California Bridge Vulnerability Index (BVI)

PCA/PFA-based Bridge Vulnerability Index for California bridges, integrating NBI inventory data, displacement susceptibility, and spaceborne monitoring availability.

How to cite

This repository contains code required to reproduce results for the paper entitled "Integrating structural and social vulnerability for equitable bridge maintenance prioritisation" published in International Journal of Disaster Risk Reduction (DOI https://doi.org/10.1016/j.ijdrr.2026.106115).

Research data: 10.5281/zenodo.22756091.

Data availability

Input data, intermediate files, and pipeline outputs that reproduce the paper are published on Zenodo (archive folders external/, intermediate/, and outputs/). Source URLs and licenses for third-party inputs are described in that dataset.

DOI: 10.5281/zenodo.22756091

Download and extract the archive, then set data_root to the extracted directory (see Configuration). Folder names match config/paths.example.yaml (external/, intermediate/pfa/, outputs/figures/, outputs/tables/, and so on). Required file names are listed under Required inputs.

Installation

Requires Python 3.10.

Poetry (local pipeline — sufficient for most users):

poetry install

Conda (same default dependencies, Python 3.10):

conda env create -f environment.yml
conda activate bridge-vulnerability-california

Optional HPC extras — only needed to regenerate OSM bridge line shapefiles on a cluster with MPI. With Poetry:

poetry install --with hpc

With conda, additionally install mpi4py, osmnx, and utm (for example from conda-forge). Most users should skip this and use pre-generated OSM line files configured in config/paths.yaml.

Optional formatting hooks for contributors:

poetry install --with dev
pre-commit install

Configuration

File paths are defined in config/paths.yaml. Analysis defaults (thresholds, PCA settings, Monte Carlo sample size, plot DPI, and related constants) are defined in config/params.yaml. Change values there rather than in scripts.

Setup (first time):

cp config/paths.example.yaml config/paths.yaml
# Edit config/paths.yaml and set data_root to the extracted Zenodo directory
# (or your local copy of that layout)

Quick override (without editing the file):

export BVI_DATA_ROOT=/path/to/extracted_zenodo_archive

Default folders are undated and match the published archive (intermediate/pfa/, intermediate/sensitivity/, outputs/figures/, outputs/tables/). To keep successive local runs in separate folders, add {run_date} to those paths in your paths.yaml and set:

export BVI_RUN_DATE=08_01_2026

HPC override (set in your Slurm script):

export BVI_DATA_ROOT=/path/to/data_root

Custom config files:

export BVI_PATHS_FILE=/path/to/my_paths.yaml
export BVI_PARAMS_FILE=/path/to/my_params.yaml

Pipeline scripts also accept the same settings as flags (no file edit needed):

poetry run python -m bridge_vulnerability.index.build_bvi \
  --data-root /path/to/extracted_zenodo_archive \
  --log-level INFO

Common flags on every step: --data-root, --paths-file, --params-file, --log-level, --dry-run (print resolved paths and exit). --run-date only applies if your paths YAML contains {run_date}. Monte Carlo additionally accepts --analysis-type, --n-samples, --batch-size, and --plot-only (replot Figs 8–9 from intermediate CSVs without re-running sampling). OSM MPI extraction accepts --min_range and --chunk-size.

In Python scripts or notebooks:

from bridge_vulnerability.config.paths import paths, p

nbi_csv = paths.intermediate.nbi.bridges_enriched_csv
index_dir = paths.intermediate.index.dir
figures_dir = paths.outputs.figures.dir
tables_dir = paths.outputs.tables.dir
county_shp = p("external", "auxiliary", "counties_shp")

Paths are grouped by role in config/paths.yaml:

Section	Meaning
`external`	Third-party or manual data — must exist before running the pipeline
`intermediate`	Files produced by one step and consumed by later steps (including glue CSVs under `intermediate/pfa/` and Monte Carlo dumps under `intermediate/sensitivity/`)
`outputs`	Paper artifacts only: `outputs/figures/` (`fig.png` + paired `fig.csv`) and `outputs/tables/` (`table*.csv`)

Required inputs

Before running the pipeline, set data_root in config/paths.yaml (or BVI_DATA_ROOT) to the extracted Zenodo directory. Paths match config/paths.example.yaml.

