NG43A-0435 1340h
Web-based Data Mining to Systematically Determine Data Quality From the EarthScope USArray Seismic Observatory Project
When fully operational, the Transportable Array (TA) and Flexible Array (FA) components of the continent-scale EarthScope USArray seismic observatory project will provide telemetered real-time data from up to 600 stations. By the fifth year of the deployment the predicted total amount of data production for the TA and FA will be approximately 1500 Gb/yr and approximately 1000 Gb/yr respectively. In addition to delivering the data to the IRIS Data Management Center (DMC) for permanent archiving, the Array Network Facility (ANF) is charged with real-time data quality control, calibration, metadata storage and retrieval, network monitoring and local archiving. The Antelope real-time processing software provides the back-bone to this effort, supported by the Storage Resource Broker data replication/archiving system and the Nagios network monitoring tool. Real-time, web-based data mining, with support for multiple database schemas, is provided by an Antelope interface to both Perl and PHP scripting languages. This allows embedding of database functions in HTML. A suite of online tools allows query and graphical display of dynamic real-time sensor network parameters such as data latency, network topologies, and data return rates. Data and metadata are also web-accessible, for example XML trees of seismic data and graphical display of instrument response functions. The purpose of these tools is to provide the ANF, IRIS and end-users of USArray data with a real-time systematic method of determining data quality for the spatio-temporal area of interest. The tools are accessible at http://anf.ucsd.edu
NG43A-0436 1340h
Hidden Markov Models for Detecting Aseismic Events in Southern California
We employ a hidden Markov model (HMM) to segment surface displacement time series collection by the Southern California Integrated Geodetic Network (SCIGN). These segmented time series are then used to detect regional events by observing the number of simultaneous mode changes across the network; if a large number of stations change at the same time, that indicates an event. The hidden Markov model (HMM) approach assumes that the observed data has been generated by an unobservable dynamical statistical process. The process is of a particular form such that each observation is coincident with the system being in a particular discrete state, which is interpreted as a behavioral mode. The dynamics are the model are constructed so that the next state is directly dependent only on the current state -- it is a first order Markov process. The model is completely described by a set of parameters: the initial state probabilities, the first order Markov chain state-to-state transition probabilities, and the probability distribution of observable outputs associated with each state. The result of this approach is that our segmentation decisions are based entirely on statistical changes in the behavior of the observed daily displacements. In general, finding the optimal model parameters to fit the data is a difficult problem. We present an innovative model fitting method that is unsupervised (i.e., it requires no labeled training data) and uses a regularized version of the expectation-maximization (EM) algorithm to ensure that model solutions are both robust with respect to initial conditions and of high quality. We demonstrate the reliability of the method as compared to standard model fitting methods and show that it results in lower noise in the mode change correlation signal used to detect regional events. We compare candidate events detected by this method to the seismic record and observe that most are not correlated with a significant seismic event. Our analysis demonstrates that in the case of most events we can rule out the possibility of the event being the result of regional transients such as weather phenomena. As a result, the implication is that these regionally observed mode changes are either the result of small-scale seismic activity or of unknown episodic aseismic activity.
NG43A-0437 1340h
A Hybrid Neuro-Fuzzy Model For Integrating Large Earth-Science Datasets
A GIS-based hybrid neuro-fuzzy approach to integration of large earth-science datasets for mineral prospectivity mapping is described. It implements a Takagi-Sugeno type fuzzy inference system in the framework of a four-layered feed-forward adaptive neural network. Each unique combination of the datasets is considered a feature vector whose components are derived by knowledge-based ordinal encoding of the constituent datasets. A subset of feature vectors with a known output target vector (i.e., unique conditions known to be associated with either a mineralized or a barren location) is used for the training of an adaptive neuro-fuzzy inference system. Training involves iterative adjustment of parameters of the adaptive neuro-fuzzy inference system using a hybrid learning procedure for mapping each training vector to its output target vector with minimum sum of squared error. The trained adaptive neuro-fuzzy inference system is used to process all feature vectors. The output for each feature vector is a value that indicates the extent to which a feature vector belongs to the mineralized class or the barren class. These values are used to generate a prospectivity map. The procedure is demonstrated by an application to regional-scale base metal prospectivity mapping in a study area located in the Aravalli metallogenic province (western India). A comparison of the hybrid neuro-fuzzy approach with pure knowledge-driven fuzzy and pure data-driven neural network approaches indicates that the former offers a superior method for integrating large earth-science datasets for predictive spatial mathematical modelling.
