Data and code for: Impact Curves Reveal Variable Post-Fire Hydrologic and Water Quality Responses Across the Conterminous United States

Paige Tunby and Ricardo González-Pinzón, University of New Mexico. Contact: gonzaric@unm.edu

This resource accompanies a manuscript in preparation for peer-reviewed publication. It contains: - the impact-curve and Area Under the Curve Ratio (AUCR) results for every fire-site pair and post-fire period - the fire-site pair and site tables - the global envelope test results (one-sided and in both directions) - the R code that implements the impact-curve algorithm, the circular block bootstrap, the AUCR and the global envelope test - the R code for every statistical analysis reported in the manuscript

The candidate fire-site pairs were identified with RIO-FINDER (Tunby and González-Pinzón, 2026, Environ. Model. Softw., 107063; HydroShare resource https://www.hydroshare.org/resource/81a7609e5a514b45aefcba032136e79c/). Raw high-frequency time series are available from the USGS National Water Information System (https://waterdata.usgs.gov/nwis). Fire perimeters and burn severity are available from Monitoring Trends in Burn Severity (https://www.mtbs.gov).

Contents

Data/

• AUC_metrics_FINAL/AUC_both_3YR_<code>.csv: the original pipeline output (AUC integrated over duration in minutes), one row per fire-site pair and post-fire period (6, 12, 24, 36 months), one file per USGS parameter code: 00060 discharge, 00010 water temperature, 00095 specific conductance, 00300 dissolved oxygen, 00400 pH, 63680 turbidity.
• Identifiers and period: Event_ID (MTBS fire), site_no (USGS gage), fire_group (single or multi-fire site), Month_Interval.
• Areas and ratio: AUC_Pre (area under the bootstrap-median pre-fire impact curve), AUC_Post, and Raw_PostOverPre (AUC_Post/AUC_Pre; for 00300 it is AUC_Pre/AUC_Post, the inverted ratio of Equation 3). AUCR = log10(Raw_PostOverPre).
• Other columns: n_points (number of duration windows), basin metrics (basin_area_km2, pct_basin_burned_full, pct_basin_Low/Moderate/High), Level I ecoregion (NA_L1NAME), Ig_Date, and Previous_fires.
• Columns to ignore: Valid_Intervals is a numeric artifact of concatenating the window minutes; the windows are listed in impact_curves_all_fire_site_pairs_long.csv. Path columns refer to the original analysis machine; the scripts use only the part after Data_Final/.
• AUC_metrics_LOGAXIS/AUC_both_3YR_<code>.csv: the AUCR used in the manuscript. Same rows and columns as AUC_metrics_FINAL, with AUC_Pre, AUC_Post and Raw_PostOverPre integrated over log10(duration), so that each decade of duration contributes equally; the original values, integrated over duration in minutes, are kept as AUC_Pre_linear, AUC_Post_linear and Raw_PostOverPre_linear, and Windows_used lists the windows (minutes). Produced by Code/analysis/00_log_axis_aucr.R. HPCC_ANALYSIS/AUC_summary_HPCC_logaxis.csv gives the same for the case study in Figures 3-4.
• impact_curves_all_fire_site_pairs_long.csv: the per-window impact-curve values behind every AUCR (47,330 rows).
• Interval_min: window length in minutes.
• Boot_Median, Boot_CI_Lower, Boot_CI_Upper: median and 2.5th/97.5th percentiles of 1,000 circular block bootstrap replicates of the pre-fire record (each replicate is one contiguous pre-fire block as long as the post-fire period).
• Comparison_Max: post-fire maximum of the moving-window mean (Comparison_Min for dissolved oxygen).
• Table_S2_fire_site_pairs.csv: the 785 fire-site-parameter combinations analyzed (Supplementary Table S2), with ignition date, ecoregion, basin area, percentage of the basin burned, and gage coordinates (WGS 84).
• sites_coordinates.csv: the 531 USGS gages, with names and coordinates.
• global_envelope_test/: results of the one-sided global envelope test (GET R package; continuous rank envelope, family-wise error control, α = 0.05), which compares each post-fire impact curve with the 1,000 bootstrapped pre-fire curves. Global_Envelope_Summary_PARAM_<code>.csv gives the result for every fire-site pair and post-fire period (global p-value, significance, and the windows outside the envelope), and Global_Envelope_Binomial_CI_ALL_PARAMETERS.csv gives the percentage of significant pairs with binomial confidence intervals (Supplementary Table S7 and Figure S6). The test in both directions (Supplementary Table S11) and its sensitivity to pre-fire record length (Supplementary Table S12) are in Code/analysis/outputs/.
• Data_Final/*run_log*.csv: bootstrap run logs, with the number of pre- and post-fire observations per fire-site pair.
• Data_Final/Boot_Output_00060/ and Data_Final/Boot_Output_MultiFire_00060/: bootstrap outputs for the discharge fire-site pairs shown in Figure 2 and Supplementary Figure S10 (Gallinas Creek near Montezuma, NM, 08380500; Gallinas River near Lourdes, NM, 08382000; Big Sur River near Big Sur, CA, 11143000). Files are organized as <Event_ID>/<site_no>/X_<code>_00000/; *_impact_summary_m<months>.csv holds the post-fire impact curve and the summary of the pre-fire bootstrap for one post-fire period, and *_iteration_log_shared_m<months>.csv holds the 1,000 bootstrapped pre-fire impact curves (one row per replicate, one column per window). For sites burned more than once, the shared pre-fire bootstrap is in PRE_SHARED/site_<site_no>/.
• HPCC_ANALYSIS/: the same bootstrap outputs and AUCR summaries for the Hermit's Peak-Calf Canyon case study at the Pecos River above Santa Rosa, NM (USGS 08382650; P5167 in the manuscript) shown in Figures 3-4.
• Saved_Variables/all_sites.rds: the RIO-FINDER site list, used for gage coordinates.