External data (download or obtain)

Path under `data_root`	Used in
`external/California_brdgs/NBI/2024del/CA24.txt`	Step 1 — raw NBI inventory
`external/California_county_codes/county_codes.csv`	Step 6 — county FIPS / names
`external/California_county_GDP_2023/California_GDP_manually_extracted.xlsx`	Steps 6, 11 — county GDP
`external/California_borders/ca_counties/CA_Counties.shp` (+ sidecars)	Steps 8–9 — county boundaries
`external/CDC_Social_Vulnerability_2022/California_county_2022.csv`	Step 8 — CDC SVI
`external/California_subsidence/vertical_displacement_Govorcin_paper/CA_VLM.tif`	Step 4 — displacement raster
`external/ps_density/merged_california.tif`	Step 5 — PS density (see External dependencies)

Bundled intermediate (skip HPC steps 2–3b)

These are not produced by the local pipeline. Use the published data bundle or regenerate on HPC:

Path under `data_root`	Used in
`intermediate/HPC_output/combined_nbi_lines.shp` (+ sidecars)	Step 5 — displacement zonal stats
`intermediate/HPC_output/nbi_segments_segment_1.shp` … `nbi_segments_segment_5.shp`	Step 5 — PS density zonal stats

Displacement susceptibility (manual if not bundled)

Step 5 reads intermediate/California_subsidence/vertical_displacement_Govorcin_paper/CA_VLM_fixed_filled2px.tif. The data bundle includes this file. To recreate: run step 4, apply 2-pixel QGIS interpolation on CA_VLM_fixed.tif, and save as CA_VLM_fixed_filled2px.tif.

Run the full pipeline (local)

For a single end-to-end run of all non-HPC steps (1, 4–11, including both Monte Carlo analyses), use:

poetry install
cp config/paths.example.yaml config/paths.yaml   # edit data_root
./run_full_pipeline.sh

The script checks required inputs, runs each step in order, and stops with QGIS instructions if the susceptibility raster is missing. Steps 2–3b (MPI OSM extraction and segment division) are skipped — pre-generated OSM shapefiles must already be present.

See the header of run_full_pipeline.sh for the same input checklist in plain text.

Running the pipeline

Run every script from the repository root with Poetry:

poetry run python -m bridge_vulnerability.<package>.<module>

After step 1, steps 2–3 and step 4 (displacement) can run in parallel. Step 3b requires step 3. Step 5 requires step 3b outputs (or bundled shapefiles) and the displacement susceptibility raster. Steps 6–11 depend on earlier outputs as noted below.

NBI inventory

Step 1. data_prep.nbi_read — Loads raw California NBI text data, filters tunnels and culverts, converts coordinates, and writes bridge tables with geometry.

Input: external.nbi.raw_dir / external.nbi.raw_file (raw NBI text file, e.g. CA24.txt). Output: intermediate.nbi.bridges_csv, bridges_geo_csv, and bridges_shp.

poetry run python -m bridge_vulnerability.data_prep.nbi_read

OSM bridge lines — optional (HPC only; can run in parallel with step 4)

Most users should skip steps 2–3b. The published data bundle includes OSM bridge shapefiles in {data_root}/intermediate/HPC_output/ (paths relative to data_root in config/paths.yaml). Step 5 reads these files directly:

File in `intermediate/HPC_output/`	`paths.yaml` key	Used in
`combined_nbi_lines.shp`	`intermediate.osm.combined_lines`	Step 5 — displacement zonal stats
`nbi_segments_segment_1.shp` … `nbi_segments_segment_5.shp`	`intermediate.osm.segments_pattern`	Step 5 — PS density zonal stats

If you are regenerating lines on HPC (poetry install --with hpc), run steps 2–3b below. They write chunked shapefiles, combined layers, and five segment shapefiles under intermediate.osm.hpc_output_dir. If your output folder or filenames differ from the defaults above, update the intermediate.osm.* keys in config/paths.yaml.

