Developing, Maintaining, and Employing Large Computational Frameworks for the Ecological Sciences

Participants work through a problem presented by a speaker.

The following was written by Lea Jenkins, Associate Professor, Mathematical Sciences Department at Clemson University and one of the program leaders for the SAMSI Mathematical and Statistical Ecology Program.

The workshop combined scientific lectures with hands-on workshop components to inform participants of the current trends in mathematical and statistical computing.  The lecture content included material on current modeling efforts for navigational flows, movement and tracking of ecological species (both aquatic and terrestrial), current trends in scientific computing languages and environments, and ecological inference.  Hand-on material included information on repository management, computation and documentation using Python and R software, and creation and documentation of tests for software and data validation and verification to ensure scientific replicability of published results.  Lightning talks preceded a poster session on the first evening of the workshop.  The lightning talks introduced participants to one another and were used to begin collaborative discussions. Scientific lectures were delivered by scientists at the USACE Engineer Research and Development Center (USACE-ERDC), the University of Illnois at Urbana-Champaign (UIUC), the University of North Carolina at Chapel Hill (UNC-CH), and Continuum Analytics.   Hands-on lecture content on effective uses of current software tools for simulation and analysis in the science disciplines were delivered by scientists from USACE, SESYNC, the University of Arizona, UIUC, and Continuum Analytics. Breakout sessions were started on Wednesday afternoon.

David LeBauer from University of Illinois at Urbana-Champaign lecturing.

There were four primary topics for the breakout sessions.  One session began an analysis of migration patterns of various species of fish, with data provided by a participant from the World Wildlife Fund.  A second session focused on developing computational models from a conceptual framework, and wrote a basic simulation tool incorporating agent-based models with stochastic movement dynamics.  A third session worked on issues surrounding data assimilation, where data exists on multiple scales and the different realizations must be combined to deliver an accurate description of the underlying dynamics.  The associated problems of interest included finer scale resolution of census data and resolution of remote sensing and direct measurements of bathymetry in near-coastal regions.  The fourth session focused on rendering and building distinct finite element structures for use in hydraulic simulation codes. Many of the participants were early career, ranging from graduate students to postdocs to early-career scientists between the postdoc and academic career realms.  Github repository accounts were created for each of the breakout groups to allow continue collaborative efforts beyond the end of the workshop.

Checking out a poster during the poster session and reception.

Enjoying some food at the reception/poster session.

SAMSI Postdoc Profile – Kimberly Kaufeld

Kimberly Kaufeld hiking in Arizona.

SAMSI postdoctoral fellow Kimberly Kaufeld always wanted to apply what she has learned to her love of the outdoors. She grew up in Rochester, Minnesota appreciating the snow and cold. She entered the University of Minnesota Duluth in 2001 where she double majored in psychology and statistics. She liked both subjects, and in her undergraduate was able to merge her studies in a funded undergraduate research project on workplace environments in the psychology department. After graduating from University of Minnesota Duluth in 2005, she joined AmeriCorps and spent a year volunteering. Some of the projects involved working with victims of Hurricane Katrina and in Oregon working on environmental projects such as removing invasive species, sparking her interest in ecology.

Kimberly with the AmeriCorps team.

When she completed her time with AmeriCorps, she applied for graduate school at Minnesota State University, Mankato. She was accepted into the mathematics/statistics program where she taught statistics courses. Upon completing her Master’s degree in 2009, she was offered an instructor position teaching statistics and mentoring undergraduate students for a year before deciding that she wanted to get her Ph.D.

Kimberly Kaufeld loves the outdoors.

