SEARCH · Engineering Papers

Results for “local clustering algorithm”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 37 records · Page 2

All sources NASA NTRS DOE OSTI

NASA NTRS2003

Archetypal TRMM Radar Profiles Identified Through Cluster Analysis

It is widely held that identifiable 'convective regimes' exist in nature, although precise definitions of these are elusive. Examples include land / Ocean distinctions, break / monsoon beahvior, seasonal differences in the Amazon (SON vs DJF), etc. These regimes are often described by differences in the realized local convective spectra, and measured by various metrics of convective intensity, depth, areal coverage and rainfall amount. Objective regime identification may be valuable in several ways: regimes may serve as natural 'branch points' in satellite retrieval algorithms or data assimilation efforts; one example might be objective identification of regions that 'should' share a similar 2-R relationship. Similarly, objectively defined regimes may provide guidance on optimal siting of ground validation efforts. Objectively defined regimes could also serve as natural (rather than arbitrary geographic) domain 'controls' in studies of convective response to environmental forcing. Quantification of convective vertical structure has traditionally involved parametric study of prescribed quantities thought to be important to convective dynamics: maximum radar reflectivity, cloud top height, 30-35 dBZ echo top height, rain rate, etc. Individually, these parameters are somewhat deficient as their interpretation is often nonunique (the same metric value may signify different physics in different storm realizations). Individual metrics also fail to capture the coherence and interrelationships between vertical levels available in full 3-D radar datasets. An alternative approach is discovery of natural partitions of vertical structure in a globally representative dataset, or 'archetypal' reflectivity profiles. In this study, this is accomplished through cluster analysis of a very large sample (0[107) of TRMM-PR reflectivity columns. Once achieved, the rainconditional and unconditional 'mix' of archetypal profile types in a given location and/or season provides a description of the local convective spectrum which retains vertical structure information. A further cluster analysis of these 'mixes' can identify recurrent convective spectra. These are a first step towards objective identification of convective regimes, and towards answering the question: 'What are the most convectively similar locations in the world?'

Boccippio, Dennis J.↗

NASA NTRS1995

Parallel processing for scientific computations

The scope of this project dealt with the investigation of the requirements to support distributed computing of scientific computations over a cluster of cooperative workstations. Various experiments on computations for the solution of simultaneous linear equations were performed in the early phase of the project to gain experience in the general nature and requirements of scientific applications. A specification of a distributed integrated computing environment, DICE, based on a distributed shared memory communication paradigm has been developed and evaluated. The distributed shared memory model facilitates porting existing parallel algorithms that have been designed for shared memory multiprocessor systems to the new environment. The potential of this new environment is to provide supercomputing capability through the utilization of the aggregate power of workstations cooperating in a cluster interconnected via a local area network. Workstations, generally, do not have the computing power to tackle complex scientific applications, making them primarily useful for visualization, data reduction, and filtering as far as complex scientific applications are concerned. There is a tremendous amount of computing power that is left unused in a network of workstations. Very often a workstation is simply sitting idle on a desk. A set of tools can be developed to take advantage of this potential computing power to create a platform suitable for large scientific computations. The integration of several workstations into a logical cluster of distributed, cooperative, computing stations presents an alternative to shared memory multiprocessor systems. In this project we designed and evaluated such a system.

Alkhatib, Hasan S.↗

NASA NTRS2012

Implementation and Testing of the First Prompt Search for Electromagnetic Counterparts to Gravitational Wave Transients

A transient astrophysical event observed in both gravitational wave (OW) and electromagnetic (EM) channels would yield rich scientific rewards. A first program initiating EM follow-ups to possible transient OW events has been developed and exercised by the LIGO and Virgo community in association with several partners. In this paper, we describe and evaluate the methods used to promptly identify and localize OW event candidates and to request images of targeted sky locations. Methods. During two observing periods (Dec 172009 to Jan 8 2010 and Sep 2 to Oct 20 2010), a low-latency analysis pipeline was used to identify OW event candidates and to reconstruct maps of possible sky locations. A catalog of nearby galaxies and Milky Way globular clusters was used to select the most promising sky positions to be imaged, and this directional information was delivered to EM observatories with time lags of about thirty minutes. A Monte Carlo simulation has been used to evaluate the low-latency OW pipeline's ability to reconstruct source positions correctly. Results. For signals near the detection threshold, our low-latency algorithms often localized simulated GW burst signals to tens of square degrees, while neutron star/neutron star inspirals and neutron star/black hole inspirals were localized to a few hundred square degrees. Localization precision improves for moderately stronger signals. The correct sky location of signals well above threshold and originating from nearby galaxies may be observed with approx.50% or better probability with a few pointings of wide-field telescopes. Key words. gravitational waves - methods: observational

