Quantifying Interactive Effects of Fire and Precipitation Regimes on Catchment Biogeochemistry of Aridlands
This project develops a reproducible, database-centered workflow to quantify how wildfire history and hydrologic variability alter stream chemistry across arid and semi-arid catchments in the southwestern United States.
The current implementation focuses on:
- Harmonizing multi-source chemistry and discharge records into a common analytical framework.
- Linking stream sites to fire exposure through catchment-level and event-level spatial aggregation.
- Estimating pre/post-fire concentration-discharge behavior with progressively structured Bayesian models.
- Producing reusable data products that support synthesis across sites and ecoregions.
This repository is designed for scientific inference and data integration. It is not currently an operational real-time telemetry or early-warning platform.
flowchart LR
A[USGS and non-USGS chemistry and discharge data]
B[Spatial fire and catchment inputs<br/>MTBS, catchments, ecoregions, hydrography]
C[PostgreSQL and PostGIS firearea schema]
D[Standardization and aggregation functions and views<br/>usgs_water_chem_std, analyte views, ranges_agg]
E[Exported analysis tables and covariates]
F[R and Stan model development<br/>pre/post, delta, unified, hierarchical]
G[Figures, interpretation, and proposal-ready synthesis]
A --> C
B --> C
C --> D
D --> E
E --> F
F --> G
The workflow is anchored in the firearea schema in PostgreSQL with PostGIS enabled. Core functions standardize chemistry records, reconcile analyte units/forms, and integrate USGS and non-USGS data streams into analyte-specific views.
Spatial and temporal fire exposure is summarized through derived views and materialized views, including event windows and cumulative burn metrics. These products connect site records to fire-event timing and extent while preserving deterministic ordering for reproducible exports.
Curated tables are exported for downstream R and Stan workflows. Model development progresses from simple concentration-discharge structure to pre/post-fire, delta, unified, and hierarchical templates.
- wildfire_database: documentation-first SQL development in Quarto with extraction and execution workflows for building schema objects and export products.
- models: Stan templates that represent increasing structural complexity from baseline CQ models to hierarchical formulations.
- model_fitting: data preparation and model execution scripts for iterative fitting and diagnostics.
- hydrologic_distance: scripts to derive distance-based hydrologic context metrics.
- spatial_covars: spatial covariate extraction and preparation workflows.
- burn_severity: burn severity estimation workflows and supporting documentation.
- conceptual_figure_one, marss_figure_one, figures: synthesis and communication outputs.
This repository is complemented by the firearea R package. firearea provides reusable functions for watershed delineation, fire-perimeter preparation, flowline-buffer operations, and hydrologic-distance calculations that support upstream data preparation and downstream database integration.
SQL development follows a documentation-first pattern: SQL logic is authored in Quarto source files, extracted into executable SQL, and run against PostgreSQL. This design improves traceability between analytical documentation and database state.
Execution is managed through a lightweight task orchestration layer and environment-variable-driven database targeting, supporting development and production contexts while preserving consistent build semantics.
The implemented workflow provides a robust foundation for assessing wildfire impacts on aquatic chemistry at site and regional scales, including ecoregion-constrained analyses and standardized multi-source data integration.
For proposal development, the strongest contribution of the current system is its end-to-end reproducible architecture: harmonized chemistry/discharge data, fire-event linkage in the database, and statistically explicit pre/post-fire modeling. Future operational capabilities can build on this foundation, but are not represented here as deployed systems.
- models includes the following Stan model templates:
- STAN_lm_template.stan: basic linear model with structure
$y = mx + b$ . - STAN_lm_prepost_template.stan: pre/post-fire split with separate CQ slopes.
- STAN_lm_delta_template.stan: derives change in slope from pre/post estimates.
- STAN_lm_delta_unified_template.stan: unified structure estimating all parameters in one model.
- STAN_lm_hierarchical_template.stan: hierarchical site and cross-site structure (under development).
- model_fitting/STAN_dev_script.R prepares analysis inputs and fits the model sequence.
- Tamara Harms, Associate Professor, Environmental Sciences Department, University of California Riverside
- Heili Lowman, Research Scholar, Rhodes Information Initiative, Duke University
- Stevan Earl, Information Manager, Central Arizona-Phoenix Long-Term Ecological Research (CAP LTER)
- Ball, G. F., et al. (2021). Wildfire impacts to aquatic networks in the western United States. Study context for burned-stream-length and downstream propagation framing.
- Dahm, C. N., et al. (2015). Post-fire runoff effects on Rio Grande water quality and ecosystem processes. Case-study basis for episodic disturbance framing.
- Dennison, P. E., et al. (2014). Large wildfire trends in the western United States, 1984-2011. Baseline evidence for long-term fire activity trends.