministry of Land, Housing, Physical Planning, and Urban Development to develop strategies and government policies to benefit people living in the Ado-Ekiti Local Government Area of the State.

PROBA-V (PRoject for On-Board Autonomy–Vegetation) was launched in May-2013 as an operational continuation to the vegetation (VGT) instruments on-board the Système Pour l’Observation de la Terre (SPOT)-4 and -5 satellites. The first reprocessing campaign of the PROBA-V archive from Collection 0 (C0) to Collection 1 (C1) aims at harmonizing the time series, thanks to improved radiometric and geometric calibration and cloud detection. The evaluation of PROBA-V C1 focuses on (i) qualitative and quantitative assessment of the new cloud detection scheme; (ii) quantification of the effect of the reprocessing by comparing C1 to C0; and (iii) evaluation of the spatio-temporal stability of the combined SPOT/VGT and PROBA-V archive through comparison to METOP/advanced very high resolution radiometer (AVHRR). The PROBA-V C1 cloud detection algorithm yields an overall accuracy of 89.0%. Clouds are detected with very few omission errors, but there is an overdetection of clouds over bright surfaces. Stepwise updates to the visible and near infrared (VNIR) absolute calibration in C0 and the application of degradation models to the SWIR calibration in C1 result in sudden changes between C0 and C1 Blue, Red, and NIR TOC reflectance in the first year, and more gradual differences for short-wave infrared (SWIR). Other changes result in some bias between C0 and C1, although the root mean squared difference (RMSD) remains well below 1% for top-of-canopy (TOC) reflectance and below 0.02 for the normalized difference vegetation index (NDVI). Comparison to METOP/AVHRR shows that the recent reprocessing campaigns on SPOT/VGT and PROBA-V have resulted in a more stable combined time series.

A fast method for the Apparent Pressure Retrieval (APR method) over land from satellite data, based on a two band ratio in the oxygen A‐band (759–770 nm), is described. This method is devoted to the cloud detection and atmospheric corrections. Parameterizations are performed from line‐by‐line calculations assuming a pure absorbing medium. Moreover, we defined a corrective factor to account for scattering effects of the atmosphere. We validated this method with measurements of the MOS sensor (Modular Optoelectronic Scanner), whose spectral characteristics are appropriate. Comparisons with ECMWF (European Centre for Medium‐Range Weather Forecasts) pressures showed the need to perform in‐flight calibrations over a reference scene to account for spectral shifts of filter responses. Therefore we selected bright surfaces for the calibration, such as deserts, because of their major contribution to the satellite signal. After calibration the accuracy of the method is about 10 hPa over bright surfaces. Comparisons for various meteorological and geographical conditions showed that deviations between ECMWF pressures and MOS apparent pressures are generally less than 30 hPa using scattering corrections. These deviations are multiplied by 2 without correction. The APR method has been included in the cloud detection and atmospheric correction algorithms for the MOS data processing over land. Theoretical studies showed that the APR method is suitable for the cloud discrimination and that an error of 30 hPa on the surface pressure retrieval has no noticeable effect on the geophysical products of these algorithms, such as aerosol optical thickness or surface reflectance. Consequently, this method is potentially applicable to the retrieval of the apparent pressure over land with the MOS algorithms, as well as other similar satellite sensors such as the Medium Resolution Imaging Spectrometer.

Sea ice modulates the transfer of shortwave radiative energy fluxes within the Arctic atmosphere-sea-ice-ocean system. Understanding and predicting these fluxes comes with greatest uncertainties during the melt and freeze-up seasons, when the sea ice surface is strongly heterogeneous and changing rapidly. Then, the partitioning of solar radiative fluxes between atmosphere, ice, and ocean has greatest impacts on the surface energy budget, controlling sea ice melt and formation. Here, we investigated changes and impacts of sea ice surface variability by analyzing high-resolution red-green-blue aerial imagery obtained during the Multidisciplinary Observatory for the Study of Arctic Climate (MOSAiC) expedition in 2020. We used pixel brightness from processed aerial images as a proxy of surface albedo, because such data are frequently available and obtainable from commercial digital cameras. The results allowed quantification of fluxes on floe-scales and also revealed the seasonality of sea ice spatial heterogeneity, which was strongest in the middle of melt season driven by melt pond processes. On scales of 10 m × 10 m, a magnitude larger than the traditional single in-situ optical observations (although many are made over larger scales), distinct surface conditions, for example, individual melt ponds, resulted in differences of energy deposition into the ice by more than 600%. The effects of spatial variability were minimized by integrating over areas 200 m × 200 m and larger. We suggest considering these scales for future energy budget studies and airborne observations, because sufficient parts of different surface features are included. The concept of surface brightness and aerial photographs might help to bridge in-situ observations to even larger scales, including fractions of open water. It may also be used to upscale observations of under-ice light regimes by providing spatially continuous surface brightness that governs the light transmittance, thus to improve our understanding of the coupled system, including ecological functions.

