scholarly journals Retrieval of Cloud Optical Thickness from Sky-View Camera Images using a Deep Convolutional Neural Network based on Three-Dimensional Radiative Transfer

2019 ◽  
Vol 11 (17) ◽  
pp. 1962 ◽  
Author(s):  
Ryosuke Masuda ◽  
Hironobu Iwabuchi ◽  
Konrad Sebastian Schmidt ◽  
Alessandro Damiani ◽  
Rei Kudo
Keyword(s):  
Neural Network ◽  
Spatial Distribution ◽  
Radiative Transfer ◽  
Convolutional Neural Network ◽  
Optical Thickness ◽  
Three Dimensional ◽  
Digital Camera ◽  
Transfer Model ◽  
Radiative Effect ◽  
Cloud Optical Thickness

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.

scholarly journals Retrieval of optical thickness and droplet effective radius of inhomogeneous clouds using deep learning

2017 ◽  
Author(s):  
Rintaro Okamura ◽  
Hironobu Iwabuchi ◽  
K. Sebastian Schmidt
Keyword(s):  
Deep Learning ◽  
Radiative Transfer ◽  
Optical Thickness ◽  
Three Dimensional ◽  
Cloud Droplet ◽  
Effective Radius ◽  
Radiative Transfer Model ◽  
Transfer Model ◽  
Transfer Effects ◽  
Cloud Properties

Abstract. Three-dimensional (3D) radiative transfer effects are a major source of retrieval errors in satellite-based optical re- mote sensing of clouds. In this study, we present two retrieval methods based on deep learning. We use deep neural networks (DNNs) to retrieve multipixel estimates of cloud optical thickness and column-mean cloud droplet effective radius simultane- ously from multispectral, multipixel radiances. Cloud field data are obtained from large-eddy simulations, and a 3D radiative transfer model is employed to simulate upward radiances from clouds. The cloud and radiance data are used to train and test the DNNs. The proposed DNN-based retrieval is shown to be more accurate than the existing look-up table approach that assumes plane-parallel, homogeneous clouds. By using convolutional layers, the DNN method estimates cloud properties robustly, even for optically thick clouds, and can correct the 3D radiative transfer effects that would otherwise affect the radiance values.

scholarly journals Retrieval of Ice Cloud Optical Thickness and Effective Particle Size Using a Fast Infrared Radiative Transfer Model

2011 ◽  
Vol 50 (11) ◽  
pp. 2283-2297 ◽  
Author(s):  
Chenxi Wang ◽  
Ping Yang ◽  
Bryan A. Baum ◽  
Steven Platnick ◽  
Andrew K. Heidinger ◽  
...  
Keyword(s):  
Particle Size ◽  
Radiative Transfer ◽  
Optical Thickness ◽  
Radiative Transfer Model ◽  
Transfer Model ◽  
Distribution Method ◽  
Ice Cloud ◽  
Effective Particle Size ◽  
Effective Particle ◽  
Cloud Optical Thickness

AbstractA computationally efficient radiative transfer model (RTM) is developed for the inference of ice cloud optical thickness and effective particle size from satellite-based infrared (IR) measurements and is aimed at potential use in operational cloud-property retrievals from multispectral satellite imagery. The RTM employs precomputed lookup tables to simulate the top-of-the-atmosphere (TOA) radiances (or brightness temperatures) at 8.5-, 11-, and 12-μm bands. For the clear-sky atmosphere, the optical thickness of each atmospheric layer resulting from gaseous absorption is derived from the correlated-k-distribution method. The cloud reflectance, transmittance, emissivity, and effective temperature are precomputed using the Discrete Ordinate Radiative Transfer model (DISORT). For an atmosphere containing a semitransparent ice cloud layer with a visible optical thickness τ smaller than 5, the TOA brightness temperature differences (BTDs) between the fast model and the more rigorous DISORT results are less than 0.1 K, whereas the BTDs are less than 0.01 K if τ is larger than 10. With the proposed RTM, the cloud optical and microphysical properties are retrieved from collocated observations from the Moderate Resolution Imaging Spectroradiometer (MODIS) and Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) in conjunction with the Modern Era Retrospective-Analysis for Research and Applications (MERRA) data. Comparisons between the retrieved ice cloud properties (optical thickness and effective particle size) based on the present IR fast model and those from the Aqua/MODIS operational collection-5 cloud products indicate that the IR retrievals are smaller. A comparison between the IR-retrieved ice water path (IWP) and CALIOP-retrieved IWP shows robust agreement over most of the IWP range.

Finding Features of Positioning Error for Large Industrial Robots Based on Convolutional Neural Network

2021 ◽  
Author(s):  
Daiki Kato ◽  
Kenya Yoshitugu ◽  
Naoki Maeda ◽  
Toshiki Hirogaki ◽  
Eiichi Aoyama ◽  
...  
Keyword(s):  
Neural Network ◽  
Convolutional Neural Network ◽  
Production Systems ◽  
Industrial Robot ◽  
Three Dimensional ◽  
Industrial Robots ◽  
Numerical Control ◽  
Robot Motion ◽  
Positioning Error ◽  
End Effector

Abstract Most industrial robots are taught using the teaching playback method; therefore, they are unsuitable for use in variable production systems. Although offline teaching methods have been developed, they have not been practiced because of the low accuracy of the position and posture of the end-effector. Therefore, many studies have attempted to calibrate the position and posture but have not reached a practical level, as such methods consider the joint angle when the robot is stationary rather than the features during robot motion. Currently, it is easy to obtain servo information under numerical control operations owing to the Internet of Things technologies. In this study, we propose a method for obtaining servo information during robot motion and converting it into images to find features using a convolutional neural network (CNN). Herein, a large industrial robot was used. The three-dimensional coordinates of the end-effector were obtained using a laser tracker. The positioning error of the robot was accurately learned by the CNN. We extracted the features of the points where the positioning error was extremely large. By extracting the features of the X-axis positioning error using the CNN, the joint 1 current is a feature. This indicates that the vibration current in joint 1 is a factor in the X-axis positioning error.

