Earth Systems Models Will Benefit
from Remote Sensing Data Archives

PROJECT LEADERS
Joseph Já Já, Director of UMIACS, Professor of Electrical Engineering Affiliate Professor with Computer Science, University of Maryland
John Townshend, Professor, Department of Geography, University of Maryland

atellites with cameras trained on the Earth's surface capture a continuous stream of data for Earth systems scientists. The process of collecting data by satellites, called remote sensing, generates data that researchers can use to monitor changes in the Earth's vegetation, such as loss of tropical rain forests or the spread of deserts. However, current computing technology cannot easily handle all the data that has been archived since the early 1980s. NPACI scientists are taking the first steps to piece together the tools for storing, managing, and analyzing repositories of remote-sensing data to paint a complete picture of the changing planet.

Geographers and computer scientists at the University of Maryland are leading a project to build just such a remote-sensing data tool--for the Web. Behind the Web page interface, the prototype tool, called KRONOS, will access satellite-collected data sets and analyze land cover information from the images on high-performance computers before presenting them to the Web browser.

DATA-INTENSIVE QUESTIONS

GLOBAL INSIGHTS

DATA-INTENSIVE QUESTIONS

"KRONOS will allow users to generate custom-tailored products from large data repositories," said Joseph Já Já, director of the University of Maryland Institute for Advanced Computer Studies (UMIACS) and co-leader of the NASA-funded Land Cover Earth System Information Partnership (ESIP) at UMIACS. "This is the main emphasis of our NPACI work, and the ESIP work fits beautifully with what we are trying to do with NPACI."

The ESIP land cover project is one of 12 ESIP-2 projects funded by NASA to develop computational tools for high-end Earth systems science research. Building on an earlier Grand Challenge project on land cover dynamics, the ESIP effort is at a more advanced stage of software development, with a better interface, additional visualization tools, and hierarchical indexing of data. NPACI participation is enabling the project to turn the prototype into production-level software.

NPACI participation is also encouraging collaborations between the KRONOS developers and other thrust areas. First, KRONOS will be integrated with the Active Data Repository (ADR), an NPACI-supported Programming Tools and Environments project at UMIACS. The ADR project, led by Joel Saltz, is developing a customizable database system that integrates storage, retrieval and processing of sparse, irregular and multi-resolution, multi-dimensional data sets. The ADR's processing can employ high-performance computers to minimize the amount of data returned across the network, making it ideally suited to the large data sets, Web-based client, and custom processing of KRONOS.

"The user can specify the processing chain they need," Já Já said, "as well as the types of data, the surface area, and the format of the output image, including an overlay of layers from multi-channel data."

In a second collaboration, KRONOS will use distributed NPACI data caches, including a large data cache being installed at UMIACS. Partly funded through NPACI's Data-intensive Computing thrust area, the cache has 260 gigabytes of disk and 10 terabytes of tape storage controlled by an IBM SP. Já Já said the two NPACI collaborations, with ADR and with distributed data caches, will be demonstrated at SC98 in Orlando, Florida.

GLOBAL INSIGHTS

"With the ESIP client, we are also aiming at a broader audience--government labs, high school students, and the general public," Já Já said. "For example, one data set available on the Web site shows the health of U.S. coastal wetlands, which students and others may want to look at in relation to environmental issues."

The ESIP land cover effort is led jointly by Já Já and John Townshend of Maryland's Geography Department. Townshend's group is working with scientists from NASA's Goddard Space Flight Center to produce a complete land cover map of the Earth at one-kilometer resolution, possibly the highest resolution ever attempted (Figure 1).

Besides studying land cover changes such as deforestation and desertification, the tools being developed by the ESIP-2 project will make possible many types of Earth systems studies. For example, scientists might augment global-scale models of climate systems with remote-sensing data, study the impact of climate change over time on vegetation, or look for sources, sinks, and changes over time in the global carbon cycle. --DH