Current Trends and Demands in Visualization in the Geosciences

Curre nt T re nds a nd De m a nds in V isua liza t ion in t he Ge osc ie nc e s Gordon Erle ba c he r De pt . of M a t he m a t ic s Florida St a t e U nive rsit y T a lla ha sse e , FL 3 2 3 0 6 -4 1 2 0 Da vid A. Y ue n Fa bie n Dubuffe t M inne sot a Supe rc om put e r I nst it ut e a nd De pt . of Ge ology a nd Ge ophysic s U niv. M inne sot a M inne a polis, M N 5 5 4 1 5 -1 2 2 0 To appear in Electronic Geosciences, also on www.msi.umn.edu/~heather then click on the left electrongeo ABST RACT Geosciences, along with many other disciplines in science and engineering, faces an exponential increase in the amount of data generated from observation, experiment and large-scale, high-resolution 3-D numerical simulations. In this communication we describe the fundamentals of visualization necessary to meet these challenges. We present several alternative methodologies such as 2D/3D feature extraction, segmentation methods, and flow topology, to help better understand the physical structure of the data. We use AMIRA from TGS to demonstrate our concepts. Examples are drawn from fields in computational fluid dynamics, 3-D mantle convection and seismic tomography. Finally, we present our perspective on the future of visualization. 11/16/01 Electronic Geosciences 2 I nt roduc t ion Today in the geosciences we are facing a very serious problem of data flooding from very large data sets produced by very high resolution numerical simulations , improved laboratory and experimental instrumentation, in particular of terrain data gathered by satellites. For instance, a state-of-the-art simulation in global atmospheric circulation, mantle convection, geodynamo, or 3-D earthquake modelling can easily generate up to one to several gigabytes of data per time-step, because of the many variables involved. Terabytes of data can be produced each day from highresolution SAR data satellite coverage.. It is quite clear that visualizing the data at full resolution is not a viable option. In this communication we will discuss some alternative routes to visualizing large data sets, not in full, but by extracting and highlighting its salient features, using currently available visualization packages. Software ) can be employed to visualize high-resolution 3-D mantle convection and seismic tomography. Finally, we address the criteria for a successful visualization system and comment on the future of visualization and its growing multidisciplinary character. Following the spirit of an online electronic journal, we will present our ideas in the condensed format of overheads, where the main ideas are encapsulated in bullet format, rather in long paragraphs with complex wordy prose. Our main objective is to find and reach an enthusiastic and eager audience, who will avail themselves of the ideas presented herein and will begin to apply these concepts to their own research and /or educational endeavors. First, we will discuss some of the fundamental paradigms in visualization. This will be followed by a description of an ideal visualization system suited to analysis and comprehension of complex data sets drawn from diverse disciplines, such as mantle convection, seismic tomography, supersonic flows around aircraft and 3-D biological structures. We will then discussed how the volume-rendering algorithm implemented in the commercial package AMIRA ( from Template Graphics 11/16/01 Electronic Geosciences 3 We w ill … Ø Discuss general visualization principles Ø Some features of particular commercial visualization package: Amira Ø Visualization using Amira We w ill not … Ø Discuss specific algorithms Ø Compare different visualization packages 11/16/01 Electronic Geosciences 4 Pe rsona l ba c k ground Gordon Erle ba c he r Ø I have conducted research in Fluid Dynamics and scientific visualization § Simulations of compressible transition and turbulence § Turbulence modeling § Numerical algorithms