NG43A-0438 1340h
Ocean Current Spatial Patterns from West Florida Shelf velocity Time Series Using the Self-organizing Map
A neural network analysis based on the Self-Organizing Map (SOM) is used to examine patterns of the ocean current variability on the West Florida Shelf from time series of moored velocity data that span the three-year interval from October 1999 to September 2001. Three characteristic spatial patterns are extracted in the 3$\times$4 SOM array: spatially coherent southeastward and northwestward flows with strong currents, and a transition pattern of weak currents. On synoptic weather time scale the variations of these patterns are coherent with the local winds. On seasonal time scale the variations of the patterns are coherent with both the local winds and complementary SST patterns. The currents are predominantly southeastward during winter months (from October to March) and northwestward during summer months (June through September). The spatial patterns extracted by the (nonlinear) SOM method are asymmetric, a feature that is not captured by the (linear) Empirical Orthogonal Function method. Thus, we find: (1) southeastward currents are generally stronger than northwestward currents, (2) the coastal jet axis is located further offshore for southeastward currents relative to northwestward currents, and (3) the velocity vector rotations with depth are larger when the currents are southeastward than when they are northwestward.
NG43A-0439 1340h
How to Leverage Existing Spectral Knowledge when Clustering Hyperspectral Data
Hyperspectral images collect large volumes of data, with observations at hundreds or thousands of different wavelengths. The large data size renders a thorough manual analysis difficult, expensive, and time-consuming. For example, the Hyperion instrument on the EO-1 spacecraft regularly produces image cubes that are over 1 gigabyte in size (256 $\times$ 7000 pixels, at 242 wavelengths). Automated techniques for analyzing and summarizing these mega-data sets provide two major benefits: 1) scientists can quickly obtain high-level views of the data contents, and 2) summaries produced on-board the spacecraft enable quick prioritization of data for transmission to make the best use of limited bandwidth. One approach for generating summaries is to partition the pixels from a given image into a set of $k$ clusters. Each cluster contains pixels that are more similar to each other than to pixels in other clusters. The image can be summarized by the set of $k$ clusters, represented by the cluster means and standard deviations (of the pixel values for each cluster). However, typical clustering algorithms are completely data-driven and will produce summaries based on the largest distinguishing factor between pixels (often, brightness), regardless of whether that distinction is physically meaningful. In contrast, we seek to include knowledge from existing spectral libraries to improve the automated summaries. In this work, we present a knowledge-driven clustering method that incorporates laboratory spectra as ``seeds'' for some, or all, of the data clusters. We contrast the summaries produced by data-driven and knowledge-driven clustering. We find that summaries that incorporate the spectral library have greater science value than those produced from the data alone. Knowledge-based summaries are more interpretable and more likely to be based on true compositional differences in the areas being imaged. We present sample results from several diverse areas to illustrate the benefits of clustering with spectral libraries.
NG43A-0440 1340h
Tracing Geologic Formations in Hyperspectral Imagery using Support Vector Machines
We are using hyperspectral imagery from the Hyperion instrument on NASA's Earth-Observing 1 satellite to investigate the feasibility of automated recognition of geologic formations from space. Different minerals have unique spectral signatures, therefore making it possible to distinguish them and classify them in a hyperspectral image. We report on some success and some challenges in using a multi-class Support Vector Machine to trace geologic formations in the Panamint Mountain Range in Southeastern California. This research represents the first step in the proposed development of an automated geologic mapping system capable of ingesting hyperspectral and elevation data and inferring the 4 dimensional structure and deformational history of a region.
