My Research (as of October 2007)

 


 

 

 

 

 

 

 

 

 

Multiscale Multigranular Image Segmentation

 

In remote sensing, land cover characterization is often required at multiple spatial and categorical scales. In this research, we introduce a framework that

allows for adaptive choice of both the spatial resolution of subregions and the categorical granularity of labels. Our framework is based upon a class of

models we call mixlets, a blending of recursive dyadic partitions and finite mixture models. The first component of these models allows for the
sparse representation of spatial structure at multiple resolutions. The second component provides a natural mechanism for capturing the varying

degrees of mixing of pure categories that accompany the use of different resolutions, and enables them to be related to a user-specified hierarchy

of labels at multiple granularities in a straightforward manner. A segmentation is produced in our framework by selecting an optimal mixlet model, through
complexity-penalized maximum likelihood, and summarizing the information in that model with respect to the categorical hierarchy. Both theoretical and

empirical evaluations of the proposed framework are presented in a series of papers. We present here sample remote sensing data, manual and the code

for disseminating the research.

 

 

Modeling the Spatial Patterns of Addiction in the U.S.

Suchi Gopal, Bill Anderson, Matt Adams, Lauren Friel, Boston University
Mark Vanelli MD, MHS, MBA, Harvard Medical School, Cambridge MA
Mark Albanese, Cambridge Health Alliance, Cambridge MA

This research project incorporates data from many sources including health survey data, arrest records, popular media, socio-economic, as well as prescription data for the period 1999-2002 at the sub-state level. The project includes mapping and spatial analysis of addiction from heroin, cocaine, opiates, marijuana, methamphetamine, ecstasy and club drugs. This research uses ESDA (Exploratory Spatial Data Analysis) and spatial analysis (predictive models) at the zip code to sub-state scale. Two physicians from Harvard Medical School of Public Health and Psychiatry are co-authors on this paper and are providing the public health policy perspective to the GIS analysis.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

The Evolving Social Geography of Blogs

 

Dr. Sucharita Gopal

Department of Geography & Center for Cognitive and Neural Systems

 

Blogs are creating new virtual communities and are changing the social geography of the Internet. The present research takes a spatial science perspective to examine the growth and evolution of blogs using the methodology of social networks.  Blogs relating to Hurricane Katrina were analyzed.  Influential leaders, message content, external links, time, actors (bloggers) and other variables of the diffusion process are analyzed.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Testing for Local Spatial Autocorrelation in Ecological Systems: Using G* and Neural Networks

 

Tigga Kingston

Boston University

 

Determine the spatial and temporal variability in assemblage composition at the local, landscape and regional scale, Investigate the processes behind this variability at the species level through studies of population and behavioral ecology, Determine the impact of population dynamics on community interactions.

 

 

 

Spatial Distribution and Analysis of Pneumococcus Infection in the United States (1990 – 2000)

 

Lauren Paletta, Sucharita Gopal, Barbara Mahon and Jason Sanders

 

Boston University Department of Geography

Boston University School of Public Health and Medicine

 

Streptococcus pneumoniae (pneumococcus) is a major cause of illness amongst children in both the developed and developing world.  Prior studies indicate that the spatial distribution of the disease is affected by location, exposure to Ultra Violet Radiation (which varies by latitude and time of year), and ethnicity (notably skin color and type of clothing).

 

The research for this paper has three objectives.  The first is to investigate the relationship between UV radiation and the prevalence of the disease.  The second is to describe the spatial distribution of the disease and examine the role of race, ethnicity and social factors.  Lastly, model the b of the disease prevalence.  The data for this study includes the Total Ozone Mapping Spectrometer (TOMS) measurements of Erythemal UV from NASA Goddard Space Flight Center and a meta-database of incidences of pneumococcus disease in the U.S. compiled by Mahon for the period of 1990-2000.  Understanding the distribution and epidemiological characteristics of pneumococcal disease will be invaluable in public health field in the design and evaluation of prevention strategies.