Ø Work in Scientific Visualization § Vector fields § I nteractivity § Distributed visualization 11/16/01 Electronic Geosciences 5 Pe rsona l ba c k ground Da vid A. Y ue n a nd Fa bie n Dubuffe t Ø We are geophysicists Ø We conducted large-scale numerical simulations Ø We are interested in the use of state-ofart techniques to help visualize and understand 3-D numerical simulations in the Earth’s mantle 11/16/01 Electronic Geosciences 6 Possible Re se a rc h T opic s in V isua liza t ion Ø Visualization of time-dependent motion Ø Change of topology Ø I nteractive feature extraction Ø I nteractive exploration Ø Use of force feedback in visualization Ø Handling of Multi-Gigabyte datasets Ø Exploration of high-dimensional spaces 11/16/01 Electronic Geosciences 7 Why V isua lize Da t a Ø Numerical simulations and experiments produce extremely large datasets Ø The size of these datasets are increasing exponentially fast Ø Numerical output (e.g., tables) does not lend itself to easy comprehension Ø We need dynamical display of the fields for unravelling new physics. 11/16/01 Electronic Geosciences 8 Sc ie nt ific V isua liza t ion Ge ne ra l Princ iple s Ø Maximize comprehension Ø Maximize information Ø Maximize accuracy Ø Minimize clutter Ø Maximize interactivity Ø I ndependence of underlying meshing Ø Minimize program response time 11/16/01 Electronic Geosciences 9 Sc ie nt ific V isua liza t ion Ø Extract from large datasets more meaningful components (called data extracts) § I sosurface, streamlines, streaklines, vector field topology, vortex tubes, cracks, fault lines, etc. § Sedimentation layers, free-surfaces, edge and surface extraction Ø Render this data with comprehension of the physics in mind, as opposed to visual realism 11/16/01 Electronic Geosciences 10 Com put e r Gra phic s M ode ling input input Camera Modeling Geometric Models input Light Modeling Rendering output input Image Storage and Display Animation Parameters Textures input 11/16/01 Electronic Geosciences 11 T a x onom y Ø Dimensionality of domain § 1D (x) à 4D (x,y,z,t) § N-D (e.g., phylogeny) Ø Dimensionality of range R n § Scalar, vector, tensor fields Ø Domain connectivity § (Un)Structured, points, graphs 11/16/01 Electronic Geosciences 12 Dom a in Conne c t ivit y (2 D) Cartesian Tree 11/16/01 Curvilinear Unstructured Graph Electronic Geosciences 13 Aux ilia ry V e rt e x Da t a Ø Coordinates (2,3,or 4) Ø Color Ø Normals (for lighting) Ø Temperature, conductivity, viscosity, etc. Aux ilia ry Edge Da t a Ø Heat Flux, forces 11/16/01 Electronic Geosciences 14 V isua liza t ion in Ge osc ie nc e s Com pone nt s Ø Vector fields (velocity) Ø Gradient fields (temperature gradient) Ø Tensor fields (stress, strain-rate, anisotropic energy spectra, momentum flux) Ø Multiple scales in time and space Ø Multi-domain, curvilinear and tetrahedral grids Ø Time dependent structures (plumes, crack propagation) Ø I nteractive data exploration in space and time 11/16/01 Electronic Geosciences 15 V isua liza t ion H a rdw a re De sire d Fe a t ure s Ø Large framebuffer memory § Double buffering (smooth animation) § Stereo (left/ right buffers) § Z-buffering (hidden line removal) x 1600x1200 frame with 32 bit color: 7.7 Mbytes x Double buffering: 15 Mbytes x Stereo: 30 Mbytes x Even higher with alpha, stencil, z-buffers § Large texture memory x 11/16/01 Used by many modern visualization algorithms Electronic Geosciences 16 For $ 3 ,7 0 0 , N ove m be r, 2001 … Ø Ø Ø Ø Ø Ø Ø Dell Precision Workstation 530 Dual pentium: 1.7 Ghz cpu 1 Gbyte memory (400 Mhz) 21 inch screen 80 Gbyte disk Read/ Write CD-rom (read/ write DVD is better) Quadro2-Pro graphics card (can not handle dual monitors and stereo) Ø Linux/ Windows X 11/16/01 Electronic Geosciences 17 I m m e rsive Environm e nt s Ø PowerWalls and other large-scale displays § (PI CTURE) 6’x8’ and larger § Rear or front projection § Enables 5-20 people to view and interact with the data simultaneously. § Only one