NG43A-0441 1340h
National Archive of Marine Seismic Surveys (NAMSS): U.S. Geological Survey Program to Provide new Access to Proprietary Data
Marine seismic reflection profile data originally acquired for purposes of offshore oil and gas exploration and development within the United States Exclusive Economic Zone represent a national scientific resource of inestimable value. Although the commercial value of these data has diminished due to technological advances and offshore development moratoria, the value to current and future scientific endeavors continues to be very high. Recently, commercial owners (including WesternGeco and ChevronTexaco) of large data holdings offshore the eastern, western, and Alaskan coasts of the United States have offered to transfer over 200,000 line kilometers of two-dimensional data (vintage 1970 to 1985) to the public domain. Recognizing the value of these data, the U.S. Geological Survey in co-operation with the Institute for Crustal Studies at UCSB, the Incorporated Research Institutions for Seismology, and the American Geological Institute) is promoting efforts to safeguard on behalf of the research community and the nation any data that may otherwise be lost, and to ensure free and open access to that data. To achieve these goals, the USGS has developed a National Archive of Marine Seismic Surveys (NAMSS). Work is underway to organize and reformat digital data currently stored on obsolete media, primarily nine-track tapes. The NAMSS web site below has further information on the project, including trackline maps of surveys that will soon be publicly available. The ultimate objective is the establishment of a data repository accessible through an on-line database, with graphical and text-based search and retrieval interface.
http://walrus.wr.usgs.gov/NAMSS/
NG43A-0442 1340h
Visual Data Mining for Remote Sensing Data Sets
NASA's Earth Observing System (EOS) produces vast quantities of satellite data intended for the study of global climate change, its consequences for life on Earth, and the impact of human activities on it. However, EOS data sets are so large and complex that it is difficult, if not impossible, to understand their large scale structure and draw conclusions that could help inform the debate over climate change. We present here a prototype data mining tool for visualizing structure and exploring these massive data sets. It's use is demonstrated on data from JPL's Multi-angle Imaging SpectroRadiometer.
NG43A-0443 1340h
Interactive Collaborative Visualization in the Geosciences
Datasets in the earth sciences continue growing in size due to higher experimental resolving power, and numerical simulations at higher resolutions. Over the last several years, an increasing number of scientists have turned to visualization to represent their vast datasets in a meaningful fashion. In most cases, datasets are downloaded and then visualized on a local workstation with 2D or 3D software packages. However, it becomes inconvenient to download datasets of several gigabytes unless network bandwidth is sufficiently high (10 Gbits/sec). We are investigating the use of Virtual Network Computing (VNC) to provide interactive three-dimensional visualization services to the user community. Specialized software [1,2] enables OpenGL-based visualization software to capitalize on the hardware capabilities of modern graphics cards and transfer session information to clients through the VNC protocol. The virtue of this software is that it does not require any changes to visualization software. Session information is displayed within java applets, enabling the use of a wide variety of clients, including hand-held devices. The VNC protocol makes collaboration and interaction between multiple users possible. We demonstrate the collaborative VNC system with the commercial 3D visualization system Amira (http://www.tgs.com) and the open source VTK (http://www.vtk.org) over a 100 Mbit network. We also present ongoing work for integrating VNC within the Naradabrokering Grid environment. [1] Stegmaier, S. and Magallon, M. and T. Ertl, "A Generic Solution for Hardware-Accelerated Remote Visualization," Joint Eurographics -- IEEE TCVG Symposium on Visualization, 2002. [2] VirtualGL--3D without boundaries http://virtualgl.sourceforge.net/installation.htm
NG43A-0444 1340h
A Grid Framework for Visualization Services
Increasingly large collaborative teams, geographically distributed, coupled with experimental and numerical data sets whose size appears to increase exponentially, demands new solutions to ease data analysis, visualization, and manipulation in a transparent manner. Rather than concentrate on where and how the job gets done, users should be able to interact with their data without concern for the underlying hardware, system load, or resource availability. We address this problem through a a unique and flexible architecture, based on the NaradaBrokering (NB) middleware application program interface (API) (http://www.naradabrokering.org, [1]). We aim to support collaborative real-time remote visualization, data analysis, and video creation with an emphasis on simplicity of use. NB connects clients to services through the use of topics, rather than IP addresses and hostnames. The flexibility of this approach makes it possible to easily incorporate logging systems, redundancy, fault-tolerance, asynchronous