Modeling habitat that’s just right

In this week’s Scientific Reports, Michigan State University researchers offer ways to take the guess work out of parameters.

Wolong Nature Reserve alpine landscape
China's Wolong Nature Reserve alpine landscape. By Thomas Connor

When modeling how an animal may be living in its habitat, scientists can find themselves like the Goldilocks of methodology – finding some study areas and resolutions of data too big or too small. In this week’s Scientific Reports, Michigan State University researchers offer ways to make a model just right.

Scientists depend on models to understand how creatures are using their space and predict how that may change under different conditions. Such considerations can be the foundation of good policy, especially when seeking to protect endangered or threatened species like China’s giant pandas. 

Thomas Connor, a PhD candidate in MSU’s Center for Systems Integration and Sustainability (CSIS), worked to take the guess work out of parameters and suggest techniques to best represent how pandas are living across vast swaths of southwestern China.

The “just right,” Connor and his colleagues learned, depends on the goal of the research. They found that in order to best model habitat in small areas you should go with small resolution-variables – think how fuzzy a photo can get it you blow it up really big and the information that is lost in the process. At larger areas, the resolution of the variables didn’t matter that much.

The researchers also recommend a willingness to combine predictions from several smaller study areas rather than the animal’s whole range.

“I found that the mountain-range models were more accurate than the panda range-wide model, meaning that it would be more effective to combine predictions from several smaller study areas than build a giant model of the whole panda range,” Connor said. “I think this is because there are important differences in panda habitat preferences between mountain ranges since they have been separated and adapted to different conditions. These differences are averaged out when we model at the whole range-scale.”

Connor has spent more than a year living and working in the remote mountains of the Wolong Nature Reserve frequented by pandas.

Besides Connor, “Interactive spatial scale effects on species distribution modeling: The case of the giant panda”

was authored by Andrés Viña, Julie Winkler, Vanessa Hull, Ying Tang, Ashton Shortridge, Hongbo Yang, Zhiqiang Zhao, Fang Wang, Jindong Zhang, Zejun Zhang, Caiquan Zhou, Wenke Bai and Jianguo “Jack” Liu, who is the director of CSIS.

The work was funded by the National Science Foundation.


Did you find this article useful?