Satellite remote sensing has been providing aerosol data with ever-increasing accuracy, representative of the MODerate-resolution Imaging Spectroradiometer (MODIS) Dark Target (DT) and Deep Blue (DB) aerosol retrievals. These retrievals are generally performed over spectrally dark objects and therefore may struggle over bright surfaces. This study proposed an analytical TERRA-AQUA MODIS relationship in the reflective solar bands for aerosol retrieval. For the relationship development, the bidirectional reflectance distribution function (BRDF) effects were adjusted using reflectance ratios in the MODIS 2.13 μm band and the path radiance was approximated as an analytical function of aerosol optical thickness (AOT) and scattering phase function. Comparisons with MODIS observation data, MODIS AOT data, and sun photometer measurements demonstrate the validity of the proposed relationship for aerosol retrieval. The synergetic TERRA-AQUA MODIS retrievals are highly correlated with the ground measured AOT at TERRA MODIS overpass time (R2 = 0.617; RMSE = 0.043) and AQUA overpass time (R2 = 0.737; RMSE = 0.036). Compared to our retrievals, both the MODIS DT and DB retrievals are subject to severe underestimation. Sensitivity analyses reveal that the proposed method may perform better over non-vegetated than vegetated surfaces, which can offer a complement to MODIS operational algorithms. In an analytical form, the proposed method also has advantages in computational efficiency, and therefore can be employed for fine-scale (relative to operational 10 km MODIS product) MODIS aerosol retrieval. Overall, this study provides insight into aerosol retrievals and other applications regarding TERRA-AQUA MODIS data.

Clouds are an important component of weather and climate change and have a significant impact on ecology, weather forecasting and aviation safety. Cloud detection is a preprocessing step in many satellite image processing and remote sensing inversion. Clouds not only cover the subsurface, but also absorb and scatter solar radiation, which affects the remote sensing analysis and study of the surface and reduces the utilization of data. Accurate detection of cloud-covered areas in satellite images can effectively reduce the interference of cloud cover on remote sensing data application. The Fengyun-4B Satellite is the latest generation of geostationary meteorological satellites in China, with significant performance advantages. In this paper, a land cloud detection algorithm for AGRI sensors is constructed based on the existing cloud detection algorithm using the AGRI full disk data of Fengyun 4B Satellite. Through the analysis of cloud detection product quality, the algorithm threshold is optimized to significantly improve the accuracy of cloud detection. The experimental results show that the algorithm has good overall cloud detection performance in full disk imagery, especially in the thick cloud region, but there are some misclassifications in the high brightness surface region such as the Tibetan Plateau. This study expands the application scenarios of domestic geostationary meteorological satellites, and provides new technical support for meteorological monitoring and remote sensing applications.

Since the launch of the first Advanced Very High Resolution Radiometer (AVHRR) instrument aboard the Television and Infrared Observational Satellite (TIROS-N), measurements in the 3.7-μm atmospheric window have been exploited for use in cloud detection and screening, cloud thermodynamic phase and surface snow/ice discrimination, and quantitative cloud particle size retrievals. The utility of the band has led to the incorporation of similar channels on a number of existing satellite imagers and future operational imagers. Daytime observations in the band include both reflected solar and thermal emission energy. Since 3.7-μm channels are calibrated to a radiance scale (via onboard blackbodies), knowledge of the top-of-atmosphere solar irradiance in the spectral region is required to infer reflectance. Despite the ubiquity of 3.7-μm channels, absolute solar spectral irradiance data come from either a single measurement campaign (Thekaekara et al.) or synthetic spectra. In the current study, the historical 3.7-μm band spectral irradiance datasets are compared with the recent semiempirical solar model of the quiet sun by Fontenla et al. The model has expected uncertainties of about 2% in the 3.7-μm spectral region. The channel-averaged spectral irradiances using the observations reported by Thekaekara et al. are found to be 3.2%–4.1% greater than those derived from the Fontenla et al. model for Moderate Resolution Imaging Spectroradiometer (MODIS) and AVHRR instrument bandpasses; the Kurucz spectrum, as included in the Moderate Spectral Resolution Atmospheric Transmittance (MODTRAN4) distribution, gives channel-averaged irradiances 1.2%–1.5% smaller than the Fontenla model. For the MODIS instrument, these solar irradiance uncertainties result in cloud microphysical retrieval uncertainties that are comparable to other fundamental reflectance error sources.

