Title: Space, Time, Sensors, and Data Semantics

Abstract:

In this talk we will discuss the issues concerning data management for 
environmental monitoring of moving phenomena in continuous media.  Specifically, 
we will concentrate on three topics: non-Newtonian notions of time, measurement 
units, and the need for better spatio-temporal data models, such as fiber bundle 
data models to model vector field data.

Classical temporal data modeling for databases has invoked a Newtonian conception 
of time, with the notion of universal simultaneous time.  Einstein's theory of 
relativity has supplanted the earlier Newtonian model of time among physicists, 
astronomers, and now GPS users.  We will discuss why and how this matters for 
DBMS systems.

Most contemporary DBMS systems, query languages, etc. entirely ignore issues of 
measurement units failing to adequately support many sensor based applications. 
We will discuss some measurement unit issues and dimensional analysis and suggest that 
they be incorporated into our type systems for DBMSs.

We note that conventional DBMSs (relational, object oriented, OLAP, and XML) are built 
from collections of discrete things (tuples, objects, "facts", or trees).  However, 
for many applications, such as weather forecasting, climate simulations, oceanography, 
water pollution studies, astrophysics, and other fluid dynamics applications, such collections 
of discrete objects (e.g., sets) are not an appropriate data model. Common to many of these 
applications is the notion of vector field data (such as velocity fields for wind or ocean 
currents).  It hardly makes sense to talk about interpolation in classical data models of 
collections of discrete objects.  We will recount some of the basic ideas of fiber bundle 
data models, first investigated by Lloyd Treinish at IBM for data visualization applications.  
Variations of such data models have sometimes been referred to as vector bundle data 
models or as sheaf data models.

We summarize by suggesting that sensor data management for environmental monitoring of 
fluid dynamics phenomena (weather, climate, oceanography, etc.) is an area of growing 
importance, rapidly growing data sets, and in need of considerable development of better data
management technology.  We will then discuss a number of funding opportunities at National 
Science Foundation that support such research.


Bio:

Frank Olken has been a computer scientist at Lawrence Berkeley National
Laboratory working primarily in scientific and engineering data management
research and standards development and in bioinformatics.  He has a Ph.D. in
Computer Science from UC Berkeley. For the past 3 years he has been detailed
to the National Science Foundation, Computer and Information Sciences and
Engineering Directorate, Information and Intelligent Systems Division,
Information Integration and Informatics Program.  As a program director at
NSF he manages proposal reviews and research awards in core data management
topics, graph and tensor data management and mining, data semantics,
ontology and semantic web technologies, cyber-physical systems, parallel
DBMS, and bioinformatics. His research interests include query optimization, 
scientific, engineering, and social science data management, electronic health 
records, data semantics, ontologies, XML and graph data management, etc. 
He is a member of ACM, IEEE, ASA, AAAI, AAAS, and ACS. He is also a metric zealot. 
He can be reached via email at folken@nsf.gov.