Integrating Remotely Sensed and Ground Observations for Modeling, Analysis, and Decision Support Andrea Donnellan, Margaret Glasscoe, Jay Dennis McLeod W. Parker, Robert Granat University of Southern California Jet Propulsion Laboratory Mail Code 0781, 3651 Trousdale Parkway California Institute of Technology Los Angeles, CA 90089 4800 Oak Grove Dr. [email protected] Pasadena, CA 91109 818-354-4737 John Rundle [email protected] Department of Physics, University of [email protected] California, Davis [email protected] One Shields Avenue [email protected] Davis, CA, 95616 [email protected] Marlon Pierce, Jun Wang, Geoffrey Fox Indiana University Lisa Grant Ludwig 2719 East 10th Street University of California, Irvine Bloomington IN 47408 Irvine, CA 92697 [email protected] [email protected] [email protected] [email protected] Abstract—Earthquake science and emergency response require 5. CLOUD COMPUTING ......................................... 6 integration of many data types and models that cover a broad range of scales in time and space. Timely and efficient 6. SUMMARY ......................................................... 7 earthquake analysis and response require automated processes ACKNOWLEDGEMENTS ......................................... 7 and a system in which the interfaces between models and REFERENCES ......................................................... 7 applications are established and well defined. Geodetic imaging data provide observations of crustal deformation from BIOGRAPHIES ........................................................ 8 which strain accumulation and release associated with earthquakes can be inferred. Data products are growing and tend to be either large in size, on the order of 1 GB per image, 1. INTRODUCTION or high data rate, such as from 1 Hz GPS solution. As a result, the products can be computationally intensive to manipulate, QuakeSim is a computational environment integrating data analyze, or model. Required computing resources can be large, and models for understanding earthquake processes. even for a few users, and can spike when new data are made QuakeSim focuses on using remotely sensed geodetic available or when an earthquake occurs. Moving to a cloud imaging data to study the interseismic part of the earthquake computing environment is the natural next extension for some cycle (Figure 1). Geodetic imaging data provide components of QuakeSim as an increasing number of data observations of crustal deformation from which strain products and model applications become available to users. An accumulation and release associated with earthquakes can additional consideration is that moving large images consumes be inferred. The goal is to understand earthquakes for a tremendous amount of bandwidth. Storing the data near the model applications improves performance for the user. mitigation and response. Remotely sensed geodetic imaging observations provide information on motions of the surface Table of Contents of the Earth’s crust, expressing the cycle of strain accumulation and release. These data products are fused with geologic and seismicity information in the QuakeSim 1. INTRODUCTION ................................................. 1 environment with visualization, modeling, and pattern 2. DATA PRODUCTS .............................................. 3 analysis tools. E-DECIDER is an earthquake response 3. MODELING AND ANALYSIS ............................... 4 decision support system that also uses remotely sensed 4. DECISION SUPPORT .......................................... 5 observations and some QuakeSim components. 978-1-4673-1813-6/13/$31.00 ©2013 IEEE 1 )*#&($ 6,7,24$7255$8/57.20,9,*4$/92+,5$ %&'#&($ ,264<=32>,$.-02:-*$2*8$5/?,$ 542:-*$7-5/:-*$:9,$5,6/,5$2*8$;,.-0/:,5$ +,-.-+/0$123.45$ !"#$ %&'()*+#"#,-$ !"#$ +0'#&'"&+$01'#-$ (0'-'#&'"&+$%#20."*!0,$ &,-#.'#&'"&+$'-.*&,$*++/"/)*!0,$ Figure 1. Schematic of earthquake cycle of strain accumulation and release with the long-term tectonic rate removed. QuakeSim and E-DECIDER are downstream data product milliseconds to millions of years, and from microns to consumers that perform forward modeling, inverse global scales. Crustal deformation data products derived modeling, forecasting, event detection, and response on the from GPS and interferometric synthetic aperture radar data products. Both E-DECIDER and QuakeSim rely on (InSAR) observations enable an understanding of the cycle processed data products, which are typically velocities and of strain accumulation and release. Geological and seismic time series from GPS, and ground range changes from data complement the spaceborne observations. Various Interferometric Synthetic Aperture Radar (InSAR). These domain experts study aspects of these