Best management practices (BMPs) are the tools which are adopted to run any activity while inflicting least negative impacts on the surrounding environment. Since, water pollution is one of the most important issues in today's world; it is indispensable to discuss the activities which are deteriorating its quality. Although, a number of treatment technologies and methods are in place for treating the point sources of pollution, there is a need to address nonpoint sources of pollution. There are many successful practices which are capable of curtailing the impacts of nonpoint source pollution, if adopted. In this paper an attempt has been made to evaluate the performance and trade‐offs of various management practices to address the issues of agricultural nonpoint source pollution. Nevertheless, there is a dearth of guiding principles which can pave a way for adoption of these practices. Therefore, an attempt has also been made to explore the aspects of policy formulations along with emphasizing the factors which can lead to adoption/ non‐adoption of such practices.

In this research work, the separation of built-up areas from bare lands in El Khroub city is carried out using a supervised classification approach involving several indices and combining spectral bands of the Sentinel-2A images sensor. The multi-index approach is based on the combination of seven indices in order to discriminate between the three main categories of land cover, which are water bodies, green areas and buildings. 3First, this operation requires the use of NDVI, BAEI, NDBI, NDTI, BUI, MNDWI and the NDVIre index, which have a strong discrimination capacity between build-up area and the other land cover features. The neo-images obtained from the combination of the above indices are then classified with the Likelihood algorithm for the extraction of the six class types of land cover (built-up areas, bare land, vegetation, forest, water bodies and asphalt). The multi-index obtained from the combination of BUI, NDTI and NDVIre is the most effective; shown by the evaluation values, where the Overall accuracy is of 96.44%, the Kappa Coefficient (K) of 95.72% and a User Accuracy for built-up class of the order of 100%, with a zero rate of commission. Therefore, the multi-index (BUI, NDTI and NDVIre) is retained for build-up area extraction due to its best discrimination capability.

Natural waters contain particulate matter that increases turbidity, imparts apparent color, and interferes with light penetration. These particles originate from erosion on watersheds and within water bodies, vegetative debris from watersheds, and microorganisms produced in water bodies. Suspended particles vary from tiny colloids (1–100 mμ) that remain suspended indefinitely in still water to larger silt and sand particles held in suspension by turbulence. The settling velocity of particles in still water is estimated by the Stoke’s law equation and it depends mainly on particle diameter and density—large, dense particles settle the fastest. Organic particles settle slowly because of their low density, but planktonic organisms also have adaptions that lessen settling velocity. Solar radiation of all wavelengths is quenched quickly—about 50 % is reflected or converted to heat within the first meter. Within the visible spectrum, red and orange light is absorbed most strongly followed by violet, and by yellow, green, and blue. Plants absorb mostly red and orange light and yellow and green light for use in photosynthesis. Turbidity absorbs all wavelengths of light and diminishes photosynthesis. Two common ways of assessing turbidity in natural waters are by measuring the depth of disappearance of a 20-cm diameter disk (the Secchi disk) and by using nephelometry to determine the amount of light reflected at a 90 ° angle from a water sample. The weight of particles retained on a fine filter allows gravimetric assessment of suspended solids.

We perform correlation analysis between air temperature and surface brightness temperature over the Adelaide Central Business District and the surrounding parklands. The results indicate that three effective surface areas associated with three different mechanisms exist. They are the effective surface area with upward sensible heat transfer to heat up and maintain the air temperature, the effective surface area with downward sensible heat transfer to cool down the air, and the effective surface area related to very localized conditions (e.g., sky-view factors). The three mechanisms occur at different times of the day and result in different air temperature and surface temperature correlations. The first effective surface area exists in the daytime after the surface is heated up by solar radiation and can persist into the night, particularly in urban environments. The second effective area occurs in night-time when the surface has been cooled down sufficiently. The third effective area coexists with the second one and has a weaker correlation between the surface and air temperatures. It was also found that in an urban area, exclusion of roof pixels in the calculation of surface brightness temperature can increase the correlation between the surface and air temperatures in the middle of the night.