Code/analysis/

R scripts that reproduce the reported statistics and figures from the files in Data/. Each script locates Data/ relative to its own position, so it runs from any working directory, for example Rscript "Code/analysis/01_spearman_correlations.R". Outputs are written to Code/analysis/outputs/, which already contains the tables reported in the manuscript. Every script that does not need the full bootstrap iteration logs (all except 04, 10, and 14) was run from this resource and reproduced the tables and figures of the manuscript exactly.

The scripts were tested with R 4.6.1. They require the dplyr, readr, tidyr, purrr, stringr, ggplot2, patchwork, cowplot, scales, lme4, lmerTest, and sf packages, and GET for script 10.

Code/pipeline_original/

R Markdown pipeline used to process the raw time series and exist for methodological reference, but require the raw time series to run: 1. TimeseriesAnalysis.Rmd: cleaning and resampling. 2. Impact_curve_bootstrapping_code.Rmd: impact curves and the circular block bootstrap.

HPCCImpactCurvePlots.Rmd and HPCCImpactCurvePlots_Q.R cover the Hermit's Peak-Calf Canyon case study. These notebooks need the raw time series, and the file paths inside them point to the original analysis machine and must be edited.

Licenses

• Data (everything in Data/ and the tables in Code/analysis/outputs/): Creative Commons Attribution 4.0 International (CC BY 4.0). See LICENSE_DATA.txt.
• Code (everything in Code/, except the output tables): MIT License. See LICENSE_CODE.txt.

How to cite

If you use these data or code, please cite both: - This resource: Tunby, P., and González-Pinzón, R. (2026). Data and code for: Impact Curves Reveal Variable Post-Fire Hydrologic and Water Quality Responses Across the Conterminous United States. HydroShare. The DOI is shown in the citation on this resource's HydroShare page. - The associated article: Tunby, P., Van Horn, D., and González-Pinzón, R. Impact Curves Reveal Variable Post-Fire Hydrologic and Water Quality Responses Across the Conterminous United States. In preparation.