Camp, Jordan↗

NASA NTRS2012

Implementation and Testing of the First Prompt Search for Gravitational Wave Transients with Electromagnetic Counterparts

Aims. A transient astrophysical event observed in both gravitational wave (GW) and electromagnetic (EM) channels would yield rich scientific rewards. A first program initiating EM follow-ups to possible transient GW events has been developed and exercised by the LIGO and Virgo community in association with several partners. In this paper, we describe and evaluate the methods used to promptly identify and localize GW event candidates and to request images of targeted sky locations. Methods. During two observing periods (Dec. 17, 2009 to Jan. 8, 2010 and Sep. 2 to Oct. 20, 2010), a low-latency analysis pipeline was used to identify GW event candidates and to reconstruct maps of possible sky locations. A catalog of nearby galaxies and MilkyWay globular clusters was used to select the most promising sky positions to be imaged, and this directional information was delivered to EM observatories with time lags of about thirty minutes. A Monte Carlo simulation has been used to evaluate the low-latency GW pipeline's ability to reconstruct source positions correctly. Results. For signals near the detection threshold, our low-latency algorithms often localized simulated GW burst signals to tens of square degrees, while neutron star/neutron star inspirals and neutron star/black hole inspirals were localized to a few hundred square degrees. Localization precision improves for moderately stronger signals. The correct sky location of signals well above threshold and originating from nearby galaxies may be observed with ∼50% or better probability with a few pointings of wide-field telescopes.

testing↗

NASA NTRS2011

Implementation and Testing of the First Prompt for Electromagnetic Counterparts to Gravitational Wave Transients

A transient astrophysical event observed in both gravitational wave (GW) and electromagnetic (EM) channels would yield rich scientific rewards. A first program initiating EM follow-ups to possible transient GW events has been developed and exercised by the LIGO and Virgo community in association with several partners. In this paper, we describe and evaluate the methods used to promptly identify and localize GW event candidates and to request images of targeted sky locations. Methods. During two observing periods (Dec 17 2009 to Jan 8 2010 and Sep 2 to Oct 20 2010), a low-latency analysis pipeline was used to identify GW-event candidates and to reconstruct-maps of possible sky locations. A catalog of nearby galaxies and Milky Way globular clusters was used to select the most promising sky positions to be imaged, and this directional information was delivered to EM observatories with time lags of about thirty minutes. A Monte Carlo simulation has been used to evaluate the low-latency GW pipeline s ability to reconstruct source positions correctly. Results. For signals near the detection threshold, our low-latency algorithms often localized simulated GW burst signals to tens of square degrees, while neutron star/neutron star inspirals and neutron star/black hole inspirals were localized to a few hundred square degrees. Localization precision improves for moderately stronger signals. The correct sky location of signals well above threshold and originating from nearby galaxies may be observed with 50% or better probability with a few pointings of wide-field telescopes.

Abadie, J.↗

NASA NTRS2010

Initialization and Restart in Stochastic Local Search: Computing a Most Probable Explanation in Bayesian Networks

For hard computational problems, stochastic local search has proven to be a competitive approach to finding optimal or approximately optimal problem solutions. Two key research questions for stochastic local search algorithms are: Which algorithms are effective for initialization? When should the search process be restarted? In the present work we investigate these research questions in the context of approximate computation of most probable explanations (MPEs) in Bayesian networks (BNs). We introduce a novel approach, based on the Viterbi algorithm, to explanation initialization in BNs. While the Viterbi algorithm works on sequences and trees, our approach works on BNs with arbitrary topologies. We also give a novel formalization of stochastic local search, with focus on initialization and restart, using probability theory and mixture models. Experimentally, we apply our methods to the problem of MPE computation, using a stochastic local search algorithm known as Stochastic Greedy Search. By carefully optimizing both initialization and restart, we reduce the MPE search time for application BNs by several orders of magnitude compared to using uniform at random initialization without restart. On several BNs from applications, the performance of Stochastic Greedy Search is competitive with clique tree clustering, a state-of-the-art exact algorithm used for MPE computation in BNs.

