Transcript cronus04-zreda-i

ITR: Collaborative research: software for
interpretation of cosmogenic isotope
inventories - a combination of geology,
modeling, software engineering and artificial
intelligence
Marek Zreda, U. Arizona - isotopes, modeling
Elizabeth Bradley, U. Colorado - artificial intelligence
Ken Anderson, U. Colorado - software engineering
Funded by: NSF/IIS/EAR
Duration: 2003 - 2008
Budget: $1.6M
Software for all occasions
Software will satisfy computation and information needs in six areas:
(1) production rate calibration and scaling;
(2) calculation of individual sample age using production rates
from item 1 above;
(3) calculation of surface exposure age (also called apparent age) of
the landform from multiple sample ages (each calculated from item
2);
(4) calculation of model age of the landform (correcting item 3 for
geological processes that affect apparent age);
(5) compiling information necessary for production rate calculations
(raw physical, chemical, and geological data);
(6) compiling information on existing and potential applications
(what can be dated and how).
Structure of software
The artificial intelligence core links
databases (knowledge, information
and data storage) and models
(geological and mathematical) on
one side, with field and laboratory
data on the other side, to produce
results (numerical output,
interpretation, speculations and
theorizing, and advice on further
course of action).
Solid lines and arrows show input
of information and data; dashed
lines and arrows show requests
and feedback.
Two challenges
Challenge 1: Integration.
Our software:
Effective analysis of geochemical data
requires a heterogeneous collection of tools
and techniques, applied in the right order to
the right data, and guided by the right
interpretation. Each analysis method rests
upon different facets of the underlying
science and demands different software
algorithms. Current analysis tools are useful
from the computational point of view, but
they share the same two shortcomings: they
do not have user-friendly interfaces, and so
they are practically unusable by scientists
other than their authors; and they are not
integrated with programs that calculate
exposure ages from cosmogenic
concentrations.
We will construct a software architecture and
component framework for the analysis and
interpretation of cosmogenic nuclides in
terrestrial environments. The use of
components will ease the construction of new
analysis tools by transforming the current
(and manual) “development from scratch”
process into an automated “development by
assembly” process. The software architecture
will provide a uniform framework that
combines the computational models,
scientific databases, and AI capabilities into a
cohesive whole accessed by a usable,
intuitive, and flexible user interface.
Two challenges
Challenge 2: Complexity.
Our software:
The complexity of the cosmogenic nuclide
dating poses another IT challenge.
We will incorporate AI techniques directly
into the software tool, in order to explicitly
capture the properties and dependencies of
the analysis process-e.g., that one should
collect additional samples if the numerical
analysis of geochemical results indicates that
there is too much noise in the data to produce
the desired temporal resolution-and advise
the user accordingly. This kind of dedicated,
knowledge-based data-analysis support
facility will be useful in assisting both naive
and experienced users. Building it will be a
significant knowledge engineering problem,
and it will require effort and communication
from all the PIs, working together.
Objectives
To develop, within a uniform framework:
(1) Geological-mathematical models to quantitatively describe the
accumulation of cosmogenic nuclides in evolving landforms, and to
invert field and isotopic data to obtain landform ages and rates and
frequencies of geological processes that act on these landforms.
(2) Databases of basic and applied knowledge and data useful in the
evaluation of cosmogenic data, drawn from all fields relevant to
cosmogenic nuclide geochemistry: nuclear physics and cosmic-ray
physics, chemistry and geochemistry, various fields of geology,
atmospheric sciences, magnetism and paleomagnetism, statistics and
mathematics.
Objectives
(3) An artificial intelligence system for analysis of cosmogenic data,
connecting logically all existing information and models (from
objectives 1 and 2) with any new, user-generated data. The system
will have three main uses: research design (construction of testable
hypotheses), analysis and interpretation of isotopic data, and
speculation under conditions of uncertain or absent information.
(4) A software architecture that supports the construction of a state-ofthe-art component-based software system that enables “plug-andplay” assembly of software tools that embody the features of
objectives 1-3 above.
Not an objective
It is not our objective to improve calibrations or scaling functions, or
to improve the interpretations of published applications of
cosmogenic dating. Our goal is to develop a new integrated analysis
tool for cosmogenic isotope dating, explore its potential and show its
versatility, computational feasibility and accuracy.
For improved calibrations and scaling - we have CRONUS!
Three-tiered architecture