SEARCH · Search NASA
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.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
The purpose of the 1988 ASEE Summer Program has been to broaden the application of texture-based cloud classification approaches to lower spatial resolution GOES imagery, and to design texture-based approaches for determining cloud cover over high albedo surfaces.
The loss of cloud-classification accuracy as a function of spatial resolution is assessed by investigating the variation of textural measures as a function of spatial resolution. Landsat MSS imagery is progressively averaged to produce degraded imagery of 1/8-km, 1/4-km, 1/2-km, and 1-km spatial resolution, and textural measures are computed from the Gray Level Difference Vector (GLDV) approach described by Chen et al. (1989). It is found that the classification accuracies obtained using the 1/8-km spatial resolution data are similar to those obtained using the full-resolution (1/16 km) texture measures, indicating that there is no advantage in using even higher spatial resolution 30-m Landsat Thematic Mapper and 10-m SPOT imagery for cloud classification.
The study examines the textural signatures of clouds, ice-covered mountains, solid and broken sea ice and floes, and open water. The textural features are computed from sum and difference histogram and gray-level difference vector statistics defined at various pixel displacement distances derived from Landsat multispectral scanner data. Polar cloudiness, snow-covered mountainous regions, solid sea ice, glaciers, and open water have distinguishable texture features. This suggests that textural measures can be successfully applied to the detection of clouds over snow-covered mountains, an ability of considerable importance for the modeling of snow-melt runoff. However, broken stratocumulus cloud decks and thin cirrus over broken sea ice remain difficult to distinguish texturally. It is concluded that even with high spatial resolution imagery, it may not be possible to distinguish broken stratocumulus and thin clouds from sea ice in the marginal ice zone using the visible channel textural features alone.
Clouds play an important role in the Earth’s energy budget, and their behavior is one of the largest uncertainties in future climate projections. Satellite observations should help in understanding cloud responses, but decades and petabytes of multispectral cloud imagery have to date received only limited use. This study describes a new analysis approach that reduces the dimensionality of satellite cloud observations by grouping them via a novel automated, unsupervised cloud classification technique based on a convolutional autoencoder, an artificial intelligence (AI) method good at identifying patterns in spatial data. Our technique combines a rotation-invariant autoencoder and hierarchical agglomerative clustering to generate cloud clusters that capture meaningful distinctions among cloud textures, using only raw multispectral imagery as input. Cloud classes are therefore defined based on spectral properties and spatial textures without reliance on location, time/season, derived physical properties, or pre-designated class definitions. We use this approach to generate a unique new cloud dataset, the AI-driven cloud classification atlas (AICCA), which clusters 22 years of ocean images from the Moderate Resolution Imaging Spectroradiometer (MODIS) on NASA’s Aqua and Terra instruments—198 million patches, each roughly 100 km × 100 km (128 × 128 pixels)—into 42 AI-generated cloud classes, a number determined via a newly-developed stability protocol that we use to maximize richness of information while ensuring stable groupings of patches. AICCA thereby translates 801 TB of satellite images into 54.2 GB of class labels and cloud top and optical properties, a reduction by a factor of 15,000. The 42 AICCA classes produce meaningful spatio-temporal and physical distinctions and capture a greater variety of cloud types than do the nine International Satellite Cloud Climatology Project (ISCCP) categories—for example, multiple textures in the stratocumulus decks along the West coasts of North and South America. We conclude that our methodology has explanatory power, capturing regionally unique cloud classes and providing rich but tractable information for global analysis. AICCA delivers the information from multi-spectral images in a compact form, enables data-driven diagnosis of patterns of cloud organization, provides insight into cloud evolution on timescales of hours to decades, and helps democratize climate research by facilitating access to core data.