person controls the interaction Ø Caves § Project in stereo onto 5 or 6 walls § Provides realistic display of data § Users interact using wands and other devices (one at a time) 11/16/01 Electronic Geosciences 18 V isua liza t ion Algorit hm s Ra nge T ype s Ø Scalar § I socontours (2D), isosurface (3D) § Volume rendering Ø Vector § § § § Streamlines, pathlines, streaklines Line integral convolution (steady state) LEA (Lagrangian-Eulerian Advection (time-dependent) Critical points, vector field topology Ø Tensor § Tensor field topology (symmetric and antisymmetric tensors) § Hyperstreamlines: streamlines along dominant eigenvector, ellipsoidal cross-section normal to the streamline, determined by other two eigenvalues 11/16/01 Electronic Geosciences 19 V isua liza t ion Algorit hm Cha lle nge s Ø Strike balance between § High- resolution versus interactive speed Ø How to do time-dependent visualization Ø Describe and view change of data topology § Vector and scalar fields § Tensor fields Ø How to navigate a Terabyte dataset? 11/16/01 Electronic Geosciences 20 V olum e Re nde ring Ø I t is often difficult to choose isosurface values that produce meaningful surfaces Ø More often, it is a collection of isosurfaces that is desired Ø Need global techniques for complex datasets Ø I n the physical world § x-rays, translucent medium Ø Solution § consider points of the physical domain as emitters and absorbers of light § Composite points along rays through the volume to produce final image 11/16/01 Electronic Geosciences 21 V olum e Re nde ring Ray casting Texture compositing Screen Screen 11/16/01 Electronic Geosciences 22 API s, Pa c k a ge s, T oolk it s Ø Low Level Graphic API s (OpenGL, Direct8X) Ø Visualization API s (Open I nventor) Ø Visual I nterfaces (Ensight, LightView) Ø Flowcharting (OpenDX, I ris Explorer, Amira) Ø Visualization Toolkits (VTK, NCAR) Ø Free specialized Solutions (Vis5d) Ø Commercial specialized solutions (Rivertools, …) 11/16/01 Electronic Geosciences 23 Am ira (from T GS) 11/16/01 Electronic Geosciences 24 Am ira Fe a t ure s Ø Flowcharts are created interactively by the user Ø Each component has an associated user interface Ø Software algorithms harness the latest graphic hardware (SGI , Nvidia, ATI ) to achieve good performance Ø Flowcharts, called networks, can be saved for re-use Ø Developer version allows users to create their own modules for specialized visualization Ø The user interface is based on Qt (free for academic use); portable on wide array of architectures (including PDA) 11/16/01 Electronic Geosciences 25 Am ira Fe a t ure s Ø Very I nteractive Ø Manipulators § I nteract with the data Ø Extensible § Users can write extension modules § API is very sophisticated Ø Highly advanced algorithms § I sosurface, volume rendering, vector visualization, ima § Combinations of the above 11/16/01 Electronic Geosciences 26 Ex a m ple Ø Read in 3D file Ø Generate several planar cross-sections Ø Generate an iso-surface Ø Generate a volumetric plot Ø Combine techniques Ø Query data (e.g., line cut, point probe) 11/16/01 Electronic Geosciences 27 Ex a m ple s 11/16/01 Electronic Geosciences 28 pow(x,3)+pow(y,3)-3*x*y+x*z+2*y*z*x Opaque isosurfaces 11/16/01 Electronic Geosciences 29 pow(x,3)+pow(y,3)-3*x*y+x*z+2*y*z*x Transparent isosurfaces 11/16/01 Electronic Geosciences 30 FEM a nd 3 D da t a visua liza t ion Visualized with Amira Vector Fields: (courtesy TGS) • illuminated field lines 11/16/01 Electronic Geosciences 31 3 -D m a nt le c onve c t ion 2573 dataset 643 subsampling Volume rendering of temperature fields 11/16/01 Electronic Geosciences 32 Adiabatic heating distribution 11/16/01 Electronic Geosciences 33 Local state of adiabaticity in convection 11/16/01 Electronic Geosciences 34 Viscous heating distribution 11/16/01 Electronic Geosciences 35 Viscous heating and illuminated streamlines 11/16/01 Electronic Geosciences 36 Convection looking upward from the bottom. Isothermal surface with