Machine learning-based evidence and attribution mapping of 100,000 climate impact studies - Data

<p>Data for the paper&nbsp;Machine learning-based evidence and attribution mapping of 100,000 climate&nbsp;impact studies</p> <p><strong>Document Metadata</strong></p> <p>0c_doc_info.csv contains basic document metadata for each document considered in our study</p> <p><strong>Predictions</strong></p> <p>In each predictions file, 1 refers to a document hand-labelled as belonging to a category, and 0 refers to a document hand-labelled as not belonging to a category. All values in between are predicted values, where for values greater than 0.5, a document is considered likely to belong to the given category.</p> <p>1_document_relevance.csv contains the predicted relevance of a document to the study.</p> <p>1_driver_predictions.csv contains the predicted climate driver of each document.</p> <p>1_impact_predictions.csv contains the predicted impact type of each document</p> <p><strong>Geographical data</strong></p> <p>Place_df.csv contains a row for each geographical entity automatically extracted from each study</p> <p>Study_gridcell_2.5.csv contains a row matching each study with each grid cell covered by the study&rsquo;s smallest mentioned geographical entity</p> <p><strong>Merged data</strong></p> <p>2_study_da.csv contains a row for each study describing the aggregated detection and attribution characteristics of the grid cells the study refers to</p> <p>2_merged_da_data.csv contains a row for each grid cell describing the attribution categories and the number of weighted grid cells for each climate driver.</p>