The hidden value of low-performers
A new study in the journal Nature Scientific Reports explores ensemble design strategies in biogeochemical ocean modelling. It illustrates the benefits of combining of the “power of many” with Machine Learning by exploring simulated oxygen concentrations at a test site (Eckernförde Bight; Baltic Sea). Even seemingly weak or deliberately extreme model versions can contribute important insights: https://www.uni-kiel.de/en/details/news/099-value