planforms illuminated 11/16/01 Electronic Geosciences 37 Isothermal surface with planforms 11/16/01 Electronic Geosciences 38 Thermal fields volume-rendered with BOB 11/16/01 Electronic Geosciences 39 3-D tomographic slice at 250 km depth Taken from Zhao, EPSL,2001 11/16/01 Electronic Geosciences 40 Zoom-in view of the whole mantle under Japan 11/16/01 Electronic Geosciences 41 Integration of time-dependent vector field animations within Amira using developer version 11/16/01 Electronic Geosciences 42 Interaction of a shock with a longitudinal vortex Pressure isosurface and contours of pressure gradient magnitude and Mach number. 11/16/01 Electronic Geosciences 43 Interaction of a shock and a longitudinal vortex (side view). Pressure isosurface, velocity vectors. 11/16/01 Electronic Geosciences 44 Interaction of a shock with a vortex ring. Pressure isosurface, contours of density gradient magnitude and Mach number. Data: Ding, Hussaini, Erlebacher 11/16/01 Electronic Geosciences 45 Shock interaction with an axisymmetric vortex Data: Ding, Hussaini, Erlebacher 11/16/01 Electronic Geosciences 46 Com put a t iona l St e e ring a nd Co-proc e ssing Ø Couple numerical simulations with scientific visualization Ø Drill down of image for data querying (i.e., visualizing metadata or underlying raw data) Ø Raw data is often not on client: need robust client/ server communication Ø Would like to query a running simulation and change its parameters (e.g., PV3, Cumulus, SciRun, etc.) 11/16/01 Electronic Geosciences 47 Fut ure of V isua liza t ion Ø Visualization is/ has become multidisciplinary, involves many fields. Ø Successful visualization system must address § § § § § § I/O Maintainability Flexibility (e.g., using plugins) Accessibility (low cost and easy to use/ install) Robust Standardization Ø The above features are not consistent with each other 11/16/01 Electronic Geosciences 48 V isua liza t ion U biquit y Ø Collaboration at a distance through visualization Ø Office walls or ceilings become visualization displays (E-I nk: thin, pliable medium capable of electronic encoding) Ø Exchange of visual data becomes as ubiquitous as exchange of text documents and graphics in 2001 11/16/01 Electronic Geosciences 49 An I de a l V isua liza t ion Syst e m Ø Reusable modules Ø Flexible Ø Ease of use Ø Low memory footprint Ø Extensible Ø Scriptable Ø Good debugging 11/16/01 Ø Portable Ø I ntelligent defaults Ø Changeable defaults Ø I nterpreted and compiled modes Ø Novice and expert modes Ø Mathematical text editor Electronic Geosciences 50 Grid Se rvic e s (Fox e t a l. 2 0 0 1 ) Ø Collaborative Portal § XML-based § Secure Ø Coupling of § § § § § § 11/16/01 Multi-scale numerical simulations / observational data 4D space-time domain (visualization) Data mining Efficient I / O mechanisms Computational Steering Databases Electronic Geosciences 51 Se le c t e d Re fe re nc e s 11/16/01 Electronic Geosciences 52 Geosciences Balachandar, S., Yuen, D.A., and D. Reuteler, Viscous and adiabatic heating effects in three-dimensional compressible convection at infinite Prandtl number, Phys. Fluids, A vol.5 (11), 2938- 2945, 1993. Jordan, K.E., Yuen, D.A., Reuteler, D.M., Zhang, S. and R. Haimes, Parallel interactive visualization of 3D mantle convection, I EEE Computational Science and Engineering, Vol. 3 , No. 4, 29 - 37, 1996. Matyska, C. and D.A. Yuen, Are mantle plumes adiabatic? Earth Planet. Sci. Lett., 189, 165- 176, 2001. Yuen, D.A., Balachandar, S. and U. Hansen, Modeling mantle convection: A significant challenge in geophysical fluid dynamics, in Geophysical and Astrophysical Convection, , edited by P.A. Fox and R.M. Kerr, pp 257- 293, Gordon and Breach Publishers, 2000. Zhao, D., “Seismic Structure and Origin of Hotspots and Mantle Plume”, Earth Planet. Sci. Lett., 192, 251-265, 2001. 11/16/01 Electronic Geosciences 53 Shock I nteractions Ding, Z., Hussaini, M. Y., Erlebacher, G., and Krothapalli, G. A., "Self-I nteraction of Acoustic Wave due to Shock/ Vortex I nteraction," AI AA Journal, Vol. 38, No. 6, pp. 1002-1009. Erlebacher, G., Hussaini, M. Y., and Shu, C.