File structure and contents

The file structure is as follows:

• info_results: This folder contains several csvs that summarize the results of the modeling exercises. They are:

• experimental_design_all.csv: describes the parameters used in each model run. The column "run_number" is a key ID that is used throughout these files to identify unique parameter sets. This file contains a column for each parameter that was calibrated in the GLUE analysis.

• all_rmses.csv: contains columns for run number, depth, maximum date (indicating whether a run completed), site, root mean squared error of volumetric water content at a given depth, percent difference between modeled and observed growing season evaporation at the site at which evaporation was available. This file contains results in the calibration period only for all parameter sets - not just those that were deemed acceptable.

• behavioral_params.csv: contains columns for run number and model performance in the calibration period for behavioral parameters only, with average values across sites.

• cal_val_objectives.csv: contains model fit statistics for each behavioral parameter set in the calibration and validation periods.

• future_summary.csv: contains a summary of the results of model simulations with downscaled GCM data. Columns are site name, run number (parameter set), GCM, year, growing season ET (mm), and mean growing season volumetric water content at the presumed rooting depth for that site.

• site folders: Each site (UP1A, Smith Lake 1, Smith Lake 2, and US-Rpf) have a folder containing model inputs and outputs. These files are formatted as SHAW model inputs and outputs. File formats, current SHAW distributions, and model documentation can all be obtained at: https://www.ars.usda.gov/pacific-west-area/boise-id/northwest-watershed-research-center/docs/shaw-model/. For each site, files include:

-site_name.wea: Gap-filled input meteorological data for SHAW model forcing. These are formatted as hourly inputs to the SHAW model and as such lack column headers. Variables are, in order: day, hour, year, air temperature (°C), wind speed (m/s), relative humidity (%), snow density (set to 0 and calculated by SHAW), and downward shortwave radiation (W/m2).

-site_name.inp: Input file for the SHAW model that gives file names to the model.

-site_name.sit: Input file for the SHAW model that gives parameters. These are best-estimate prior parameters based on site information and literature review.

-GFDL-CM3.wea and NCAR-CCSM4.wea: Downscaled CMIP5 GCM outputs for the collocated 20km grid cell, bias-corrected to available meteorological data for each site as described in the manuscript. Formatting is the same as the site meteorological data.

-run_xxxxxxxxx: These folders contain the inputs and outputs specific to a given model instantiation with a particular parameter set. The file name matches the run_number column in experimental_design_all.csv. In each of these folders is a set of SHAW model inputs and outputs for the given parameter set, using observed historical climate forcing. Within each folder is also a subfolder for each GCM, containing SHAW outputs when forced with data from that GCM. If file formats are unclear, please see SHAW documentation or contact the authors of this dataset.