Research and Supercomputing Visitor Program

RSVP collaboration
RSVP visitor Siddartha Nandy, a statistics graduate student at Michigan State University, talks with Doug Nychka, a scientist in IMAGe, who hosted his RSVP visit. When one considers how physical variables are correlated it is often the case that the strength of the dependence between measurements depends not only on how far apart they are but also on the direction. Siddartha’s research at NCAR was concerned with creating models that represent correlation patterns that are not the same in all directions. The plot below illustrates a way to measure this.
Correlation graphic
This figure illustrates how a hypothetical measurement in the center of this region is correlated with surrounding observations. Were there no preference for direction, the pattern would indicate circular contours for regions of equal correlation. The elliptical contours indicate longer range (or stronger) correlations in roughly an east-west direction with less dependence in a north-south direction. Moreover this example also illustrates a pattern that is slightly tilted from a horizontal alignment. Although this kind of anisotropic pattern for correlation is standard in spatial statistics, Siddartha Nandy has implemented this idea in a new statistical method that scales to large problems and is described in a novel way by three statistical parameters. In this way he is helping to make statistical tools more appropriate for analyzing large geophysical data sets. Moreover, this work ties into his dissertation research on parameter selection.

The Research and Supercomputing Visitor Program (RSVP) is designed to bring university faculty, researchers, and students to NCAR to foster collaboration with CISL staff and to provide training opportunities for underrepresented groups. The program pays for travel and living expenses for stays of up to three months.

These extended visits help establish stronger relationships that lead to long-term collaborations. For many visitors, this program represents a unique opportunity to interact with CISL and NCAR scientists and staff on topics ranging from high-performance computing and Earth System modeling to applied mathematics and statistics.

In FY2014, CISL used some RSVP funds to attract students from minority-serving and EPSCoR-state institutions to participate in the SEA Conference and Scalable Profiler Workshop, and to partially support travel expenses for students from EPSCOR states to attend NCL workshops in Boulder.

This program supports CISL’s education imperatives of integrating research and education and broadening participation by being a key component that integrates CISL’s education, outreach, and training efforts.

RSVP has sponsored graduate students, junior faculty, senior faculty, and scientist visitors from around the U.S. and the world.

Visitors in FY2014 included applied mathematicians to work with staff on numerical methods for geophysical models (e.g., new dynamical cores) and also graduate students in statistics to expand their mathematical research by including more focused methods for large geophysical data sets.

This program is made possible through NSF Core funding.