Mengshoel, Ole J.↗

NASA NTRS2012

Classification of Ascension Island and Natal Ozonesondes Using Self-Organizing Maps

Ozone profiles from balloon-borne ozonesondes are used for development of satellite algorithms and in chemistry-climate model initialization, assimilation and evaluation. An important issue in the application of these profiles is how best to treat variations where varying photochemical and dynamical influences can cause the ozone mixing ratio in the tropospheric segments of the profile to change by of a factor of 2-3 within a day. Clustering techniques are an ideal way to approach the statistical classification of profile data and we apply self-organizing maps to tropical tropospheric SHADOZ data, hypothesizing that the data will sort according to various influences on ozone, namely anthropogenic sources like biomass burning, meteorological conditions, and stratospheric or extra-tropical intrusions. Self-organizing maps, that use a learning algorithm to reveal the most prominent features of a data set according to a specified number of clusters, have been determined for the 1998-2009 SHADOZ profiles over Ascension Island (512 profiles, 7.98 deg. S, 14.42 deg. W) and Natal, Brazil (425 profiles, 5.42degS, 35.38degW). The 2 × 2 self-organizing map, which creates 4 clusters, reveals that deviations from the average ozone in the free troposphere include both increased ozone resulting from seasonal biomass burning in Africa and locally reduced ozone brought about by convective lifting of unpolluted boundary-layer air. Expanding to a 4 × 4 self-organizing map shows how biomass burning influences the yearly cycle of tropospheric ozone at Ascension Island and captures the seasonality of ozone at both Ascension Island and Natal. Comparing Ascension Island and Natal using a 4 × 4 self-organizing map at each site reveals similarities in mid-tropospheric ozone, but shows differences in lower-tropospheric ozone due to Ascension Island being closer to African biomass burning and more affected by descent from the mean Walker circulation, with less convective activity, than Natal.

algorithms↗

NASA NTRS1996

Adaptive grid methods for RLV environment assessment and nozzle analysis

Rapid access to highly accurate data about complex configurations is needed for multi-disciplinary optimization and design. In order to efficiently meet these requirements a closer coupling between the analysis algorithms and the discretization process is needed. In some cases, such as free surface, temporally varying geometries, and fluid structure interaction, the need is unavoidable. In other cases the need is to rapidly generate and modify high quality grids. Techniques such as unstructured and/or solution-adaptive methods can be used to speed the grid generation process and to automatically cluster mesh points in regions of interest. Global features of the flow can be significantly affected by isolated regions of inadequately resolved flow. These regions may not exhibit high gradients and can be difficult to detect. Thus excessive resolution in certain regions does not necessarily increase the accuracy of the overall solution. Several approaches have been employed for both structured and unstructured grid adaption. The most widely used involve grid point redistribution, local grid point enrichment/derefinement or local modification of the actual flow solver. However, the success of any one of these methods ultimately depends on the feature detection algorithm used to determine solution domain regions which require a fine mesh for their accurate representation. Typically, weight functions are constructed to mimic the local truncation error and may require substantial user input. Most problems of engineering interest involve multi-block grids and widely disparate length scales. Hence, it is desirable that the adaptive grid feature detection algorithm be developed to recognize flow structures of different type as well as differing intensity, and adequately address scaling and normalization across blocks. These weight functions can then be used to construct blending functions for algebraic redistribution, interpolation functions for unstructured grid generation, forcing functions to attract/repel points in an elliptic system, or to trigger local refinement, based upon application of an equidistribution principle. The popularity of solution-adaptive techniques is growing in tandem with unstructured methods. The difficultly of precisely controlling mesh densities and orientations with current unstructured grid generation systems has driven the use of solution-adaptive meshing. Use of derivatives of density or pressure are widely used for construction of such weight functions, and have been proven very successful for inviscid flows with shocks. However, less success has been realized for flowfields with viscous layers, vortices or shocks of disparate strength. It is difficult to maintain the appropriate mesh point spacing in the various regions which require a fine spacing for adequate resolution. Mesh points often migrate from important regions due to refinement of dominant features. An example of this is the well know tendency of adaptive methods to increase the resolution of shocks in the flowfield around airfoils, but in the incorrect location due to inadequate resolution of the stagnation region. This problem has been the motivation for this research.