She then moved to Colorado to attend University of Northern Colorado for her Ph.D. where she was able to do an independent study in spatial statistics. “I liked the program there because it was small enough that I could do the research I really wanted to do,” she explained. She started working on a research project involving the spread of the mountain pine beetle, a pest that has devastated vast swaths of forests in Colorado and other mountain states. She has been looking at developing new methodology and spatio-temporal modeling to measure this spread. Kimberly said, “My advisor, Trent Lalonde, got me involved me in studying health effects as well which was a good pairing for my research.” She became interested in Adolescent Health, in particular, after an internship with the Educational Innovation Institute at Northern Colorado. She went to several statistical, health and ecology conferences during her time there. Kimberly seeks opportunities to expand her knowledge and collaborate with others. After her first year at Northern Colorado she reached out to the Colorado Forestry Service to inquire about their forestry data. She got to see firsthand what real world problems they had by participating in workshops on aerial survey techniques. In her second year she reached out to the National Center for Atmospheric Research (NCAR). ” I talked with NCAR to see if there were any student visitor opportunities, so I gave a talk at the mesa campus and was given a student visitor position.” She was involved with the IMAGe program at NCAR for about two years continuing her research on beetle kill and climate modeling. She worked with people such as Doug Nychka and Steve Sain. “During the summer, we would meet as a group with the other graduate students there and talk about the work we were doing. We would give a 5 minute presentation and get feedback on where each person could go next,” she explained. During her time at University of Northern Colorado Kimberly was recognized as a graduate scholar in her college and awarded the Dean’s Citation for Excellence in recognition of her academic research and all the outreach efforts she did including the work at NCAR, Colorado Forestry Service. Kimberly learned about SAMSI from Doug Nychka and Dorit Hammerling, who was a former postdoc at SAMSI. The program at SAMSI in Statistical/Mathematical Ecology was right in line with her interests so she decided to apply to the program. She also thought that the Triangle region was a good place to be because there would be opportunities to reach out to the various universities in the area to give talks and to collaborate with different people. “It has been great to work with ecologists, statisticians and mathematicians this year, getting all the different perspectives and hear about the different types of research that they do and how we can put all these different ideas together,” said Kimberly, “Learning about joint species distribution models has been really great because you learn about how each species interact with each other. When you look across space and across time, how does competition impact another as the species grow together? How do you create a model that will account for these various interactions? And then you have to think about how other factors such as precipitation, temperature and other many others that can have an impact on all these species.” The modeling can get very large and complex vary fast due to large number of species. This is a focus of the multivariate models group that Kimberly is involved in, developing dimension reduction techniques to handle large number of species. These problems are not just in ecology but are also being discussed in the bioinformatics program too, so she has also been attending some workshops that are focusing on similar issues as well. Kimberly will be spending her second year primarily at NC State, but will be continuing the research she began while at SAMSI. She will be working with Montse Fuentes and will be looking at air pollution and mortality rates and collaborating with the EPA.

Kimberly Kaufeld speaking at the RTP 180 event.

Kimberly has been very helpful promoting SAMSI this year. She was selected by SAMSI’s communications director to give a talk to the RTP 180 lecture series, which is open to all people in the Research Triangle region. She also helped with a video that was promoting the various research organizations that are in the Triangle University Center for Advanced Studies Inc (TUCASI) campus, which is where SAMSI is located. She also gave a short presentation on mountain pine beetles at RTP 180 with the theme “Creepy Crawlers”. When she is not working, Kimberly likes to spend time with her husband of eight years, Kevin, and go backpacking or hiking. They both love the outdoors, so any time spent in nature is a good day to both of them. They are both enjoying exploring North Carolina. She also enjoys skiing, snowboarding and running.

Affiliates Meeting in Miami Covers Review Processes, Availability and Quality of Data and More

Neung Soo, NISS, and John Eltinge, BLS.

The following was written by Neung Soo Ha, postdoctoral fellow for SAMSI and NISS.

The National Institute of Statistical Sciences (NISS) and the Statistical and Applied Mathematical Sciences Institute (SAMSI) held its annual Affiliates meeting on March 15 at Hyatt Regency Hotel in downtown Miami, Florida, in conjunction with the Eastern North American Region (ENAR) meeting. About 30 affiliates and board members attended the meeting. The acting director, Nell Sedransk, provided a brief history of NISS. She explained that the purpose of being an affiliate is about active involvement with the research program development, such as planning conferences and workshops. She finished by talking about the new website for NISS and asked for input from every affiliate member.

NISS Acting Director Nell Sedransk gives an overview.

Sujit Ghosh, SAMSI’s deputy director, gave a description about SAMSI and some of its programs, including workshops, working groups, and education outreach programs for undergraduates and graduates. He showed how affiliates can be involved with developing future programs.

SAMSI Deputy Director Sujit Ghosh speaking to the affiliates.

The morning session ended with presentations from three postdoc researchers from NISS and SAMSI. First presenter was Daniel Taylor-Rodriguez who is a first year postdoc from SAMSI. His presentation was about analyzing the jaguar population in Central and South America. Second presenter was Neung Soo Ha, who is a second year postdoc from NISS. He presented an analysis of insurance rates for Florida in 2010. The last presenter was Hang Kim, who is a third year postdoc from NISS, who talked about data confidentiality.

SAMSI post doc Daniel Taylor-Rodriguez, shares his research with the affiliates.

In session 2, there was a panel about effective reviewers. The panel members included: Michelle Dunn, National Institute of Health program director, Sujit Ghosh, SAMSI deputy director and former National Science Foundation program director, and Xihong Lin, professor at Harvard University, School of Public Health.