communications, and collaboration as additional web services into the system. Furthermore, it becomes possible to insert additional services into the system such as databases and storage archives. After describing a prototype system within the context of Web-IS [2-3], we will demonstrate fault tolerance with respect to the faulty nodes in the NB network and with respect to the visualization servers, and the ability to share data. References [1] S. Pallickara and G. Fox, "NaradaBrokering: A Middleware Framework and Architec-ture for Enabling Durable Peer-to-Peer Grid", in Proceedings of ACM/IFIP/USENIX International Middleware Conference Middleware-2003. pp 41-61, (2003). [2] Y. Wang, D.A. Yuen, Z. Garbow, and G. Erlebacher, "Web-based Service of a Visualization Package ` Amira' for the Geosciences", Visual Geosciences, Springer-Verlag, (2003), [3] Php-based imaging service, view http://tomo.msi.umn.edu/~max/webis
NG43A-0445 1340h
A Grid Service for Automatic Land Cover Classification Using Hyperspectral Images
Hyperspectral images are collected using Airborne Visible/Infrared Imaging Spectrometer (Aviris) optical sensors [1]. 224 contiguous channels are measured across the spectral range, from 400 to 2500 nanometers. We present a system for the automatic classification of land cover using hyperspectral images, and propose an architecture for deploying the system in a grid environment that harnesses distributed file storage and CPU resources for the task. Originally, we ran the following data mining algorithms on a 300x300 image of a section of the Sevilleta National Wildlife Refuge in New Mexico [2]: Maximum Likelihood, Naive Bayes Classifier, Minimum Distance, and Support Vector Machine (SVM). For this, ground truth for 673 pixels was manually collected according to eight possible land covers: river, riparian, agriculture, arid upland, semi-arid upland, barren, pavement, or clouds. The classification accuracies for these algorithms were of 96.4%, 90.9%, 88.4%, and 77.6%, respectively [3]. In this study, we noticed that the slope between adjacent frequencies produces specific patterns across the whole spectrum, giving a good indication of the pixel's land cover type. Wavelet analysis makes these global patterns explicit, by breaking down the signal into variable-sized windows, where long time windows capture low-frequency information and short time windows capture high-frequency information. High frequency information translates to information among close neighbors while low frequency information displays the overall trend of the features. We pre-processed the data using different families of wavelets, resulting in an increase in the performance of the Naive Bayesian Classifier and SVM to 94.2% and 90.1%, respectively. Classification accuracy with SVM was further increased to 97.1 % by modifying the mechanism by which multi-class is achieved using basic two-class SVMs. The original winner-take-all SVM scheme was replaced with a one-against-one scheme, in which k(k-1) binary classes are trained for a k class problem . To deploy the data mining system as a grid service, we are using the Globus Toolkit to build a distributed environment over a computational grid. It is consisted of two major components: a) a service-oriented infrastructure b) a set of client tools to communicate with the service-oriented infrastructure, i.e., a web service based on the Kepler system, a visual modeling system for designing and executing scientific workflows that access distributed data and tools, and uses a semantic-mediation engine to integrate those resources [4] and the SKIDLkit data mining toolkit for high-dimensional data mining [5] to train and test new classification models. References [1] http://aviris.jpl.nasa.gov [2] http://sev.lternet.edu [3] Pennington, D., H. Jasso, P. Shin, & T. Fountain. The effect of landscape heterogeneity on classification accuracy: a comparison of classifier prediction in sub-opotimal sampling conditions. Seventh Workshop on Mining Scientific and Engineering Datasets, 2004 SIAM International Conference on Data Mining (SDM 2004), Orlando, Florida, 2004. [4] I. Altintas, C. Berkley, E. Jaeger, M. Jones, B. Ludäscher, S. Mock. Kepler: Towards a Grid-Enabled System for Scientific Workflows, In the Workflow in Grid Systems Workshop in GGF10 - The Tenth Global Grid Forum, Berlin, Germany, March 2004. [5] http://daks.sdsc.edu/skidl/skidldownloads.html
NG43A-0446 1340h
A Flexible Turbulent Vector Field Generator
Analysis and generation of turbulent vector fields is a necessity in many areas, such as Atmospheric Science. A candidate model of vector field must be flexible enough to tune some features, such as the spacial distribution of vortices, sinks and sources, according to physical measures. To achieve that goal, we propose a model that depends upon a given matricial function called "topolet" and a law of random vectors family. This model has a hierarchical structure. Its spinal column is a tree: the encoding tree of the domain where the vector field lives. The sets of vortices, sinks and sources are driven by some Bernouilli subtrees, directly giving their fractal dimension. At each node of the tree is attached a rate of energy loose giving the spectral slope. All those quantities are independantly identifiable on the base of mathematical proofs. A primitive version of this model have been proposed for generating clouds.
