SGER: Intelligent Data Fusion and Model-based Reasoning to Acquire and Transfer Knowledge from LTER Sites

Investigator(s):

Gilbert Starks (PI) ; Stuart Rubin (Co-PI)

Sponsor:

Central Michigan University, MI 48859 9897746777

Start Date/Expiration Date

2000-01-01 to 2002-05-31 (amended 2001-07-24)

Awarded Amount to Date:

$37,000

Abstract: Award Abstract DBI 9982954; SGER: Intelligent Data Fusion and Model-based Reasoning to Acquire and Transfer Knowledge from LTER Sites The development of algorithms and a prototype software system for intelligent mining and fusion of ecological data is proposed. The software system, with human guidance, will produce a rule-based model that can be used to predict the response of a landscape to environmental driving variables, disturbances, or management events. The particular event of focus is prescribed burning for dry prairies. The tool will help ecologists to acquire knowledge about soils and vegetation at one site, for which data are abundant, and transfer the knowledge to another site, for which data are scant. Although the approach is general, the project will be conducted using data from the Cedar Creek Long Term Ecological Research (LTER) site in Minnesota (with abundant data) and the Newaygo Prairies Research Natural Area (RNA) in Michigan (with scant data). The site in Michigan has similar soils and vegetation to that in Minnesota. The prototype is expected to lead to advances in ecological knowledge engineering. Future versions should be useful to scientists trying to formulate research questions at new research sites where data are scant. Such a tool must accommodate both imprecision and lack of data, but at the same time, there is a need for greater spatial resolution in ecological research. Therefore, the work will emphasize rapid spatial collection of data in a natural language format using fuzzy sets. The proposed software tool will accept as input fuzzy data collected at high spatial resolution. In addition, an auxiliary voice recognition software/hardware system will be developed to facilitate collection of the fuzzy data in the form of natural language. The data will be recorded as fuzzy descriptions, like "Andropogon scoparius (crisp variable) is somewhat (fuzzy hedge) dominant (fuzzy primary term) in a plot that is heavily (hedge) vegetated (primary)." The natural language technique should provide for greater spatial resolution and perhaps no loss of precision: There is psychological evidence in the literature that response consistency across investigators might be more consistent if they are allowed to use natural language rather than forcing them to provide a graded, numerical estimate (like percent cover) to describe a fuzzy concept. Using a World Wide Web, client-side (Java) version of the rule-based simulator, an attempt will be made to elicit knowledge from ecologists working at the Cedar Creek LTER site as well as other experts. The experts will be contacted and invited to explore the online, rule-based simulation, which will have explanatory capability and will allow for storing suggested rule changes in the database. Thus, the simulator will aid communication among scientists. The overall result will be a new technique for knowledge acquisition and transfer from LTER and other databases.

NSF Org:

DBI - Division of Biological Infrastructure

Award Number:

9982954

Award Instrument:

Standard Grant

Program Manager:

Sylvia J. Spengler
DBI Division of Biological Infrastructure
BIO Directorate for Biological Sciences

NSF Program(s):

BIOLOGICAL DATABASES & INFORMA

Field Application(s):

Biology of Pest Organisms, Role-Terrestrial Ecosystem, Software Development

Program Reference Code(s):

BIOTECHNOLOGY - INFRASTRUCTURE, 9184
SMALL GRANTS-EXPLORATORY RSRCH, 9237

Program Element Code(s):

1165