This repository contains all scripts and data files to reproduce the analysis of the manuscript "Competitive hierarchies in bryozoan assemblages mitigate network instability by keeping short and long feedback loops weak" Abstract Competitive hierarchies in diverse ecological communities have long been thought to lead to instability and prevent coexistence. However, system stability has never been tested and the relation between hierarchy and instability has never been explained in complex competition networks parameterised with data from direct observation. Here we test model stability of 30 multispecies bryozoan assemblages, using estimates of energy loss from observed interference competition to parameterise both the inter- and intraspecific interactions in the competition networks. We find that all competition networks are unstable. However, instability is mitigated considerably by asymmetries in the energy loss rates brought about by hierarchies of strong and weak competitors. This asymmetric organisation results in asymmetries in the interaction strengths, which reduces instability by keeping the weight of short (positive) and longer (positive and negative) feedback loops low. Our results support the idea that interference competition leads to instability and exclusion but demonstrate that this is not because of, but despite, competitive hierarchy. Data Our data set contains records of overgrowth competition in 30 high-latitude bryozoan assemblages. Rocks were collected by hand from shallow subtidal coastal locations at Rothera Island, West Antarctic Peninsula, Signy Island in the maritime Antarctic and Spitsbergen in the Arctic. For each assemblage, the data set contains one .csv file with abundance per species and one .csv file containing the species-contact-matrix. All bryozoans were identified to species and counted, giving abundance data in colonies per species. Then, all pairwise contests between colonies were classified as win, draw or loss and the results were compiled in the species-contact-matrices. For details, see the methods section of the paper. Analysis The analysis is subdivided into the following sections: 0 Random matrices: Stability of random matrices with symmetric and asymmetric interactions. 1 Preparation: Define functions to calculate asymmetry measures and set plotting parameters 2 Read and process raw data: Converts raw data to Jacobian matrices 3 Analysis of empirical matrices: Calculates stability, asymmetry measures, loop weights of empirical matrices. 4 Analysis of randomised matrices: Randomises empirical matrices and analyses the effect on stability, asymmetry measures and loop weights. 5 Sensitivity: Effect of model assumptions (cost-values / replacement of missing values) on the results. Details on how to reproduce the full analysis, including all figures and tables in the manuscript can be found in the ReadMe file.

This project was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under project number 451967415 (AL 2563/1-1)