Step 2. data_prep.osm_extract_mpi — Extracts OpenStreetMap road lines for each bridge using MPI on an HPC cluster (submit hpc/run_osm_line_extraction.sbatch).

Input: intermediate.nbi.bridges_geo_csv. Output: chunked shapefiles in intermediate.osm.hpc_output_dir (nbi_lines_<start>_<end>.shp, nbi_polygons_<start>_<end>.shp).

srun python -m bridge_vulnerability.data_prep.osm_extract_mpi --min_range 0

Step 3. data_prep.osm_combine_lines — Merges chunked OSM line and polygon shapefiles from the HPC output folder into combined layers.

Input: intermediate.osm.hpc_output_dir (nbi_lines_*.shp, nbi_polygons_*.shp from step 2). Output: intermediate.osm.combined_lines and intermediate.osm.combined_polygons.

poetry run python -m bridge_vulnerability.data_prep.osm_combine_lines

Step 3b. data_prep.divide_into_segments — Divides each bridge centerline into five along-length segments for PS density zonal statistics.

Input: intermediate.osm.combined_lines, intermediate.osm.combined_polygons. Output: intermediate.osm.centerlines (nbi_segments.shp) and intermediate.osm.segments_pattern (nbi_segments_segment_1.shp … _5.shp).

poetry run python -m bridge_vulnerability.data_prep.divide_into_segments

Displacement rasters (can run in parallel with steps 2–3)

Step 4. data_prep.displacement_fix_nodata — Replaces NaN nodata values in the Govorcin VLM displacement raster with a fixed placeholder value (999) for downstream zonal statistics.

Input: external.displacement.vlm_raster (e.g. CA_VLM.tif). Output: intermediate.displacement.fixed_raster (e.g. CA_VLM_fixed.tif).

poetry run python -m bridge_vulnerability.data_prep.displacement_fix_nodata

Displacement susceptibility raster (required for step 5)

Step 5 reads **intermediate.displacement.susceptibility_raster** (default: CA_VLM_fixed_filled2px.tif).

Provided in the data bundle: This file is included under {data_root}/intermediate/California_subsidence/vertical_displacement_Govorcin_paper/ so most users do not need to recreate it.

How it was produced: Step 4 converts missing VLM pixels (NaN) to a fixed nodata value of 999 (CA_VLM_fixed.tif). The susceptibility raster adds 2-pixel interpolation in QGIS to fill remaining gaps so displacement values exist over all bridge locations — including areas where the original VLM data had no coverage due to loss of coherence over water or vegetated areas.

From the paper: “For this work, interpolation was applied to the VLM dataset to fill missing data pixels and ensure coverage over all bridges not covered by the original data due to loss of coherence over water or vegetated areas.”

To recreate from scratch: Run step 4, open CA_VLM_fixed.tif in QGIS, apply 2-pixel interpolation to fill nodata pixels, and save the result as CA_VLM_fixed_filled2px.tif (or update intermediate.displacement.susceptibility_raster in config/paths.yaml).

Bridge-level raster extraction

Step 5. data_prep.extract_bridge_raster_stats — Computes zonal statistics of displacement susceptibility and PS density along bridge lines and writes per-bridge CSV/shapefile outputs.

Input: intermediate.osm.combined_lines, intermediate.osm.segments_pattern (segments 1–5), intermediate.displacement.susceptibility_raster, external.ps_density.raster. Output: intermediate.bridge_lines.displacement_csv, monitoring_csv, and combined_csv.

poetry run python -m bridge_vulnerability.data_prep.extract_bridge_raster_stats

NBI enrichment

Step 6. data_prep.nbi_enrich — Joins bridge inventory with county names, county GDP, and keeps only bridges that have OSM line statistics from step 5.