Thornburg, Hugh J.↗

NASA NTRS1988

Transonic Navier-Stokes computations of strake-generated vortex interactions for a fighter-like configuration

Transonic Euler/Navier-Stokes computations are accomplished for wing-body flow fields using a computer program called Transonic Navier-Stokes (TNS). The wing-body grids are generated using a program called ZONER, which subdivides a coarse grid about a fighter-like aircraft configuration into smaller zones, which are tailored to local grid requirements. These zones can be either finely clustered for capture of viscous effects, or coarsely clustered for inviscid portions of the flow field. Different equation sets may be solved in the different zone types. This modular approach also affords the opportunity to modify a local region of the grid without recomputing the global grid. This capability speeds up the design optimization process when quick modifications to the geometry definition are desired. The solution algorithm embodied in TNS is implicit, and is capable of capturing pressure gradients associated with shocks. The algebraic turbulence model employed has proven adequate for viscous interactions with moderate separation. Results confirm that the TNS program can successfully be used to simulate transonic viscous flows about complicated 3-D geometries.

Reznick, Steve↗

NASA NTRSDate not supplied

NASA’s Prototype Spectral Water Inversion Processor and Emulator (SWIPE): Towards Global Coastal and Inland Water Quality and Algal Biodiversity Monitoring

Degradation of Earth’s inland water resources due to anthropogenic perturbations and climate anomalies at both local and global scales continues to place human health at substantial risk. There is now a growing necessity to develop pragmatic approaches that allow timely and effective extrapolation of local processes, to spatially resolved global products, and to promote operational and sustainable resource policy management. This presentation will provide updates on NASA’s prototype open-source aquatic modeling platform, Spectral Water Inversion Processor and Emulator (SWIPE), which is a comprehensive, multi-faceted modeling platform for both forward and inverse modeling of diverse aquatic ecosystems from the benthos to top-of-atmosphere (TOA). SWIPE provides a cohesive application which leverages recent advancements in particle modeling, Big Data analytics, and machine learning to develop a high-fidelity synthetic training ground for sensitivity studies and algorithm development for multispectral or upcoming hyperspectral missions. Some of the prominent features of SWIPE to be discussed include: 1. Advanced hyperspectral modeling of globally diverse algal and non-algal particles using a novel two-layer coated sphere scattering model and radiative transfer modeling, 2. Massive, highly detailed synthetic spectral libraries of Analysis-Ready-Data (ARD) which include spectral libraries of particle microphysics, water biogeophysical and optical properties, as well as surface and TOA reflectances at 1 nm resolution, 3. An ensemble of pre-built analytic, machine learning, and deep learning inversion algorithms for various water quality and biodiversity related retrieval parameters and uncertainty quantification, 4. Sensor-agnostic water quality inversion at wide ranging spatial and spectral resolutions including a codebase for seamless application in the Google Earth Engine and NASA Earth Exchange (NEX) for planetary scale analysis. SWIPE will be a fully open-source platform based in python with comprehensive documentation, tutorials, and options for distributed computing on high performance computing clusters or on single, local machines. Further, we will discuss how we envision SWIPE contributing towards a global analysis of coastal and inland water quality dynamics.

top-of-atmosphere (TOA)↗

NASA NTRS2010

Extracting Independent Local Oscillatory Geophysical Signals by Geodetic Tropospheric Delay

Zenith Tropospheric Delay (ZTD) due to water vapor derived from space geodetic techniques and numerical weather prediction simulated-reanalysis data exhibits non-linear and non-stationary properties akin to those in the crucial geophysical signals of interest to the research community. These time series, once decomposed into additive (and stochastic) components, have information about the long term global change (the trend) and other interpretable (quasi-) periodic components such as seasonal cycles and noise. Such stochastic component(s) could be a function that exhibits at most one extremum within a data span or a monotonic function within a certain temporal span. In this contribution, we examine the use of the combined Ensemble Empirical Mode Decomposition (EEMD) and Independent Component Analysis (ICA): the EEMD-ICA algorithm to extract the independent local oscillatory stochastic components in the tropospheric delay derived from the European Centre for Medium-Range Weather Forecasts (ECMWF) over six geodetic sites (HartRAO, Hobart26, Wettzell, Gilcreek, Westford, and Tsukub32). The proposed methodology allows independent geophysical processes to be extracted and assessed. Analysis of the quality index of the Independent Components (ICs) derived for each cluster of local oscillatory components (also called the Intrinsic Mode Functions (IMFs)) for all the geodetic stations considered in the study demonstrate that they are strongly site dependent. Such strong dependency seems to suggest that the localized geophysical signals embedded in the ZTD over the geodetic sites are not correlated. Further, from the viewpoint of non-linear dynamical systems, four geophysical signals the Quasi-Biennial Oscillation (QBO) index derived from the NCEP/NCAR reanalysis, the Southern Oscillation Index (SOI) anomaly from NCEP, the SIDC monthly Sun Spot Number (SSN), and the Length of Day (LoD) are linked to the extracted signal components from ZTD. Results from the synchronization analysis show that ZTD and the geophysical signals exhibit (albeit subtle) site dependent phase synchronization index.

