Improving Biodiversity Modeling
Biodiversity is the variety of life in a given ecosystem. Environmental changes affect biodiversity, but predicting the impact of those changes is complicated. Scientists have tools like species distribution models (SDMs), which rely on computer algorithms. But there are no best-practice standards to ensure models’ accuracy.
It’s a significant problem because scientists often use results from their modeling to inform “conservation, management, and risk assessment.” In other words, they’re drawing conclusions about the future from present-day data, gathered from models that don’t have to follow any particular guidelines.
In a new review published in the journal Science Advances, Professor Robert P. Anderson (The City College of New York) and a team proposed a “framework for presenting best-practice standards together with detailed guidelines for scoring key aspects of SDMs.”
“This new field has seen very rapid development over the past two decades, and only recently has it begun converging toward agreed-upon key concepts and methodological best practices,” Anderson told SUM. “The work presented here was motivated by the needs of initiatives to assess the state of biodiversity, for example the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES).”
The researchers built their framework around “four levels of standards”: gold, silver, bronze, and deficient. On the high end of the scale, gold is “aspirational” because it relies on “ideal data,” while deficient is “unacceptable.” They then suggested applying those standards to key components of SDMs — like predictor variables and model evaluation —and scoring each one.
Putting their framework into practice, Anderson and the team evaluated a sampling of 400 SDMs used between 1995 and 2015. Surprisingly, 46 percent of the studies were classified deficient in at least one area. But, encouragingly, each year the studies improved by 3.8 percent.
The researchers write: “The aims of establishing best-practice standards for models in biodiversity assessments are to provide a hierarchy of reliability, ensure transparency and consistency in the translation of scientific results into policy, and encourage improvements in the underlying science.”