New hybrid machine learning predicts the lake ecosystem’s responses to climate change

Throughout the mid-20th century, the supply of phosphorus from detergents and fertilizers deteriorated the water quality of Lake Geneva in Switzerland, prompting officials to take action in the 1970s to remedy the pollution.

“The obvious remedy was to undo the phosphorus load, and this simple idea helped a lot, but it didn’t return the lake to its former state, and that’s the problem,” said George Sugihara, a biological oceanographer at the Center for Biotechnology. UC San Diego’s Scripps Institute. of Oceanography.

Sugihara, Ethan Deyle of Boston University and three international colleagues spent five years looking for a better way to predict Lake Geneva’s ecological response to the threat of phosphorus pollution, to which must now be added the effects of climate change. to manage. The team, including Damien Bouffard of the Swiss Federal Institute of Aquatic Sciences and Technology, will publish its new hybrid empirical dynamic modeling (EDM) approach on June 20 in the journal Proceedings of the National Academy of Sciences

“Nature is much more interconnected and interdependent than scientists would often like to think,” said Sugihara, the McQuown Chair Professor of Natural Science at Scripps. EDM can help in this context as a form of supervised machine learning, a way for computers to learn patterns and teach researchers about the mechanisms behind the data.

“You pull one lever and everything else changes, in a knife-like style. Single-factor experiments, the hallmark of 20th-century science where everything is held constant, can teach you a lot in principle, but it’s not how the world works,” he said.

“If it weren’t for this, if nature behaved more like the single-factor experiments and less connected and interdependent, we could predict outcomes with simple models in which relationships don’t change.”

Interdependence and changing relationships are the reality of ecosystems and they are also the reality of financial markets where forecasting is so challenging, Sugihara noted. EDM was honed in the melting pot of financial forecasting in the mid-1990s through the early 2000s, when Sugihara was a director at Deutsche Bank.

Sugihara has used his financial background for the past 20 years at Scripps to design market tools to support sustainable marine fisheries. He calls EDM ‘math without equations’.

But EDM is not a black box method, Deyle said, referring to quantitative methods based on mysterious mathematical or computational formulas. It is a criticism he says is often made of machine learning.

“Rather, it uses the data to tell you what’s going on in the most direct way, with minimal assumptions. What are the important variables? How do relationships change over time? It has a mechanism and transparency that comes straight from the data.”

What Sugihara’s team has attempted is a departure from the traditional modeling methods used over the decades. As Deyle points out, parts of the established models are represented by constants.

“The fixed and constant gravity, or the shape and depth of a lake, for example. Consequently, physical processes in the lake can be modeled very well with simple equations,” he said.

Not so for the changing ecology and biochemistry.

“The organisms that cause changes in an ecosystem like Lake Geneva have changed over the past two decades. The food web has changed and is constantly changing, along with the biochemistry of the lake,” Bouffard said.

“The standard tools are not suited to such problems,” said Deyle, who obtained her PhD in biological oceanography from Scripps Oceanography in 2015 with advisor Sugihara.

“Lake Geneva is one of the best-studied systems in the world. It’s no accident that it was an opportunity to push the boundaries with a machine learning approach to ecological forecasting,” Deyle said.

The authors show that their hybrid approach not only leads to a significantly better prediction, but also to a more useful description of the processes (such as biogeochemical and ecological) that determine water quality.

In particular, the hybrid model suggests that the impact on water quality of increasing air temperature by 3 degrees Celsius (5.4 degrees Fahrenheit) would be of the same order as the phosphorus pollution of the last century, and that best management practices may no longer be. a single control lever, such as reducing the phosphor input only.

“One of the intellectual cornerstones of it all is minimalism,” said Sugihara. “Getting information from data with as few assumptions as possible.”

A simple model that predicts target data yet to be collected is more persuasive than a complex model that may be in line with current thinking and remarkably adaptable to history, but which doesn’t really ‘predict’ events yet to be seen. This has been the biggest problem in financial applications, where it’s easy to find things that ‘fit’, but almost impossible to find something that really ‘predicts’.

“The more complicated something is, the easier it is to fool yourself,” he said. “Our hybrid approach seems to have a balance that works.”

This research was funded by the US Department of Defense Strategic Environmental and Development Program, National Science Foundation, Department of the Interior, the McQuown Fund and the McQuown Chair of Natural Sciences, UC San Diego. Study co-authors are Victor Frossard, Université Savoie Mont Blanc; Robert Schwefel and John Melack, University of California at Santa Barbara.

Leave a Comment

Your email address will not be published. Required fields are marked *