Graduate STEM Fellow Profile

Hunter Glanz

Project Title: GLobAl Change Initiative: Education and Research (GLACIER)
Thesis: Bayesian Statistics and Discrete Inference
College/University: Boston University
Research Advisor: Surajit Ray
Degree Sought: Ph.D., Statistics
Department: Mathematics and Statistics
Research Focus: Land-cover Classification for MODIS Data
Teaching Partner(s): Donna Peruzzi

Description of Research

Urban growth and subsequent loss of agricultural lands, forests, wetlands and other highly valued biomes is one of the growing concerns for land planning and management agencies. Thus, acquiring land cover information has become increasingly important. Sources of data have ranged from aerial photographs to digital images captured from satellites. Recently, increasing emphasis has been placed on satellite imaging for many reasons including accuracy and the ease with which it lends itself to analysis. That is not to say the current methods are perfect or easy to implement. In fact, with data ranging from at most 30 years ago provided by multiple satellites at different resolutions, the methods of analysis used to acquire land class information have been evolving for some time and still are. Additionally, the abundance of data does not preclude the existence of missing data (for certain parts of the year in some regions). The methods for adapting the analyses to handle missing data are equally important and vary as well. The goals of my advisor and I are to produce best statistical analyses and methodology for this problem to date. Our current work employs some functional data analysis and graphical model tools to overcome the burden of missing data and some of the difficulties in the current multivariate methods of analysis. We hope to build a model that exploits the structure of the satellite image pixels in space and time, and by doing so not only classify a pixel as certain land cover, but determine whether that classification has changed from the previous time step to the current one. Using a Bayesian perspective is one key way we allow for the incorporation of important subject matter knowledge into the model.

Example of how my research is integrated into my GK-12 experience

Statistics is the science of analyzing data. I am fortunate that the classroom I am in provides labs and hands-on activities almost every single day as the main medium of learning. All of these labs involve making hypotheses, testing them, collecting data and then drawing conclusions about important scientific concepts from that data. As a statistician, I’ve enjoyed emphasizing the importance of collecting and analyzing data in order to make more sense out of the world. I am still searching for a suitable place in the curriculum to insert activities related to land-cover classification, but I have spent quite a bit of time helping to acquaint the students to the research process and the importance of analyzing data in any endeavor they may undertake.