Strong spatial partitioning of mesozooplankton in a dynamic frontal zone shapes trophic transfer efficiency

This dataset contains the code and data that were used in the publication "Strong spatial partitioning of mesozooplankton in a dynamic frontal zone shapes trophic transfer efficiency" in Limnology & Oceanography. This paper was based on images collected with the mDPI (also commonly referred to as ISIIS-DPI) in the North Sea in June 2023. The mDPI images were processed via a newly developed segmentation routine and Convolutional Neural Network classification described in this manuscript. This dataset contains i) the training set of 13.417 North Sea zooplankton images of the mDPI on which the CNN was trained, ii) the Python code to run the image segmentation, the training and inferrence of the CNN and the code to reproduce the images of the manuscript, and iii) the data that is necessary to reproduce the figures and ecological results. To use the code, download the folder zip-archive 'code' and 'data', check the required libraries in requirements.txt and, if desired, modify the paths to data files in data_analysis_init.py. The figures were generated in the .ipynb-notebooks (the names of the notebooks referr to the figures numbers in the manuscript), and the imaging processing was executed in the scripts run_cnn.py, run_segmentation.py and train_cnn.py. In 'data', a sample of mDPI full-frames is provided in order to run the segmentation. Future updated versions of the code may be posted at the related github-repository. The data folder contains the trained CNN model, the flat-fielding backgrounds of the mDPI full-frames, and csv-files with a.o. the zooplankton densities linked to the environmental data along each transect, the calculated size of each ROI and several other files to generate the figures in the provides notebooks. In this manuscript, a total of 73 million Regions of Interest was used. Since this is ~2 TB of images, the full set of ROIs could not be shared here. Please contact the first author if you have a serious interest in the full set of ROIs.