-W., "I nteraction of a shock with a longitudinal vortex," Journal Fluid Mechanics, Vol. 337, 1997, pp. 129- 153. 11/16/01 Electronic Geosciences 54 Vector and Tensor Field Visualization Cabral, B. and Leedom , L. C., "I maging Vector Fields Using Line I ntegral Convolution," Computer Graphics Proceedings, ACM, 1993, pp. 263- 269. Forssell, L. K. and Cohen, S. D., "Using Line I ntegral Convolution for Flow Visualization: Curvilinear Grids, Variable-Speed Animation, and Unsteady Flows," I EEE Transactions on Visualization and Computer Graphics, Vol. 1, No. 2, 1995, pp. 133-141. Helman, J. L. and Hesselink, L., "Representation and Display of Vector Field Topology in Fluid Flow Data Sets," Visualization in Scientific Computing 1989. Jobard, B., Erlebacher, G., and Hussaini, M. Y., "Lagrangian-Eulerian Advection for Unsteady Flow Visualization," Proceedings I EEE Visualization 2001, I EEE Computer Society, New York, 2001. Stalling, D. and Hege, H.-C., "Fast and Resolution I ndependent Line I ntegral Convolution," Proceedings of SI GGRAPH '95, 1995, pp. 249- 256. Tricoche, X., Scheuermann, G., and Hagen, H., "Topology-Based Visualization of Time-Dependent 2D Vector Fields,". 11/16/01 Electronic Geosciences 55 Volumetric techniques Kniss, J., Kindlmann, G., and Hansen, C., "I nteractive volume rendering using multidimensional transfer functions and direct manipulation widgets," Proceedings I EEE Visualization 2001, I EEE Computer Society, New York, 2001. Lichtenbelt, B., Crane, R., and Naqvi, S., I ntroduction to volume rendering, Prentice Hall, New Jersey, 1998. Lum, E. B., Ma, K.-L., and Clyne, J., "Texture Hardware Assisted Rendering of TimeVarying Volume Data," Proceedings Visualization 2001, I EEE Computer Society, New York, 2001. Levoy, M. S., "Display of Surfaces from Volume Data," Ph.D., Chapel Hill University, 1989, 91 pages. Max, N. L., "Optical methods for direct volume rendering," I EEE Transactions on Visualization and Computer Graphics, Vol. 1, No. 2, 1995, pp. 99-108. 11/16/01 Electronic Geosciences 56 Computational Steering and Grid Services Fox, G. C., Ko, S.-H., Pierce, M., Basloy, O., Kim, J., Lee, S., Kim, K., Oh, S., Rao, X., Varank, M., Bulut, H., Gunduz, G., Qiu, X., Pallickara, S., Uyar, A., and Youn, C., "Grid Services for Earthquake Science," ACES 2001: Special I ssue of Concurrency and Computation: Practice and Experience, 2001. Haimes, R., PV3: A distributed system for large-scale unsteady CFD visualization. AI AA paper 94-0321. 1994. AI AA. Parker, Steven G. and Johnson, Christopher R., "SCI Run: A scientific programming environment for computational steering", http: / / www.sci.utah.edu/ publications/ sc95_pj/ PARKER_JOHNSON.html. Swann, J. Edward, Lanzaborta, Marco, Maxwell, Doug, Kuo, Eddy, Uhlmann, Jeff, Anderson, Wendell, Shyu, Haw-Jye, and Smith, William, "A Computational Steering System for Studying Microwave I nteractions with Missile Bodies", http: / / www.ait.nrl.navy.mil/ vrlab/ projects/ CompSteering/ Vis_00_Comp_Steering_Slid es.pdf. 11/16/01 Electronic Geosciences 57 Useful Links "OpenDX", www.opendx.org. "Template Graphics Software", www.tgs.com. The Visualization Toolkit", http: / / public.kitware.com/ VTK. 11/16/01 Electronic Geosciences 58 Acknowledgements We would like to thank discussions with Geoffrey C. Fox. This work was supported by the NSF visualization grant NSF-0083792 and by the geophysics ( D.A.Y.) and information science (G.E.) programs of the National Science Foundation and the complex fluids program of the Dept. of Energy. Fabien Dubuffet has been supported by the Visiting Scholar program of the Minnesota Supercomputing Institute. Finally we thank ACES, GEM and SCEC , earthquake research organizations for sponsoring Maui workshop, which laid the seed for this interdisciplinary collaborative effort. 11/16/01 Electronic Geosciences 59