Botai, O. J.↗

NASA NTRS2020

An Algorithm for Characterization of Fiber Aggregation in Composite Microstructures

Composite structures are susceptible to localized flaws, or variability, that drive global failure. To capture this variance, a multiscale model needs to be introduced that not only accurately represents the statistical nature of the composite microstructure but is also efficient. At the microscale, fiber aggregation creates local stress concentrations, where failure is likely to occur sooner than expected. A method of rapid microstructure generation, in conjunction with cluster characterization, is examined to develop accurate reproductions of 2D composite cross-sections. Parameterization of shape and size of fiber clusters is used to characterize representative volume elements. The discrete element method is used to generate pseudo-microstructures. The clustering parameters from the pseudo- and actual microstructure arrangements (obtained from micrographs) can be compared to determine the validity of the representative volume element generated with the discrete element method. Results show a promising approach to evaluating randomness of fiber distributions and how to accurately recreate microstructures for strength analysis.

Carbon fiber↗

NASA NTRS2016

A Spatiotemporal Indexing Approach for Efficient Processing of Big Array-Based Climate Data with MapReduce

Climate observations and model simulations are producing vast amounts of array-based spatiotemporal data. Efficient processing of these data is essential for assessing global challenges such as climate change, natural disasters, and diseases. This is challenging not only because of the large data volume, but also because of the intrinsic high-dimensional nature of geoscience data. To tackle this challenge, we propose a spatiotemporal indexing approach to efficiently manage and process big climate data with MapReduce in a highly scalable environment. Using this approach, big climate data are directly stored in a Hadoop Distributed File System in its original, native file format. A spatiotemporal index is built to bridge the logical array-based data model and the physical data layout, which enables fast data retrieval when performing spatiotemporal queries. Based on the index, a data-partitioning algorithm is applied to enable MapReduce to achieve high data locality, as well as balancing the workload. The proposed indexing approach is evaluated using the National Aeronautics and Space Administration (NASA) Modern-Era Retrospective Analysis for Research and Applications (MERRA) climate reanalysis dataset. The experimental results show that the index can significantly accelerate querying and processing (10 speedup compared to the baseline test using the same computing cluster), while keeping the index-to-data ratio small (0.0328). The applicability of the indexing approach is demonstrated by a climate anomaly detection deployed on a NASA Hadoop cluster. This approach is also able to support efficient processing of general array-based spatiotemporal data in various geoscience domains without special configuration on a Hadoop cluster.

big data↗

NASA NTRS1986

The creation of local clusters in arbitrarily given grids

A method is presented to smoothly insert pointwise clusters into any given grid regardless of its origin, its topology, or its dimensionality. The process amounts to a local movement of the given coordinate curves or surfaces to more highly resolve an object. The object about which clustering is created can be a point, a curve, a surface, or segments of a curve or surface. The basic clustering capability is established by forming a grid operator for a single cluster. With a view toward multiple clusters being created about various objects, the basic operator is seen as an elementary operator. An algorithm is presented to execute the general elementary operation in three dimensions. In FORTRAN, this assumes the form of a subroutine which is fully operational and is presented to serve as a basic model for any such elementary clustering operation.

