Navigating a new era: Specialized models help researchers process ‘Big Data’

Navigating a new era: Specialized models help researchers process

February 4, 2014

Assistant Professor Andrew Finley, MSU Forestry

Navigating a new era: Specialized models help researchers process ‘Big Data’

The United States Environmental Protection Agency (EPA) estimates that forests occupy approximately 751 million acres of land in the United States. Forests provide many benefits and services to society. In Michigan, they play an important role in the state’s economy as a source of both employment and recreation. The EPA, however, is concerned that climate change will affect forest growth and productivity by altering the frequency and intensity of forest disturbances such as insect outbreaks, invasive species migration, wildfires and storms.

Wanting to understand more about the effects of climate and other ecological changes, the National Science Foundation (NSF) launched the National Ecological Observatory Network (NEON), a $434 million initiative that will create an open-access infrastructure that can be used to map, explore and predict changes in the nation’s ecosystem.

Michigan State University (MSU) AgBioResearch statistician Andrew O. Finley is collaborating with NEON scientists to develop important statistical models, overcome technology challenges and educate future researchers who will use the observatory’s data to manage the nation’s natural resources.

NEON will collect approximately 600 billion raw ecological measurements per year for 30 years from 106 locations throughout the United States. These measurements, once refined into accessible “data products,” will enable the understanding and forecasting of climate, land-use and invasive species change.

Ecological projections are typically based on data collected from microsystems, reflect a specific methodology, and are not generalizable to larger scales or systems. However, NEON’s data products — defined as cataloged groups of synthesized information — will provide necessary data so researchers can explore pressing environmental questions at appropriate space and time resolutions at continental scales.

“NEON recognizes that its data products must be statistically robust,” said Finley, MSU associate professor in the departments of Forestry and Geography.
“Here, the central challenge is in specifying valid statistical models for creating data products from the high-dimensional spatial and temporal data collected from
environmental sensors. Without the use of appropriate modeling techniques, the resulting spatial data products will amount to pretty maps that should not be used to inform policy or subsequent modeling efforts.”

Time and space are two important factors in statistical projections. To make valid, large-scale predictions, researchers must employ models that acknowledge the uncertainty that the two factors create. Finley will work to supply such models.

“In the summer of 2012, I was a visiting scientist in the NEON Data Products group,” said Finley, who is also an MSU adjunct professor in the Department of Statistics and Probability. “During that time, I was able to identify key modeling challenges in NEON’s progression from raw data to data products. My research focuses on creating modeling frameworks and open-source software that address those challenges; it also has a huge education effort designed to draw together students and experts from multiple disciplines and institutions.”

Finley explained that an understanding of the theories, methodologies and tools needed to employ advanced instruments and statistical models isn’t usually included in ecology or natural resource education. His project also focuses on creating transdisciplinary educational opportunities for students in those fields.

Specifically, he plans to enrich science education in 23 southwestern Michigan K-12 schools by collaborating with the MSU Kellogg Biological Station; develop cross-college curricula for undergraduate and graduate students; and explore contemporary topics with graduate students and early-career scientists in advanced courses, workshops and symposia.

He is especially excited about the series of graduate student workshops he will lead in Boulder, Colo., each summer for the next four years. NEON and the National Center for Atmospheric Research (NCAR) will host the workshops.

“The idea is to get graduate students from diverse disciplines together to explore their research challenges and teach them new modeling and computing techniques,”he said. He explained that successful navigation of the “Big Data” era requires next-generation researchers to have statistical and software tools that helps them develop valid answers to pressing questions.

“I’m excited to help projects like NEON succeed — there’s substantial value in these kinds of initiatives,” he concluded. “NEON’s plans to develop and distribute these data will ultimately help forest scientists, like myself, manage and sustain forest systems capable of persisting in changing environments and continuing to meet society’s demand for ecosystem services.”


Originally printed in AgBioResearcher's 2013 Annual Report, available here.
Michigan State University Michigan State University Close Menu button Menu and Search button Open Close