Consistent retrieval of cloud/aerosol single scattering properties and surface reflectance

2021 ◽  
Author(s):  
Marta Luffarelli ◽  
Yves Govaerts
Keyword(s):  
Radiative Transfer ◽  
Optical Thickness ◽  
Solution Space ◽  
Single Scattering ◽  
Aerosol Optical Thickness ◽  
Radiative Transfer Model ◽  
Transfer Model ◽  
Surface Reflectance ◽  
Earth System ◽  
System Components

<p>The CISAR (Combined Inversion of Surface and AeRosols) algorithm is exploited in the framework of the ESA Aerosol Climate Change Initiatiave (CCI) project, aiming at providing a set of atmospheric (cloud and aerosol) and surface reflectance products derived from S3A/SLSTR observations using the same radiative transfer physics and assumptions. CISAR is an advance algorithm developed by Rayference originally designed for the retrieval of aerosol single scattering properties and surface reflectance from both geostationary and polar orbiting satellite observations.  It is based on the inversion of a fast radiative transfer model (FASTRE). The retrieval mechanism allows a continuous variation of the aerosol and cloud single scattering properties in the solution space.</p><p> </p><p>Traditionally, different approaches are exploited to retrieve the different Earth system components, which could lead to inconsistent data sets. The simultaneous retrieval of different atmospheric and surface variables over any type of surface (including bright surfaces and water bodies) with the same forward model and inversion scheme ensures the consistency among the retrieved Earth system components. Additionally, pixels located in the transition zone between pure clouds and pure aerosols are often discarded from both cloud and aerosol algorithms. This “twilight zone” can cover up to 30% of the globe. A consistent retrieval of both cloud and aerosol single scattering properties with the same algorithm could help filling this gap.</p><p> </p><p>The CISAR algorithm aims at overcoming the need of an external cloud mask, discriminating internally between aerosol and cloud properties. This approach helps reducing the overestimation of aerosol optical thickness in cloud contaminated pixels. The surface reflectance product is delivered both for cloud-free and cloudy observations.  </p><p> </p><p>Global maps obtained from the processing of S3A/SLSTR observations will be shown. The SLSTR/CISAR products over events such as, for instance, the Australian fire in the last months of 2019, will be discussed in terms of aerosol optical thickness, aerosol-cloud discrimination and fine/coarse mode fraction.</p>

Development of a Fast Three-Dimensional Dynamic Radiative Transfer Solver for Numerical Weather Prediction Models


2021 ◽  
Author(s):  
Richard Maier ◽  
Bernhard Mayer ◽  
Claudia Emde ◽  
Aiko Voigt
Keyword(s):  
Radiative Transfer ◽  
Numerical Weather Prediction ◽  
Prediction Models ◽  
Weather Prediction ◽  
Three Dimensional ◽  
Transfer Model ◽  
Heating Rates ◽  
Time Step ◽  
Numerical Weather ◽  
Numerical Weather Prediction Models

<div> <div> <div> <div> <p>The increasing resolution of numerical weather prediction models makes 3D radiative effects more and more important. These effects are usually neglected by the simple 1D independent column approximations used in most of the current models. On top of that, these 1D radiative transfer solvers are also called far less often than the model’s dynamical core.</p> <p>To address these issues, we present a new „dynamic“ approach of solving 3D radiative transfer. Building upon the existing TenStream solver (Jakub and Mayer, 2015), radiation in this 3D model is not solved completely in each radiation time step, but is rather only transported to adjacent grid boxes. For every grid box, outgoing fluxes are then calculated from the incoming fluxes from the neighboring grid cells of the previous time step. This allows to reduce the computational cost of 3D radiative transfer models to that of current 1D solvers.</p> <p>Here, we show first results obtained with this new solver with a special emphasis on heating rates. Furthermore, we demonstrate issues related to the dynamical treatment of radiation as well as possible solutions to these problems.</p> <p>In the future, the speed of this newly developed 3D dynamic TenStream solver will be further increased by reducing the number of spectral bands used in the radiative transfer calculations with the aim to get a 3D solver that can be called even more frequently than the 1D two-stream solvers used nowadays.</p> <p>Reference:<br><span>Jakub, F. and Mayer, B. (2015), A three-dimensional parallel radiative transfer model for atmospheric heating rates for use in cloud resolving models—The TenStream solver, Journal of Quantitative Spectroscopy and Radiative Transfer, Volume 163, 2015, Pages 63-71, ISSN 0022-4073, . </span></p> </div> </div> </div> </div>

Sign in / Sign up

Export Citation Format

Share Document