Global climate change seems to cause a heated debate nearly every day in Congress. Many argue about its origins and effects without knowing that most of the research involved in understanding global climate change requires statistics.
“Many environmental problems are inherently statistical problems,” said William Christensen of the BYU Department of Statistics. “Any time you look at a set of numbers over the years you will see trends; environmental statistics is really about pulling out important signal from the noise that is embedded in any environmental data.”
Environmental statisticians like William Christensen focus their research on variables that affect the environment, such as air quality, pollution, and temperature. In his upcoming publication in the journal Biometrics, Christensen provides a new method for estimating trends in spatial data, and illustrates the approach in modeling climate.
His new method involves kriging — predicting climate variables at geographical locations based on data gathered from neighboring locations. For example, kriging could be used to calculate the temperature in Provo from temperatures gathered from surrounding cities.
Christensen’s research uses statistical processes to weight data according to accuracy. This optimizes spatial prediction when variability exists. In the local example, his method would focus on how to predict the temperature in Provo when the individual thermometers at surrounding cities do not have the same level of trustworthiness.
The research Christensen has done throughout his career is often targeted toward answering specific questions, but this paper’s purpose was to be a tool for future researchers.
“As a statistician, I sometimes am involved in specific scientific questions, but I also work on building ‘hammers,’” he said. “I build tools for people who are working on their own projects or questions in a wide variety of areas.”
His new method will likely serve geologists, statisticians, ecologists and scientists from other areas as well. Using climate data on the Hudson Strait in northern Canada, Christensen demonstrated the utility of his new approach. He used kriging with the location-specific measurement error taken into account for this set of data. Christensen’s new kriging approach is becoming known as a successful, improved predictor in climate modeling.
Research like his will assist other scientists in pursuing questions about global climate change.
“Hopefully it really affects how people do research in a variety of areas,” he said.