Within-night training data

This repository contains the data used for training and deploying gradient boosted trees that analyze the associations between bird migration intensity and predictor variables. By integrating radar (NEXRAD) data with weather and terrestrial factors within 2-hour segments, we are able to assess changes in these associations within a single migration night, as birds transition between terrestrial and atmospheric habitat. Due to the large size of the input datasets (hundreds of GB), we have only included a subset of the data for the year 2012.<br>We include code for analyzing these data as well as a suggested data storage structure on GitHub [https://github.com/a-khalighifar/migration_forecast/tree/main/within-night].<br><b>Description of files included</b><i>01_cell-coordinates/1_all-trainingPoints-fromStations.csv</i>: This csv includes training points for within the coverage of each NEXRAD radar station.Unique.ID - A numerical identifier for each training pointRadar Station - The four-letter abbreviation for each NEXRAD radar siteX - The longitude or easting coordinate of each pointY - The latitude or northing coordinate of each point*Note - we also include coordinates for points at varying distances and directions from each sampling point, denoted in the header title (e.g. X.75.N = longitude coordinate for point 75km north of a given sampling point).<i>02_Radar_Atm_combined_Annual/2012/2012.rds</i>: Full dataset of predictor variables for sampling points in 2012.<br>