Eiseman, Peter R.↗

NASA NTRS2013

Mixture-Tuned, Clutter Matched Filter for Remote Detection of Subpixel Spectral Signals

Mapping localized spectral features in large images demands sensitive and robust detection algorithms. Two aspects of large images that can harm matched-filter detection performance are addressed simultaneously. First, multimodal backgrounds may thwart the typical Gaussian model. Second, outlier features can trigger false detections from large projections onto the target vector. Two state-of-the-art approaches are combined that independently address outlier false positives and multimodal backgrounds. The background clustering models multimodal backgrounds, and the mixture tuned matched filter (MT-MF) addresses outliers. Combining the two methods captures significant additional performance benefits. The resulting mixture tuned clutter matched filter (MT-CMF) shows effective performance on simulated and airborne datasets. The classical MNF transform was applied, followed by k-means clustering. Then, each cluster s mean, covariance, and the corresponding eigenvalues were estimated. This yields a cluster-specific matched filter estimate as well as a cluster- specific feasibility score to flag outlier false positives. The technology described is a proof of concept that may be employed in future target detection and mapping applications for remote imaging spectrometers. It is of most direct relevance to JPL proposals for airborne and orbital hyperspectral instruments. Applications include subpixel target detection in hyperspectral scenes for military surveillance. Earth science applications include mineralogical mapping, species discrimination for ecosystem health monitoring, and land use classification.

Thompson, David R.↗

NASA NTRS1985

Severe Storms Branch research report (April 1984 April 1985)

The Mesoscale Atmospheric Processes Research Program is a program of integrated studies which are to achieve an improved understanding of the basic behavior of the atmosphere through the use of remotely sensed data and space technology. The program consist of four elements: (1) special observations and analysis of mesoscale systems; (20 the development of quanitative algorithms to use remotely sensed observations; (3) the development of new observing systems; and (4) numerical modeling. The Severe Storms Branch objectives are the improvement of the understanding, diagnosis, and prediction of a wide range of atmospheric storms, which includes severe thunderstorms, tornadoes, flash floods, tropical cyclones, and winter snowstorms. The research often shed light upon various aspects of local weather, such as fog, sea breezes, air pollution, showers, and other products of nonsevere cumulus cloud clusters. The part of the program devoted to boundary layer processes, gust front interactions, and soil moisture detection from satellites gives insights into storm growth and behavior.

Dubach, L.↗

NASA NTRS1995

Stereo-Based Region-Growing using String Matching

We present a novel stereo algorithm based on a coarse texture segmentation preprocessing phase. Matching is performed using a string comparison. Matching sub-strings correspond to matching sequences of textures. Inter-scanline clustering of matching sub-strings yields regions of matching texture. The shape of these regions yield information concerning object's height, width and azimuthal position relative to the camera pair. Hence, rather than the standard dense depth map, the output of this algorithm is a segmentation of objects in the scene. Such a format is useful for the integration of stereo with other sensor modalities on a mobile robotic platform. It is also useful for localization; the height and width of a detected object may be used for landmark recognition, while depth and relative azimuthal location determine pose. The algorithm does not rely on the monotonicity of order of image primitives. Occlusions, exposures, and foreshortening effects are not problematic. The algorithm can deal with certain types of transparencies. It is computationally efficient, and very amenable to parallel implementation. Further, the epipolar constraints may be relaxed to some small but significant degree. A version of the algorithm has been implemented and tested on various types of images. It performs best on random dot stereograms, on images with easily filtered backgrounds (as in synthetic images), and on real scenes with uncontrived backgrounds.

Mandelbaum, Robert↗

NASA NTRS2017

H31G-1596: DeepSAT's CloudCNN: A Deep Neural Network for Rapid Cloud Detection from Geostationary Satellites

Cloud and cloud shadow detection has important applications in weather and climate studies. It is even more crucial when we introduce geostationary satellites into the field of terrestrial remote sensing. With the challenges associated with data acquired in very high frequency (10-15 mins per scan), the ability to derive an accurate cloud shadow mask from geostationary satellite data is critical. The key to the success for most of the existing algorithms depends on spatially and temporally varying thresholds,which better capture local atmospheric and surface effects.However, the selection of proper threshold is difficult and may lead to erroneous results. In this work, we propose a deep neural network based approach called CloudCNN to classify cloudshadow from Himawari-8 AHI and GOES-16 ABI multispectral data. DeepSAT's CloudCNN consists of an encoderdecoder based architecture for binary-class pixel wise segmentation. We train CloudCNN on multi-GPU Nvidia Devbox cluster, and deploy the prediction pipeline on NASA Earth Exchange (NEX) Pleiades supercomputer. We achieved an overall accuracy of 93.29% on test samples. Since, the predictions take only a few seconds to segment a full multispectral GOES-16 or Himawari-8 Full Disk image, the developed framework can be used for real-time cloud detection, cyclone detection, or extreme weather event predictions.

GOES-1↗