This description outlines the dataset associated with the research paper "Discovering the Mesoscale for Chains of Conflict," published in PNAS Nexus (click here to view the article). Please cite this paper if you use our dataset. The dataset is divided into two main folders: "voronoi_grids": This folder contains 100 realizations of Voronoi tessellations over the African continent. These tessellations were generated during the research process detailed in the aforementioned paper. "avalanches": The "avalanches" folder comprises conflict avalanches derived from each of the realizations of the Voronoi tessellations. These avalanches were generated using the technique described in the research paper. Reading avalanches from the pickle files: The avalanches are stored in pickle files, and the following Python code can be used to read them: import dill as pickle # Temporal scale. Choose one of the following values: 2, 4, 8, 16, 32, 64, 128, 256, 512 time = 32 # Spatial scale. Choose one of the following values: 20, 28, 40, 57, 80, 113, 160, 226, 320, 453, 640, 905, 1280 dx = 453 # Voronoi grid index. Choose from 0 to 99 gridix = 3 file_path = f"avalanches/battles/gridix_{gridix}/te/te_ava_{str(time)}_{str(dx)}.p" with open(file_path, "rb") as f: ava = pickle.load(f) ava_event = ava["ava_event"] The extracted avalanches are in "event form," represented as a list of lists, where each element corresponds to the index of an individual conflict event in the ACLED dataset. Each sublist represents an avalanche. The total number of avalanches at the given temporal and spatial scale can be determined by the length of ava_event (len(ava_event)). ACLED Dataset: Due to copyright issues, we are providing a filtered version of the ACLED dataset that we used in our study. To obtain the full version of the ACLED dataset containing all the information regarding each conflict event, please refer to the final bullet point of the manual installation instructions of our python pacakge arcolanche.