Supporting data and model archives for physically-informed models of streamflow correlation for the contiguous United States, 1981-2020

This data release provides an archive of random forest regressions predicting streamflow correlation as a function of climatic similarity over the contiguous United States, based on streamflow information from water years 1981 to 2020. This data release also contains the underlying datasets used to generate those regressions and a series of scripts that facilitate their exploration and use. The scripts expect a specific file and folder structure, which is described in the "streamflow_correlation_expected_folder_structure.txt' file.

rf_models.zip - contains random forest models for all 135 model nodes, stored as R (.rds) objects readable using the 'readRDS' function in R. Also includes a csv defining the training and validation gage-pairs used for each model node, and a csv giving the coordinates of the central XY point of each model node.
scripts.zip - R scripts designed to guide access and use of the supplied models. Includes an example analysis. Scripts require the datasets stored in the 'supporting datasets' child item arranged in a specific folder structure.
streamflow_correlation_expected_folder_structure.text - simple text file laying out expected folder/file structure for scripts to function as written.

Model input data are all stored in a child item of this data release ('Supporting datasets for physically-informed models of streamflow correlation'). These data are required for the scripts presented here to run properly.