A feasibility study of automated classification of satellite images is described. Satellite images were characterized by the textures they contain. In particular, the detection of cloud textures was investigated. The method of second-order gray level statistics, using co-occurrence matrices, was applied to extract feature vectors from image segments. Neural network technology was employed to classify these feature vectors. The cascade-correlation architecture was successfully used as a classifier. The use of a Kohonen network was also investigated but this architecture could not reliably classify the feature vectors due to the complicated structure of the classification problem. The best results were obtained when data from different spectral bands were fused.
The present study reexamines the applicability of texture-based features for automatic cloud classification using very high spatial resolution (57 m) Landsat multispectral scanner digital data. It is concluded that cloud classification can be accomplished using only a single visible channel.
An essential component in global climate research is accurate cloud cover and type determination. Of the two approaches to texture-based classification (statistical and textural), only the former is effective in the classification of natural scenes such as land, ocean, and atmosphere. In the statistical approach that was adopted, parameters characterizing the stochastic properties of the spatial distribution of grey levels in an image are estimated and then used as features for cloud classification. Two types of textural measures were used. One is based on the distribution of the grey level difference vector (GLDV), and the other on a set of textural features derived from the MaxMin cooccurrence matrix (MMCM). The GLDV method looks at the difference D of grey levels at pixels separated by a horizontal distance d and computes several statistics based on this distribution. These are then used as features in subsequent classification. The MaxMin tectural features on the other hand are based on the MMCM, a matrix whose (I,J)th entry give the relative frequency of occurrences of the grey level pair (I,J) that are consecutive and thresholded local extremes separated by a given pixel distance d. Textural measures are then computed based on this matrix in much the same manner as is done in texture computation using the grey level cooccurrence matrix. The database consists of 37 cloud field scenes from LANDSAT imagery using a near IR visible channel. The classification algorithm used is the well known Stepwise Discriminant Analysis. The overall accuracy was estimated by the percentage or correct classifications in each case. It turns out that both types of classifiers, at their best combination of features, and at any given spatial resolution give approximately the same classification accuracy. A neural network based classifier with a feed forward architecture and a back propagation training algorithm is used to increase the classification accuracy, using these two classes of features. Preliminary results based on the GLDV textural features alone look promising.
Stratocumulus, cumulus, and cirrus clouds were identified on the basis of cloud textural features which were derived from a single high-resolution Landsat MSS NIR channel using a stepwise linear discriminant analysis. It is shown that, using this method, it is possible to distinguish high cirrus clouds from low clouds with high accuracy on the basis of spatial brightness patterns. The largest probability of misclassification is associated with confusion between the stratocumulus breakup regions and the fair-weather cumulus.
Arctic clouds and ice-covered surfaces are classified on the basis of textural and spectral features obtained with AVHRR 1.1-km spatial resolution imagery over the Beaufort Sea during May-October, 1989. Scenes were acquired about every 5 days, for a total of 38 cases. A list comprising 20 arctic-surface and cloud classes is compiled using spectral measures defined by Garand (1988).
Using the gray-level difference vector approach, classification accuracies with 1/8-km spatial-resolution data are similar to those obtained using the full spatial-resolution features. Hence no advantage is to be gained in cloud classification accuracies by using even higher spatial resolutions obtained from Landsat TM or SPOT imagery. The optimum spatial resolution is 1/4 km. However, significant improvement in cloud-classification accuracy compared to that available from the 1-km resolution of AVHRR and GOES imagery is obtained using 1/2-km-resolution data. Cirrus-classification accuracy is especially compromised as spatial resolution is degraded. However, texture measures defined at the combination of pixel separations d = 1,4 improve classification accuracies by several percent, even for 1-km spatial-resolution data. Cirrus-classification accuracy is significantly improved by the use of multiple distance features.
