Data for Three-dimensional distribution of groundwater residence time metrics in the glaciated United States using metamodels trained on general numerical simulation models

Residence time distribution (RTD) is a critically important characteristic of groundwater flow systems; however, it cannot be measured directly. RTD can be inferred from tracer data with analytical models (few parameters) or with numerical models (many parameters). The second approach permits more variation in system properties but is used less frequently than the first because large-scale numerical models can be resource intensive. With the data and computer codes in this data release users can (1) reconstruct and run 115 General Simulation Models (GSMs) of groundwater flow, (2) calculate groundwater age metrics at selected GSM cells, (3) train a boosted regression tree model using the provided data, (4) predict three-dimensional continuous groundwater age metrics across the Glacial Principal Aquifer, and (5) predict tritium concentrations at wells for comparison with measured tritium concentrations. The computer codes in this data release are in the form of Python scripts and Jupyter Notebooks. Users will need to have these Python resources installed on their computers to run the codes. Instructions for creating the Python environment can be found in the file Creating the Python environment.txt. Users who would rather not run the scripts but who wish to obtain the final data sets can do so by downloading the file Output--Predictions.7z. Users who wish to reproduce the data sets in this release can do so by downloading, unzipping, and running the data workflow in Starn_GW_Residence_Time_Data_and_Scripts.7z. The codes in this file use relative pathnames, so the directory structure within this file should not be changed. The ".7z" file extension indicates 7-Zip files, http://www.7-zip.org Executables--MODFLOW and MODPATH executable files provided for convenience. These are Windows 64-bit versions. Step 1--Create General Simulation Models--Codes to create 115 GSMs Step 2--Data preparation--Calculate residence time distributions at selected GSM cells Step 3--Metamodel training--Train a boosted regression tree metamodel (XGBoost) Step 4--Metamodel prediction--Predict age metrics throughout the Glacial Aquifer Step 5--Tritium simulation --Calculate tritium concentration at selected wells