README: Post-Fire Rainfall and Stream Response Data Release

Author: Megan Sears
Affiliation: Colorado State University, Department of Ecosystem Science and Sustainability
Contact: megan.sears@colostate.edu

Overview

This data release supports the analysis of rainfall and stream response in burned and unburned watersheds in northern Colorado. Data were collected following the Cameron Peak Fire (CPF) and East Troublesome Fire (ETF), both of which burned in 2020 and represent two of the largest wildfires in Colorado history.

The dataset includes in-situ tipping bucket rain gauge records, stream stage observations, spatially distributed rainfall products (MRMS, PRISM, GridMET), spatial data (catchments, DEM, burn severity), and derived metrics of rainfall-runoff response. Data span June–September of 2021, 2022, and 2023.

Repository Structure

1. TippingBucketEvents

Event-level precipitation summaries derived from tipping bucket rain gauges deployed across CPF and ETF watersheds. Each file represents one gauge station. A new event was defined when more than 0.3 mm of rainfall fell after at least six hours with no rainfall.

Variables:

2. StreamResponse

Event-level paired rainfall-runoff metrics linking streamflow response (binary: flow / no-flow) to co-occurring rainfall events. Files are organized by fire (CPF, ETF), rainfall source (MRMS radar or tipping bucket), and spatial aggregation method (maximum pixel or watershed-weighted mean).

2a. Rainfall-Runoff Event Files

Files follow the naming convention: {fire}_{source}_{aggregation}.csv

• {fire}: cpf or etf
• {source}: mrms (MRMS radar) or tb (tipping bucket)
• {aggregation}: max (maximum MRMS pixel over catchment) or wmean (area-weighted mean)

Variables:

2b. Kappa Statistics

Cohen's kappa coefficients quantifying agreement between MRMS-based and tipping-bucket-based flow/no-flow event detection, calculated per site. Files follow the naming convention: kappa_{metric}_{aggregation}.csv

• {metric}: mi30, mi60, or eventsum (rainfall metric used for event matching)
• {aggregation}: max or wmean

Variables:

Files: kappa_mi30_max.csv, kappa_mi30_wmean.csv, kappa_mi60_max.csv, kappa_mi60_wmean.csv, kappa_eventsum_max.csv, kappa_eventsum_wmean.csv

3. SpatialData

Spatial layers describing the study watersheds, sensor locations, terrain, and burn severity.

4. SpatialRainfall

Gridded seasonal and event-based rainfall products for the study area, covering June–September (JJAS) of 2021, 2022, and 2023.

4a. Gridded Seasonal Totals

4b. Probability of Exceedance

mi60_pover10.csv — Point-based MRMS values with probability of MI60 exceeding 10 mm/hr.

4c. Radar Quality Index (RQI) Stacks

stack_RQI_{year}.RData (2021, 2022, 2023) — R raster stack objects containing MRMS Radar Quality Index (RQI) values over the study area for each season. RQI ranges from 0 to 1; values ≥ 0.8 indicate high-quality radar estimates less influenced by beam blockage or range degradation. Load in R with load("stack_RQI_2021.RData").

5. PixelvsTB

Event-matched comparisons of MRMS pixel-level rainfall estimates against co-located tipping bucket gauge observations, used to evaluate MRMS accuracy across the study area.

Variables (CpfEtf_TBpixelCompare.csv):

Note: .x suffix = MRMS-derived values; .y suffix = tipping bucket-derived values. NA in tipping bucket columns indicates no matched gauge event detected.

6. DryCreekEvent

High-resolution time series for a focal precipitation-runoff event at Dry Creek (CPF), used for detailed event analysis.

Data Notes

• Temporal coverage: June–September 2021, 2022, and 2023 for most datasets; tipping bucket records extend back to 2020 for some stations.
• MRMS filtering: MRMS precision is 0.01 (_fil01 suffix) and values less than that were set to 0.
• Event delineation: Rainfall events were separated using a minimum inter-event period (6 hours). Streamflow events were matched to rainfall events using overlapping time windows.
• Spatial reference: All rasters are in WGS84 (EPSG:4326) or a projected coordinate system — see individual .prj files and raster metadata.
• Burn severity: Rasters are derived from dNBR (differenced Normalized Burn Ratio) calculated from Landsat imagery, obtained from MTBS (Monitoring Trends in Burn Severity).