The TOA/Cloud Albedo data contain albedo values, including finely-sampled or local (2.2 km) TOA albedos registered to the RLRA, and two coarsely-sampled (35.2 km resolution) TOA albedos projected to 30-km altitude. The local (2.2 km) albedos do not take the obscuration of cloud features into account, so they should only be treated as traditional albedos when the number of obscured pixels is low. The restrictive and expansive albedos are both available at 35.2 km resolution: the restrictive albedos are only calculated using the radiation upwelling from the pixel under consideration, whereas the expansive albedos use all the radiation emanating from the surrounding area. Therefore, the expansive albedo is closer to the traditional definition of top-of-atmosphere albedos. [Location=GLOBAL] [Temporal_Coverage: Start_Date=2000-02-24; Stop_Date=] [Spatial_Coverage: Southernmost_Latitude=-90; Northernmost_Latitude=90; Westernmost_Longitude=-180; Easternmost_Longitude=180] [Data_Resolution: Latitude_Resolution=35.2 km; Longitude_Resolution=35.2 km; Horizontal_Resolution_Range=10 km - < 50 km or approximately .09 degree - < .5 degree; Temporal_Resolution=about 15 orbits/day].
The TOA/Cloud Albedo data contain albedo values, including finely-sampled or local (2.2 km) TOA albedos registered to the RLRA, and two coarsely-sampled (35.2 km resolution) TOA albedos projected to 30-km altitude. The local (2.2 km) albedos do not take the obscuration of cloud features into account, so they should only be treated as traditional albedos when the number of obscured pixels is low. The restrictive and expansive albedos are both available at 35.2 km resolution: the restrictive albedos are only calculated using the radiation upwelling from the pixel under consideration, whereas the expansive albedos use all the radiation emanating from the surrounding area. Therefore, the expansive albedo is closer to the traditional definition of top-of-atmosphere albedos. [Temporal_Coverage: Start_Date=2000-02-24; Stop_Date=] [Spatial_Coverage: Southernmost_Latitude=-90; Northernmost_Latitude=90; Westernmost_Longitude=-180; Easternmost_Longitude=180] [Data_Resolution: Latitude_Resolution=1.1 km; Longitude_Resolution=1.1 km; Temporal_Resolution=about 15 orbits/day].
The spectral and textural characteristics of polar clouds and surfaces for a 7-day summer series of AVHRR data in two Arctic locations are examined, and the results used in the development of a cloud classification procedure for polar satellite data. Since spatial coherence and texture sensitivity tests indicate that a joint spectral-textural analysis based on the same cell size is inappropriate, cloud detection with AVHRR data and surface identification with passive microwave data are first done on the pixel level as described by Key and Barry (1989). Next, cloud patterns within 250-sq-km regions are described, then the spectral and local textural characteristics of cloud patterns in the image are determined and each cloud pixel is classified by statistical methods. Results indicate that both spectral and textural features can be utilized in the classification of cloudy pixels, although spectral features are most useful for the discrimination between cloud classes.
A technique has been developed to display GOES satellite cloud images in perspective over a topographical map. Cloud heights are estimated using temperatures from an infrared (IR) satellite image, surface temperature observations, and a climatological model of vertical temperature profiles. Cloud levels are discriminated from each other and from the ground using a pattern recognition algorithm based on the brightness variance technique of Coakley and Bretherton. The cloud regions found by the pattern recognizer are rendered in three-dimensional perspective over a topographical map by an efficient remap of the visible image. The visible shades are mixed with an artificial shade based on the geometry of the cloud-top surface, in order to enhance the texture of the cloud top.
This paper compares the results of cloud-field classification derived from two simplified vector approaches, the Sum and Difference Histogram (SADH) and the Gray Level Difference Vector (GLDV), with the results produced by the Gray Level Cooccurrence Matrix (GLCM) approach described by Welch et al. (1988). It is shown that the SADH method produces accuracies equivalent to those obtained using the GLCM method, while the GLDV method fails to resolve error clusters. Compared to the GLCM method, the SADH method leads to a 31 percent saving in run time and a 50 percent saving in storage requirements, while the GLVD approach leads to a 40 percent saving in run time and an 87 percent saving in storage requirements.
