SCIRAD - Data Technologies - Cyberinfrastructure Laboratory for Environmental Observing Systems

CLEOS LAB NEWS

CLEOS releases USArray SOA for Observing System Instrument Management technical report

La Jolla, CA. April 14, 2006. A technical report has been released based on CLEOS work to implement an SOA-based strategy for instrument management in environmental observing systems. This work is being done within the context of the EarthScope USArray seismic observing system.

Paper link: http://www.sdsc.edu/techreports/SDSC-TR-2006-6-USArray.pdf

CLEOS demos instrument management tools at CREON & GLEON workshop

Townsville, Australia. March 29-31, 2006. The CLEOS group successfully demonstrated cyberinfrastructure products and tools for instrument management within the context of the USArray seismic monitoring applications, and the North Temperate Lakes observatory. The demo included the CLEOS observing system cyberdashboard that automates many of the processes required for sensor deployment and state-of-health monitoring. The cyberdashboard includes a suite of web services for interfacing with sensor network middleware and a user interface composed of GridSphere portal and GoogleEarth technologies. Working sessions in Australia focused on extending the CI tools and applications to address additional instrument management tasks and observing system communities.

Workshop webpage: http://www.coralreefeon.org/workshop.cfm

CLEOS presence at CEOA Symposium on Science & Technology in GEOSS at Calit2

La Jolla, Ca. November, 20-22, 2005. The CLEOS group presented six posters and a live demo of the cyberdashboard for observing system management, tailored for the USArray seismic observing system, at the CEOA Symposium. The cyberdashboard integrates several CLEOS cyberinfrastructure products with Google Earth (tm) to create rich and powerful management tool. The mission of the Center for Earth Observations and Applications (CEOA) is to stimulate, support and coordinate research and applications in Earth observations.

CEOA website: http://ceoa.ucsd.edu

SKIDL becomes CLEOS

San Diego, CA, August 30, 2005. SKIDL is officially reborn as CLEOS: Cyberinfrastructure Laboratory for Environmental Observing Systems. SKIDL has been a participant in a variety of observing system projects over the past several years, and has amassed a wealth of experience both on the analytical side and the architectural/infrastructure side. SKIDL developed machine learning methods to analyze sensor data, as well as applied open standards technologies such as web services and service oriented architectures to observing systems applications. Thus, CLEOS is being formed to explicitly support the design and evolution of cyberinfrastructure for sensor network based observing systems in a variety of scientific domains.

CLEOS poster on Command and Control Web Services for Sensor Networks presented at GEON All-hands Meeting

San Diego, CA, May 5-6, 2005. CLEOS poster on Command and Control Web Services for Sensor Networks presented at GEON All-hands Meeting. The poster is based on the services suite being prototyped by the CLEOS team within the initial context of the EarthScope US Array project. Scalable means of command and control are critical for large-scale, evolving observing systems. Even after initial deployments, sensing hardware must be replaced or reconfigured to accommodate specific experiments or observable phenomena. Such relatively straight-forward tasks rapidly become more complex and daunting as the number of sensors, or remote sensing sites in a sensor network increases. Workflows can then be constructed, either as a series of dialogues within a web portal, or in a workflow authoring tool, to encapsulate the common observing system administration tasks. The initial set of CLEOS command & control web services supports the adding of new data loggers, defining and starting/stopping new worker tasks, and other status and configuration actions for the sensor network middleware system being used by the US Array project.

Conference website: http://www.geongrid.org/AM05/
Poster can be downloaded here: (PDF, 4MB)

EarthScope US Array website: http://anf.ucsd.edu/

CLEOS Hyperspectral poster accepted to AGU 2004 Conference

San Francisco, CA, December 13-17, 2004. CLEOS Hyperspectral Landcover Classification poster displayed at American Geophysical Union (AGU) Fall Meeting. The poster is based on work in various areas related to automatic landcover classification using hyperspectral data. Tasks include data collection, cleaning, processing using wavelets, and publishing grid-enabled classification services. One of the findings related to the classification task is that the use of wavelet preprocessing enhances the accuracy by explicitly capturing neighboring information among features.

