7.3 EXPERT SYSTEM ANALYSIS
The underlying structure of the expert system has been previously described in
Hendee (1998, 2000). Production rules (basically, if/then-type heuristics) utilized in
CREWS are drawn from published data, field observations, and from discussions in the
literature.
The approach presented here reflects first-order laboratory and field-based testing
of instruments, in conjunction with programming of the expert system. The expert
system acts as a model, which will eventually help to further elucidate the role of the
physical environment in biological events which are influenced by the physical
environment, and which can be measured with a robust sensor.
Table 2 presents the terms used for the inference engine within the expert system.
Basically, these subjective interpretations help to reduce complexity in modeling the
environment and in assessing its role in influencing a marine behavioral event (e.g.,
coral bleaching). Data from each sensor are categorized according to the table into
subjective data ranges (e.g., drastically low, very low, etc.), as described by experts
who use the sensors or work with the parameters in question. The terms “unbelievably
low” or “unbelievably high” represent thresholds beyond which the measurements
would be considered unrealistic in nature. The subjective periods of the day explained
in Table 2 are those perceived by humans, and which also quite often correspond to
periods of biotic behavior (e.g., crepuscular feeding behavior at “dawn” or “sunset”).
When an observed condition (e.g., high sea temperature) holds to the same subjective
data range beyond one of the basic periods, which are three hours each, the condition is
reassigned to the next larger category. For instance, if sea temperature is “very high”
for “dawn” and “morning” (each of which is a three hour period) it becomes reassigned
as “dawn-morning,” a six hour period. Similarly, if the high sea temperatures persist
for all daylight hours, the condition is reassigned as occurring for “daylight-hours.”
8. Research Application
The CREWS stations have been designed to provide an extensible architecture so
that instrumentation may be added relatively easily to provide answers to selected
research questions. One of the primary goals of the CREWS stations has been to
elucidate the role of light and temperature in the phenomenon of coral bleaching.
Another research question of value is determining what role carbon dioxide flux plays
in coral growth, bleaching, and coral larval settlement. In select situations, where
sediment resuspension and turbid outflow may have deleterious effects on the coral
reef, turbidity sensors might also be added. The list of possible research questions that
can be addressed by quality long term investigation of specific parameters includes, in
addition to the above mentioned parameters, monitoring of nutrients, dissolved oxygen,
pH (potential hydrogen), Eh (electrochemical potential), and video-based ecosystem
analysis. For the present discussion, research into the role of light and high sea
temperature in coral bleaching events will serve as an example of the effectiveness of
the CREWS station design and expert system analysis.
145
Meteorological and Oceanographic Instrument Array
The underlying structure of the expert system has been previously described in
Hendee (1998, 2000). Production rules (basically, if/then-type heuristics) utilized in
CREWS are drawn from published data, field observations, and from discussions in the
literature.
The approach presented here reflects first-order laboratory and field-based testing
of instruments, in conjunction with programming of the expert system. The expert
system acts as a model, which will eventually help to further elucidate the role of the
physical environment in biological events which are influenced by the physical
environment, and which can be measured with a robust sensor.
Table 2 presents the terms used for the inference engine within the expert system.
Basically, these subjective interpretations help to reduce complexity in modeling the
environment and in assessing its role in influencing a marine behavioral event (e.g.,
coral bleaching). Data from each sensor are categorized according to the table into
subjective data ranges (e.g., drastically low, very low, etc.), as described by experts
who use the sensors or work with the parameters in question. The terms “unbelievably
low” or “unbelievably high” represent thresholds beyond which the measurements
would be considered unrealistic in nature. The subjective periods of the day explained
in Table 2 are those perceived by humans, and which also quite often correspond to
periods of biotic behavior (e.g., crepuscular feeding behavior at “dawn” or “sunset”).
When an observed condition (e.g., high sea temperature) holds to the same subjective
data range beyond one of the basic periods, which are three hours each, the condition is
reassigned to the next larger category. For instance, if sea temperature is “very high”
for “dawn” and “morning” (each of which is a three hour period) it becomes reassigned
as “dawn-morning,” a six hour period. Similarly, if the high sea temperatures persist
for all daylight hours, the condition is reassigned as occurring for “daylight-hours.”
8. Research Application
The CREWS stations have been designed to provide an extensible architecture so
that instrumentation may be added relatively easily to provide answers to selected
research questions. One of the primary goals of the CREWS stations has been to
elucidate the role of light and temperature in the phenomenon of coral bleaching.
Another research question of value is determining what role carbon dioxide flux plays
in coral growth, bleaching, and coral larval settlement. In select situations, where
sediment resuspension and turbid outflow may have deleterious effects on the coral
reef, turbidity sensors might also be added. The list of possible research questions that
can be addressed by quality long term investigation of specific parameters includes, in
addition to the above mentioned parameters, monitoring of nutrients, dissolved oxygen,
pH (potential hydrogen), Eh (electrochemical potential), and video-based ecosystem
analysis. For the present discussion, research into the role of light and high sea
temperature in coral bleaching events will serve as an example of the effectiveness of
the CREWS station design and expert system analysis.
145
Meteorological and Oceanographic Instrument Array