processes using and other data products such as seismicity, geologic fault experimental, theoretical, modeling, or statistical tools. Over data, or civil infrastructure, are integrated and assimilated the last decade, in part due to the advent of new into models for understanding, forecasting, and response. As computational methods, there has been an increasing focus a result of this reliance on data product providers, it is on mining and integrating the heterogeneous data products imperative that data specifications, standards, and locations into complex models of solid Earth processes. are well understood. QuakeSim focuses on the cycle of strain accumulation and QuakeSim is distributed and heterogeneous because of the release that is well addressed by geodetic imaging data and many different data products and numerous different related model applications. Coseismic offset is the slip that applications. Decomposing QuakeSim into its different occurs during an earthquake. Postseismic deformation is parts, strategically using cloud resources where necessary, continued motion that occurs following an earthquake and and using other resources as practical result in greater deviates from the long-term slip rate. Postseismic efficiency. In cases where data volumes are large, such as deformation is usually some combination of afterslip, InSAR, it is important to ensure that model applications relaxation, or poroelastic rebound. Interseismic strain reside close to the data products to reduce latency from accumulation occurs as a result of plate tectonic motion. bandwidth limitations. Alternatively, data products can be QuakeSim addresses the size and location of events, connected to applications through high bandwidth coseismic offsets associated with the earthquakes, and the pathways. time varying displacements that occur throughout the earthquake cycle. Seismicity and geology yield information Earthquakes are the result of complex solid Earth processes. on size and location of events. GPS provides position time Plate tectonics drive motion of the Earth’s crust, series. Interferometric synthetic aperture radar from airborne accumulating deformation primarily along plate boundaries. and spaceborne platforms provide images of displacement. Strain stored as elastic energy is released in earthquakes All data sources provide information on geometry of faults along faults. These processes occur on scales of and time varying slip. 978-1-4673-1813-6/13/$31.00 ©2013 IEEE 2 !"#"$%&'()*#+$ ,--./*"0'1+$ 2)#-)#+$ !"#$%&'"() 3").#+$ 9+#-%"$&;(6)+*"1)0& !"#$%&'()*(%+,(-& *+%',-.+)/012%+3-#,(4) !"#$%&-$,.&+"%(-& #5+%32)&363)('6(4) /%+",0&*,'+"1)0& %,7'""'&)7"#8)8#&'2() 2"+%34#"5(&6"#$%&-<-%(*& 914,:;<,=4,:$ -,*#$"%)+-& >!?4)=*C) 2"+%34#"5(&6"#$%&7(3"8,)+& :">$("#"$ :5+-'1+5$ 8%4$ >!?4)*@!=;4)9<=A;@4) >,*(&-(+,(-&"0"$<-,-& 9+#-%"$&:(6)+*"1)0& 9<B) "0)*"$,(-& 45/+6/*/#7$ 9*:*4)*;*<4)<;*<4) =<**) !)+(="-%& 2"+%34#"5(&$,5($,3)):& Figure 2. QuakeSim inputs, outputs, and feedback to data providers are shown. Multiple data products are used from numerous providers. These are used in applications for understanding faults and forecasting earthquake likelihood. At times output from the applications identifies issues with the data products, which can be in the form of previously unrecognized faults, unmodeled error sources, or processing issues. While we use seismicity as a data source for size, location, data products that have typically been generated elsewhere; and mechanism of events, we do not focus on seismic raw data is not ingested or processed but higher level data waveforms. Understanding crustal deformation and fault products from other data providers are made available behavior leads to improved forecasting, emergency through Open Geospatial Consortium (OGC)-compliant planning, and disaster response. Analysis of crustal web services (Donnellan, et al., 2012, Wang, et al., 2012). deformation data can be used to indicate the existence of otherwise unknown faults (e.g., Donnellan et al., 1993), 2. DATA PRODUCTS particularly when they are not exposed at the surface. Accurate fault models are required for understanding QuakeSim applications use multiple data products from earthquake processes and require the integration of multiple multiple providers (Figure 2). Data products include faults, types of data. Identifying, characterizing, modeling and GPS position time series and velocities, Radar repeat pass considering the consequences of unknown faults, and interferometry (RPI) or Interferometric Synthetic Aperture improving the models of known faults contributes to seismic Radar (InSAR) ground range change images, and seismicity. risk mitigation. It is important that the data products come from trusted The goal of QuakeSim