NG43A-0447 1340h
An optimal wavelet for the detection of surface waves in Marine Sediments
We study seismic surface wave propagation in stratified shallow marine sediments media. Our goal is to predict dynamic (shear velocity, attenuation) and physical properties (stiffness, density) of sediments from seismoacoustic records of surface waves propagating along the water-seabed interface. To estimate and invert propagational parameters of surface waves (group and phase velocity) into shear velocity as a function of distance and depth we are using a multiscale wavelet cross-correlation technique. Standard wavelet transform series has indeed proven very useful for imaging different surface waves modes. However, to achieve a better resolution of each mode imaging we need to develop a new wavelet transform that includes optimality and adaptivity, based on the seismic data itself. Our main tool to develop such an optimal wavelet is the Karhunen-Loeve decomposition of the data series. This requires two steps: first, we calculate set of covariance matrices from the pairs of time series. Second, we estimate the corresponding eigenvalues and eigenfunctions. The calculated eigenfunctions have to be further regularized to obtain a new wavelet series. This new eigenfunctions basis has an optimal convergence in the sense of the least squares. It is sufficient to take a small number of the above set of eigenfunctions. They are naturally adapted to surface waves modes propagation in terms of scales values: time and periods (frequencies). Our approach makes it possible to decompose highly correlated reference data series into eigenvectors and then to use it to decompose field data records in the frequency and time domains with significant improvement of the image quality. We have processed different seismic records with surface waves. The results were compared with the wavelet analysis using standard wavelet kernel ('Morlet', 'Gaussian', 'Mexican hat'). We show that our new developed adaptive wavelet discriminates better between different surface wave modes propagating along seabed interface and also along interfaces separating sediment layers. This raises a new potential in using wavelet analysis to study seismic wave propagation and other kinds of problems such as elecromagnetic induction.
NG43A-0448 1340h
Adaptive Multi-Resolution Data Structure for Large-Scale Visualization and Modeling of Multi-Scale Geological Processes.
Numerical modeling of geological phenomena frequently requires dealing with processes of significantly different scales. Wavelets provide a convenient and efficient approach to resolving such processes, which would be hard-pressed to be solved by conventional finite-difference techniques. The system of nonlinear partial differential equations can be solved on an adaptive grid using wavelet based algorithms. The relevant features of the obtained datasets can be efficiently extracted by wavelet-based visualization and analysis tools. However, the efficiency of visualization tools as well as the efficiency of adaptive solvers strongly depends on the access time to the large datasets. We are presenting an effective data-structure for multi-resolution adaptive grids. Tree-like structure provides rapid access to the grid nodes both in sequential and parallel implementations.
NG43A-0449 1340h
Vizualization Challenges of a Subduction Simulation Using One Billion Markers
Recent advances in supercomputing technology have permitted us to study the multiscale, multicomponent fluid dynamics of subduction zones at unprecedented resolutions down to about the length of a football field. We have performed numerical simulations using one billion tracers over a grid of about 80 thousand points in two dimensions. These runs have been performed using a thermal-chemical simulation that accounts for hydration and partial melting in the thermal, mechanical, petrological, and rheological domains. From these runs, we have observed several geophysically interesting phenomena including the development of plumes with unmixed mantle composition as well as plumes with mixed mantle/crust components. Unmixed plumes form at depths greater than 100km (5-10 km above the upper interface of subducting slab) and consist of partially molten wet peridotite. Mixed plumes form at lesser depth directly from the subducting slab and contain partially molten hydrated oceanic crust and sediments. These high resolution simulations have also spurred the development of new visualization methods. We have created a new web-based interface to data from our subduction simulation and other high-resolution 2D data that uses an hierarchical data format to achieve response times of less than one second when accessing data files on the order of 3GB. This interface, WEB-IS4, uses a Javascript and HTML frontend coupled with a C and PHP backend and allows the user to perform region of interest zooming, real-time colormap selection, and can return relevant statistics relating to the data in the region of interest.