Daily global imaging by the Mars Color Imager (MARCI) continues the record of the Mars Orbiter Camera (MOC) and has allowed creation of a long-duration record of Martian dust storms. We observe dust storms over the first two Mars years of the MARCI record, including tracking individual storms over multiple sols, as well as tracking the growth and recession of the seasonal polar caps. Using the combined 6 Mars year record of textured dust storms (storms with visible textures on the observed dust cloud tops), we study the relationship between textured dust storm activity and meteorology (as simulated by the MarsWRF general circulation model) and surface properties. We find that textured dust storms preferentially occur in places and seasons with above average surface wind stress. Textured dust storm occurrence also has a modest linear anti-correlation with surface albedo (0.43) and topography (0.40). Lastly, we perform an empirical orthogonal function (EOF) analysis on the distribution of occurrence of textured dust storms and find that over 50 of the variance in textured dust storm activity can be explained by two EOF modes. We associate the first EOF mode with cap-edge storms just before Ls = 180deg and the second EOF mode with flushing dust storms that occur from Ls = 180-210deg and again near Ls = 320deg.
Various papers on remote sensing (RS) for the nineties are presented. The general topics addressed include: subsurface methods, radar scattering, oceanography, microwave models, atmospheric correction, passive microwave systems, RS in tropical forests, moderate resolution land analysis, SAR geometry and SNR improvement, image analysis, inversion and signal processing for geoscience, surface scattering, rain measurements, sensor calibration, wind measurements, terrestrial ecology, agriculture, geometric registration, subsurface sediment geology, radar modulation mechanisms, radar ocean scattering, SAR calibration, airborne radar systems, water vapor retrieval, forest ecosystem dynamics, land analysis, multisensor data fusion. Also considered are: geologic RS, RS sensor optical measurements, RS of snow, temperature retrieval, vegetation structure, global change, artificial intelligence, SAR processing techniques, geologic RS field experiment, stochastic modeling, topography and Digital Elevation model, SAR ocean waves, spaceborne lidar and optical, sea ice field measurements, millimeter waves, advanced spectroscopy, spatial analysis and data compression, SAR polarimetry techniques. Also discussed are: plant canopy modeling, optical RS techniques, optical and IR oceanography, soil moisture, sea ice back scattering, lightning cloud measurements, spatial textural analysis, SAR systems and techniques, active microwave sensing, lidar and optical, radar scatterometry, RS of estuaries, vegetation modeling, RS systems, EOS/SAR Alaska, applications for developing countries, SAR speckle and texture.
Severe weather phenomena not only are responsible for damages to property, infrastructure, and agriculture, they are also difficult to measure in-situ and have been associated with large errors and uncertainties in precipitation estimation that pose challenges to establishing the current climatology of severe storms and their future trend in a changing Earth system. Severe convection exhibits distinct signatures in remote-sensing datasets, where it is manifested as textured or overshooting cloud tops in visible/IR imagery, or resulting a prominent brightness temperature depression in passive-microwave imagery. These distinct deep convective signatures in spaceborne datasets have been leveraged to analyze severe thunderstorms, create climatologies, improve prediction, and provide a method of detection around the globe where traditional ground-based data may be inconsistent or unavailable. Visible/IR and passive-microwave instruments are powerful tools for detecting severe thunderstorms, however, these datasets exhibit key limitations when examined individually: passive-microwave channels have large footprint and exhibit non-uniform beam filling. The visible/IR instruments have fine horizontal resolution but are limited by their insensitivity to processes occurring below cloud top. We investigate the nearly simultaneous and colocated MODIS (visible/IR) and AMSR-E (passive-microwave) onboard the Aqua satellite to leverage both datasets together and assess the extent to which these datasets can be combined to improve severe thunderstorm detection. We pair passive microwave, IR, and visible signatures of severe convection with ground-based weather radar, severe weather reports, and environmental parameters defined by the MERRA-2 reanalysis to characterize potentially severe convective storms observed by Aqua MODIS and AMSR-E. We will discuss how MODIS and AMSR-E signatures may be used together to diagnose storm properties and processes, and how the interrelationships between the signatures varies seasonally and geographically.