Perturbed Parameter Ensemble of Dissolved Oxygen Simulations in Eckernförde Bight (Baltic Sea)

This dataset supplements the publication “The Hidden Value of Low-Performers: Ensemble Design Strategies for Coupled Ocean-Circulation Biogeochemical Modelling”, published in Nature Scientific Reports (https://doi.org/10.1038/s41598-026-54424-0). The dataset contains a perturbed-parameter ensemble of hindcast simulations of dissolved oxygen concentrations near the bottom of Eckernförde Bight (Baltic Sea), covering the period 2000–2015. The underlying coupled model (MOMBE, coupled with the oxygen module EckO2) simulates ocean circulation and oxygen dynamics in Eckernförde Bight at an ultra-high spatial resolution of 100 m. The model was specifically developed for Eckernförde Bight because of the recurrent mass fish-kill events triggered by low dissolved oxygen concentrations. The six model versions differ in their parameter settings for background mixing and local oxygen sources and sinks (Dietze & Löptien, 2021). This dataset serves as an example of a machine-learning-based post-processing approach applied to a perturbed-parameter ensemble in order to improve reliability and provide uncertainty estimates. The model simulations are provided together with the corresponding Python scripts and a selection of visualization scripts written in Python and Ferret (https://ferret.pmel.noaa.gov/Ferret/).