The goal of the session was to foster a clear understanding of review processes and grant proposals from reviewer and reviewee’s point of view. Dr. Dunn described about how the program directors are usually the ones who decide on the funding decisions for the project, yet those people might not be the experts in the field for the proposed program. Thus, when the principal investigators write the proposals, they should always keep in mind about whom would read and make decisions on the project.

There were about 30 people at the affiliates meeting in Miami, FL.

The next two panel members talked about the review processes for proposals and journals. Dr. Ghosh said that the reviewing the proposals should focus on overall ideas rather than being technical and should make judgements on contents. Dr. Lin talked about reviewing for journals. It’s important for the journal reviewer to see what is novel about the article and to provide constructive criticism on methodologies.

Session 3 was about availability and quality of the data. The panel members were: David Madigan, professor, Columbia University; Patrick Ryan, research associate, Janssen Research and Development; Rima Izem, researcher, Food and Drug Administration (FDA); and John Eltinge, director of research at the Bureau of Labor Statistics (BLS). They all talked about the concept and the availability of “Big” data from different fields, and how the data can be used for analysis and how results can propagate to the population.

Sally Morton, professor, University of Pittsburgh; and Ruth Pfeiffer, director, National Cancer Institute.

The last session was about the leadership skills. The panel members included: Sally Morton, professor, University of Pittsburgh; Ruth Pfeiffer, director, National Cancer Institute. They talked about what makes an effective leader and agreed that a leader should be a person with a vision and clear objectives. Dr. Morton emphasized that a leader should also have an understanding about both the financial and human resources, and it is important for institutions to have a program for leadership skills for junior researchers. Dr. Pfeiffer described that a person with authority does not equate to a person with leadership, and that a lower ranked person can also have a leadership skills. She also mentioned that a person can acquire leadership skills through proper trainings.

All in all it was a very informative meeting.

Learning about the Human Microbiome

The following was written by Nur Majida Shahir, graduate student, Bioinformatics and Computational Biology at the University of North Carolina at Chapel Hill.

Nur Majida Shahir at the Microbiome workshop at SAMSI.

This past month, I had the opportunity to attend a workshop at the Statistical and Applied Mathematical Sciences Institute (SAMSI) on the human microbiome. While I was only able to attend the first day, the information and insight I gained over the course of the day was amazingly useful.

Dr. Susan Holmes gave the first talk of the morning session on Multi-Table Data Analysis. While the talk itself was interesting, the thing that stood out most to me was being introduced to an R package that was created by Dr. Holmes’ group called phyloseq. Prior to this workshop, the only downstream analysis program I knew of was Explicet but after being exposed to the flexibility of phyloseq, I have a feeling that I may be using the latter more in my research.

Bill Shannon (L), Washington U. at St. Louis, and Timothy Randolph, (R) Fred Hutchison Cancer Center

Dr. Vanni Bucci gave the second talk of the morning on predictive modeling of microbiome dynamics. In contrast to the previous talk, this was approaching the microbiome from an applied mathematics perspective with a focus on creating and using a minimal model to study microbiota dynamics in enteric infections. I found this talk particularly fascinating in part due to my background in mathematics as well as the fact that looking at the microbial community dynamics in the gut makes sense due to the transient nature of some of the flora seen in the gut.

Nur participates in her first breakout working group session.

After the morning sessions and lunch, I had my first experience with a working group breakout meeting at this workshop. On one hand, it was a good experience to hear what people were thinking with regards to various datasets and the analysis of said dataset. There were many concepts and approaches that were thrown around that I honestly hadn’t thought of. On the other hand, I found it disorienting because I had a very superficial idea of what they were discussing. It would have been more beneficial if I had access to the data or at least the papers to which they were referencing prior to the working group breakout meeting.

One of the things that I enjoyed about this workshop was the varied backgrounds of the presenters. While the majority of the presentations were focused on statistics approaches and problems regarding the analysis of the microbiome, others approached the microbiome from a much more theoretical perspective as seen in Dr. Giseon Heo’s talk, which if I recall correctly, approached it from the perspective of knot theory.

People discussing their work at the poster session.

The poster session was held at the end of the day. Reflective of the talks, the poster session content was fairly diverse content-wise as well with posters ranging from the “standard data analysis + results + future directions” to more methodology oriented approaches regarding how to approach the data. I personally enjoy poster sessions because they allow me to approach the material at my own pace and to interact with the presenter in a more direct manner.

All in all, I left the workshop very content. I’ve attended a few conferences where halfway through I’m utterly exhausted and dreading the next 4 hours. At this workshop, I felt that my time there was both well spent and informative.