![]() |
|||
![]()
|
![]() |
![]() |
![]() |
6.5. CERES: CALIFORNIA ENVIRONMENTAL RESOURCE EVALUATION SYSTEMCalifornia Resource Agency (CRA), 1996 The purpose of CERES is to improve information services regarding California's rich and diverse environments. By combining databases and expert systems techniques and making them available on the Internet, a person can get access to a diverse set of electronic data on the natural resources of the state. The ultimate purpose of the CRA is to improve environmental analysis and planning. This is sought and achieved by integrating natural and cultural resources information from multiple contributors and made available and useful to a wide variety of users. In this sense, it matches much of the same idea as the Whale Watcher Expert System introduced previously. The system is able to help the public with problems such as how to obtain a hunting or fishing licence. It can also respond to queries related to the status of current water supplies in the state. CERES can be characterized as an intelligent database that is accessible through the World Wide Web. The functions of the knowledge-base part in CERES can be compared to a librarian. The main body of CERES is a library with several departments. In addition, it includes GIS capabilities. The library consists of data like satellite imagery and photography, information on environmental impacts, reports, and vegetation and wildlife data. The system contains electronic data catalogs and advanced searching capabilities. It is defined as a cooperative system using agents to provide the type of decision support specified. It applies several multimedia features that can be handled through the use of ordinary Web browsers. 6.6. SEAMason TSDSYS and SIRENAS discussed above are systems intended to preserve the environment on a local level. The impact they address will have a serious effect on the environment and the life that this environment accommodates. Yet, on a global level, a single event causing a chemical release or waste disposal will usually be confined to a fairly limited region. Hence, local monitoring is sufficient. Both TSDSYS and SIRENAS address different views on this type of environmental threat. The scope of the NASA Ames system, SEA (Seismic Event Analyzer) is much wider. It applies to a global threat that may have devastating effects on the whole earth if not controlled properly. SEA analyses data from the NORESS experimental seismic array station in Norway (Mason 95). The principal purpose of SEA is to monitor seismic reactions registered by NORESS and transmitted via satellite to the NASA Ames Research Center. Inherent in this is the need to monitor nuclear test activity and distinguish this from earthquakes. Different states have signed nuclear weapon test ban treaties that need to be enforced. Enforcement has an immediate political motivation, but its ultimate objective is to ensure that the global environment is preserved. Neither war acts nor gradual contamination through re-armament processes must be allowed. Obedience to both comprehensive test treaties as well as so called low-yield test ban treaties are required. Seismic monitoring is the most reliable means available to verify compliance with treaties regulating underground nuclear weapons testing. However, it is a difficult task. Treaty violations can be disguised. Nuclear tests can be designed to produce weak seismic signals. Events of this type are characterized by a low signal-to-noise ratio often hidden in background noise or discharged concurrently with other seismic occurrences. According to Mason, up to 20,000 events need to be analyzed for potential ban treaty violations. Since few experts are available for this, an automated system was required to help interpret and classify seismic events. SEA interprets data from multiple sensors. The data transmitted is stored onto an optical disk. An event-detection program analyzes the raw data for a possible seismic event. If a seismic event is registered, the data is stored in a unique file and passed on for further examination. SEA checks the wave pattern that represents the event. Different types of events produce different types of patterns. Such signatures are characterized by a distinct frequency, phases, and the paths followed through the earth. The latter determines the velocity of the propagating wavefront and therefore also the time before detection. The distinctions caused by this phenomenon enables a seismologist to estimate the distance of an event. The direction is determined by means of signal-processing tools to calculate phase characteristics. In this way, the location of the event can be determined. In real cases the signals detected may be of poor quality and wavefronts following certain paths may be totally missing. This makes the signatures blurred and complicates interpretation. In other situations, interference between two concurrent events may obscure impressions. In order to handle the challenging task of identifying ban violations, SEA applies a technique of partial interpretations (see Figure 7). This implies building a series of intermediate conclusions, each one an extension of the one before. Finally, each part is examined for inconsistencies; parts of that kind are rejected.
A belief-revision system is applied in the partial interpretation. If the belief in the validity of some step is contradicted, it is judged incorrect. The dependencies on the rejected part is kept track of so that these can also be revised. The type of truth maintenance used has been coined "multicontext truth maintenance." It is typically applied in situations where a number of solutions must be found. The use of this has been qualified by the fact that seismologists are seldom able to arrive at a single interpretation. This is due to lack of sufficient discriminating knowledge or data. The principal knowledge representation scheme is semantic networks. The network includes concepts such as Event, Phase, Segment, etc. They are interlinked by relationships such as Number-Of-Phases, Has-Phase, Phase-Id, and Begin-Time. IF-THEN rules are used to encode the actual interpretation knowledge. The rule-inference is data driven, operating in a forward-chaining manner. The action part of a rule may call a function in C, Fortran, or LISP. The truth-maintenance element is supported by an assumption-based reasoning technique. The system tags each deduced fact with a justification indicating the rule and antecedent facts used to create it and the assumptions that supports the fact that is believed. SEA uses this both for producing explanations as well as for rule operation trace. It aids in determining the current set of beliefs and supports the required dependency-directed backtracking. The assumption-based reasoning part consists of both an assumption database and truth-maintenance rules. SEA may operate on its own and as an interactive partner for a person. It was partially written in LISP, C, and Fortran. The interface is accommodated in the MIT X Window System. It is supported by a SUN3/160. SEA has been field tested with fair success and development continues. In order to cope with the great amounts of data related to this kind of interpretation, a multi-agent kind of architecture is considered.
|
![]() |
|
Use of this site is subject certain Terms & Conditions. Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details. |