Input: intermediate.nbi.bridges_geo_csv, intermediate.bridge_lines.displacement_csv, external.auxiliary.county_codes_csv, external.auxiliary.county_gdp_xlsx. Output: intermediate.nbi.bridges_enriched_csv (e.g. nbi_bridges_geo_county.csv).

poetry run python -m bridge_vulnerability.data_prep.nbi_enrich

Required for steps 7–10: downstream scripts (build_bvi, bridge_count_maps, monte_carlo) read intermediate.nbi.bridges_enriched_csv, which only this step produces.

Build BVI (PCA/PFA)

Step 7. index.build_bvi — Merges enriched NBI data with displacement/monitoring stats, aggregates to county level, and runs PCA/PFA to produce vulnerability weights and scaled indicators.

Input: intermediate.nbi.bridges_enriched_csv, intermediate.bridge_lines.combined_csv. Output: glue CSVs in intermediate.index (county_level_df.csv, final_weights.csv, full_county_scaled.csv); paper tables 7–9 in outputs.tables.dir; Figs 4–5 (fig4_*.png / fig5_*.png plus matching .csv) in outputs.figures.dir.

poetry run python -m bridge_vulnerability.index.build_bvi

Result maps

Step 8. plotting.vulnerability_maps — Produces county-level vulnerability maps, bivariate plots, and indicator contribution figures from BVI outputs.

Input: intermediate.index (county_level_df.csv, full_county_scaled.csv, final_weights.csv), external.auxiliary.counties_shp, external.auxiliary.cdc_svi_csv. Output: paper maps as PNG in outputs.figures.dir (each with a same-name .csv); intermediate.index.weighted_subindicators_csv and bivariate_bins_csv for later steps.

poetry run python -m bridge_vulnerability.plotting.vulnerability_maps

Step 9. plotting.bridge_count_maps — Maps the number of bridges per county (only needs enriched NBI from step 6; can run before step 7).

Input: intermediate.nbi.bridges_enriched_csv, external.auxiliary.counties_shp. Output: outputs.figures.dir/fig2_Number of Bridges_map.png and matching .csv.

poetry run python -m bridge_vulnerability.plotting.bridge_count_maps

Sensitivity analysis

Step 10. sensitivity.monte_carlo — Runs Monte Carlo / Sobol sensitivity analysis (slow; re-runs the index many times). Each analysis can take up to a few hours. The script supports two separate analyses; each run executes only one of them:

`analysis_type`	What varies	What stays fixed
`"threshold"` (default)	Bridge-indicator thresholds (`ADT_THRESHOLD`, detour miles, waterway/scour ratings, displacement)	Base PCA settings in `base_pca_params`
`"pca"`	PCA/index settings (capping, skew/variance thresholds, aggregation method, Mahalanobis cutoff, dropped indicator)	Default thresholds in `default_thresholds`

Run both analyses to get the full sensitivity picture. Either set BVI_ANALYSIS_TYPE or pass --analysis-type (no file edit needed):

# 1) Threshold sensitivity
poetry run python -m bridge_vulnerability.sensitivity.monte_carlo
# or: BVI_ANALYSIS_TYPE=threshold poetry run python -m bridge_vulnerability.sensitivity.monte_carlo

# 2) PCA / index sensitivity
poetry run python -m bridge_vulnerability.sensitivity.monte_carlo --analysis-type pca

# Replot Figs 8–9 from saved intermediate CSVs (no resampling)
poetry run python -m bridge_vulnerability.sensitivity.monte_carlo --analysis-type threshold --plot-only
poetry run python -m bridge_vulnerability.sensitivity.monte_carlo --analysis-type pca --plot-only

run_full_pipeline.sh runs both automatically.

Input: intermediate.nbi.bridges_enriched_csv (re-loaded internally for a full run), intermediate.index.weighted_subindicators_csv from step 8 (reference ranks for plots).

