During the 20th century, the water quality of Switzerland's LakeGeneva was degraded due to pollution from detergents andfertilizers.
The obvious remedy was to reverse the phosphorus loading, and this simple idea helped enormously, but it didn't return the lake to its former state.
Five years of searching for a better way to forecast and manage the lake's ecological response to the threat of pollution resulted in a new approach. The team, including Damien Bouffard of the Swiss Federal Institute of Aquatic Sciences and Technology, published their new hybrid empirical dynamic modeling approach in the journal.
The McQuown Chair Professor of Natural Science at Scripps stated that nature is more interdependent than scientists think. Supervised machine learning is a way for computers to learn patterns and teach researchers about the data.
It's whack-a-mole style when you pull one lever. He said that single-factor experiments can teach you a lot in principle, but it is not how the world works.
If nature was less connected and interdependent, we would be able to predict outcomes with simple models.
Interdependence and changing relationships are both realities of financial markets where prediction is difficult. In the late 1990s and early 2000s, Sugihara was a managing director atDeutsche Bank.
For the last 20 years, Sugihara has designed market tools to support sustainable marine fisheries. He calls it math without equations.
Deyle said that the method is not a black box method. He says that it's a criticism of machine learning.
The data is used to tell you what is happening with minimal assumptions. The important variables are what they are. The relationships change over time. There is a mechanism that comes from the data.
Traditional modeling methods have been used in the past. The models are represented by constants.
The shape and depth of a lake is an example. Simple equations can be used to model physical processes in the lake.
It's not possible for the ecology and biochemistry to change.
Over the last two decades, the organisms that drive change in the lake have changed. Bouffard said that the food web and the lake have changed.
According to Deyle, the standard tools are not suited for such problems.
One of the most well- studied systems in the world is LakeGeneva. Deyle said that it was an opportunity to push the envelope with a machine- learning approach to ecological forecasting.
The authors show that their hybrid approach leads to better prediction and a more actionable description of the processes that drive water quality.
The hybrid model suggests that raising air temperature by 3 degrees Celsius would have the same effect on water quality as the previous century's pollution, and that best management practices may no longer involve a single control lever.
One of the intellectual cornerstones of this is simplicity. Information can be pulled out of data with the least assumptions.
A simple model that predicts target data yet to be collected is more convincing than a complex model that agrees with current thinking but does not actually predict events yet to be seen. It is easy to find things that fit, but difficult to find things that actually predict.
He said that the easier it is to fool yourself, the more complex it is. The hybrid approach seems to work.
The co-authors of the study are VictorFrossard, Université Savoie Mont Blanc.
More information: A hybrid empirical and parametric approach for managing ecosystem complexity: Water quality in Lake Geneva under nonstationary futures, Proceedings of the National Academy of Sciences (2022). DOI: 10.1073/pnas.2102466119. Journal information: Proceedings of the National Academy of Sciences
All Rights Reserved © 2024
