RSFME: An Assessment of Annual Load Estimation Methods in Small Watersheds for Cross Site Comparisons

Authors: Nic Gubbins (Colorado State University), Weston M. Slaughter (UMD), Michael J. Vlah (Duke), Spencer Rhea (Duke), William H. McDowell (UNH/FIU), Emily S. Bernhardt (Duke), Matthew R.V. Ross (CSU)

Target journal: Hydrology and Earth System Sciences (HESS)

Contact: gubbins@colostate.edu

Overview

This repository contains the code and data pipeline for evaluating four common solute load estimation methods (linear interpolation, Beale ratio, rating, composite) by coarsening high-frequency sensor data from three monitoring networks:

• Hubbard Brook Experimental Forest (HBEF) — 15-min sensor data, Watershed 3
• Plynlimon Research Catchments — 7-hour auto-sampler data, Upper Hafren
• NEON — Daily sensor data from 6 first-order stream sites

The methods are then applied to the MacroSheds synthesis dataset (~93 sites, 112 solutes), producing a publicly available dataset of 16,489 site-years of annual load estimates.

Repository Structure

RSFME/
├── source/                      # Core analysis functions
│   ├── flux_methods.R           #   Load estimation methods (PW, Beale, Rating, Composite)
│   ├── plot_theme.R             #   Shared ggplot2 theme and palettes
│   └── calculate_annual_flux.R  #   MacroSheds annual load computation
├── data/                        # All input and intermediate data (see data/README.md)
├── paper/
│   ├── source/                  # Numbered analysis scripts (01–12)
│   │   └── 00_run_all.R         #   Runner script for full pipeline
│   └── figures/                 # All output figures (PNGs)
└── README.md                    # This file

Reproducing the Analysis

Prerequisites

• R 4.4+
• Required packages: tidyverse, here, feather, lfstat, lubridate, RiverLoad, zoo, patchwork, EGRET, macrosheds, imputeTS

Data Setup

1. MacroSheds data: Download EDI package edi.1262.2 from the EDI Data Portal and unzip into data/macrosheds/.
2. HBEF sensor data: Place w3_sensor_wdisch.feather in the repository root. This file is proprietary and not included — contact the lead author for access.
3. Other data files are included in the repository under data/.

See data/README.md for detailed provenance of each data file.

Running

From the repository root in R:

source("paper/source/00_run_all.R")

To resume from a specific script (e.g., skip the long-running analysis scripts):

Rscript paper/source/00_run_all.R 5

Scripts 01, 04, and 06 run coarsening analyses with 100 repetitions each and may take 30+ minutes. Scripts 02–03, 05, 07–12 generate figures and complete in seconds.

Outputs

• Figures are written to paper/figures/, named by figure number (e.g., fig07_hbef_ca_coarsening.png).
• Intermediate data (coarsening results) are written to subdirectories of data/.
• Annual load estimates are written to data/load_annual.csv.

Script Run Order

#	Script	Description
01	01_coarsen_analysis_hbef.R	HBEF coarsening experiment (Ca, NO3)
02	02_coarsen_figure_hbef.R	HBEF coarsening figures (Figs 7–8)
03	03_coarsen_example_figure.R	Coarsening example illustration (Fig 6)
04	04_coarsen_analysis_plynlimon.R	Plynlimon coarsening experiment
05	05_coarsen_figure_plynlimon.R	Plynlimon coarsening figures (Figs 9–10)
06	06_coarsen_analysis_neon.R	NEON coarsening experiment
07	07_coarsen_figure_neon.R	NEON coarsening figures (supplement)
08	08_macrosheds_compare.R	MacroSheds method comparison figures
09	09_macrosheds_descriptive.R	MacroSheds load distribution figure
10	10_ca_correlation.R	Ca–SpCond regression (helper, sourced by 11)
11	11_misc_figures.R	Raw data and C:Q plots (Figs 2–5)
12	12_hbef_method_comparison.R	HBEF method comparison (Fig 11)

License

Data availability and usage terms are described in the paper's Data Availability section.