Conference website: http://www.agu.org/meetings/fm04/
Poster can be downloaded here: (PDF, 11MB) (PNG, 2MB)


CLEOS collaborates with ROADNet, demonstrates Web Services for Sensor Networks at workshop

21 September, 2004. Close collaboration with the ROADNet team helps CLEOS prototype their latest Web Services for Sensor Networks and integrate them into the Antelope real-time monitoring system. Several web services were built and demonstrated at the Environmental Sensor Networks for Research and Education workshop held at Scripps Institute of Oceanography. These web services were designed to serve as building blocks for intelligent workflows involving sensor networks. Both preliminary analysis and control web services were demonstrated. This effort was motivated by the TeraBridge project: an NSF-funded ITR collaborative research project that focuses on sensor-based monitoring of the structural health of bridges and civil infrastructure. For the demo, live structural sensor data was streaming from an instrumented composite deck on a surface street at the UCSD campus, through the Antelope system, and acted upon by the web services demonstrated. The wealth of knowledge and first-hand experience provided by the ROADNet team in dealing with sensor networks, and the Antelope system in particular, proved invaluable to the integration of web services command, control, and analysis functions to the underlying sensor network infrastructure.

ROADNet Homepage

TeraBridge Homepage

New SKIDLkit data mining toolkit version released

August 2004. The latest version of the SKIDLkit data mining toolkit is now online and available for download. The toolkit includes routines for selecting key features in high dimensional datasets to classify a target variable in new samples. Updated version includes user access to feature scores and statistics for the filter method, as well as updated versions of the underlying machine learning algorithms. See downloads page for more details.

Downloads Page

CLEOS Machine Learning Algorithms Comparison Paper Accepted to SIAM 2004 Data Mining Conference

Lake Buena Vista, FL, Apr. 24, 2004. The CLEOS team has a paper accepted to and attends the 2004 SIAM International Conference on Data Mining. The paper is summarized as follows.

"Sub-optimal sampling conditions like autocorrelations and mislabelings represent a challenge for the automatic classification of landscapes using hyperspectral data. This paper compares algorithm performance of four machine learning algorithms (Support Vector Machines (SVMs), Naive Bayesian Networks, Maximum Likelihood and Minimum Distance) on limited and full training sample sets of the Sevilleta National Wildlife Refuge in New Mexico, and shows the importance of complementing algorithm accuracy measures based on confusion matrices with expert domain knowledge."

CLEOS Makes Headway on the Collaborative Lakes Metabolism Project

April 2004. CLEOS is proud to be a part of the growing international collaboration known as the Lakes Metabolism Project. The project is summarized as follows.

Lake metabolism is a fundamental lake characteristic that helps describe the source of carbon incorporated into all trophic levels of the ecosystem. The project is an international collaboration within the ILTER community to further understand lake metabolism by comparing Trout Lake in US and Yuan Yang Lake in Taiwan. In addition, the project will put in place an infrastructure for exchange of data, with near real-time data from sensors via wireless connections to the databases. Furthermore, the data infrastructure, once demonstrated and in place, has been designed to scale to include other lake systems with sensors with a minimum of reconfiguration.

Lake Metabolism Project Homepage

CLEOS Unveils I2G at LTER All Scientists Meeting 2003

Seattle, WA, Sep. 20, 2003. The CLEOS team introduces the International Information Grid for Ecology and the Environment (I2G) at the Long Term Ecological Research Network (LTER) All Scientists Meeting 2003. I2G is a standard web services infrastructure for integrating the geographically distributed LTER and PRAGMA resources. It addresses the dynamic cross-site data synthesis and analysis issues faced by these research communities. Partners include LTER, PRAGMA, GEON et al.
Click here to download poster (pdf, 5MB).

SKIDLkit data mining toolkit v1.0 released

May 28, 2003: The initial version of the SKIDLkit data mining toolkit is now online and available for download. The toolkit includes routines for selecting key features in high dimensional datasets to classify a target variable in new samples. See downloads page for more details.

CLEOS Presence at NPACI All Hands Meeting 2003

March 18: Tony Fountain, aided by Dave Archbell, gave the Data Mining for Scientific Applications tutorial again this year to a packed crowd. Many relevant techniques and examples were presented, complete with guest speakers from the various CLEOS collaborations.

March 19: Tony Fountain also presented the Spatial Data Workbench at the Collection Management session.

CLEOS debuts SDW at SC2002

Baltimore, Nov. 19, 2002. The CLEOS team demos the Spatial Data Workbench (SDW) at Supercomputing 2002. The SDW is a tool for spatio-temporal data management and mining for ecology. The SDW brings together Imagery Collection Management, Processing pipelines enabling high throughput computational ecology, and a Web services architecture for distributed ecological data analysis. Partners include LTER Network and NACSE.