Output: ranking dumps in intermediate.sensitivity.dir (for replotting), tagged by analysis_type:

sensitivity_results_threshold_rankings.csv / sensitivity_results_pca_rankings.csv
sobol_indices_threshold_rankings_wide.csv / sobol_indices_pca_rankings_wide.csv

Paper Figs 8–9 (fig8a–fig8c for threshold, fig9a–fig9c for PCA, each with a paired .csv) are written to outputs.figures.dir. Use --plot-only to regenerate those figures from the intermediate CSVs without re-running Monte Carlo.

Validation

Step 11. validation.benchmark_poor_bridges — Compares county BVI rank with poor-bridge replacement-cost rank (requires step 8 for bivariate_bins.csv).

Input: intermediate.index.bivariate_bins_csv, external.auxiliary.county_gdp_xlsx. Output: outputs.tables.table10_csv (table10_bvi_poorbrdgs_benchmarking.csv: BVI vs traditional replacement-cost rank, with BVI/SVI categories).

poetry run python -m bridge_vulnerability.validation.benchmark_poor_bridges

Pipeline overview

flowchart TD
    A[NBI raw data] --> B[nbi_read]
    B --> E[osm_extract_mpi]
    E --> F[osm_combine_lines]
    G[VLM raster] --> H[displacement_fix_nodata]
    H --> H2[QGIS 2px interpolation]
    I[PS density raster] --> J[extract_bridge_raster_stats]
    F --> J
    H2 --> J
    J --> C[nbi_enrich]
    J --> K[build_bvi]
    J --> N[monte_carlo]
    C --> K
    K --> L[vulnerability_maps]
    C --> M[bridge_count_maps]
    C --> N
    L --> N
    L --> P[benchmark_poor_bridges]

Repository structure

src/bridge_vulnerability/
├── config/                    # Path configuration loader
├── data_prep/                 # Steps 1–6: ingest and spatial extraction
├── index/                     # Step 7: PCA/PFA index construction
├── plotting/                  # Steps 8–9: result maps
├── sensitivity/               # Step 10: Monte Carlo / Sobol analysis
│   └── helpers/               # Plotting functions used by monte_carlo (not run directly)
├── validation/                # Step 11: poor-bridges benchmark
└── utils/                     # Shared helpers (OSM, monitoring class, PCA, plotting)

config/paths.example.yaml      # Path template (committed)
config/paths.yaml              # Your local paths (gitignored)
config/params.yaml             # Analysis defaults (thresholds, PCA, Monte Carlo)
run_full_pipeline.sh           # End-to-end local pipeline (steps 1, 4–11)
hpc/                           # Slurm batch scripts for MPI OSM extraction
environment.yml                # Conda environment (Python 3.10)

Modules that are not run directly

These are imported by pipeline scripts; do not invoke them with poetry run python -m ...:

Location	Purpose
`sensitivity/helpers/monte_carlo_plots.py`	Visualization helpers for `monte_carlo`
`data_prep/bridge_divisions.py`	Geometry helpers for `divide_into_segments`
`utils/osm_bridge_lines.py`	OSM line/polygon extraction logic for `osm_extract_mpi` (requires `--with hpc`)
`utils/monitoring_class.py`	Spaceborne monitoring classification for `extract_bridge_raster_stats`
`utils/pca.py`	PCA/PFA utilities for `build_bvi` and `monte_carlo`
`utils/plotting.py`	Map styling helpers for `vulnerability_maps` and `bridge_count_maps`
`config/paths.py`	Loads `config/paths.yaml`
`config/params.py`	Loads `config/params.yaml`
`config/cli.py`	Shared command-line flags for pipeline scripts

Pipeline I/O reference

Step	Script	Reads	Writes
1	`nbi_read`	`external.nbi.*`	`intermediate.nbi.bridges_csv`, `bridges_geo_csv`, `bridges_shp`
2–3	`osm_extract_mpi` / `osm_combine_lines`	`intermediate.nbi.bridges_geo_csv`, `intermediate.osm.hpc_output_dir/nbi_lines_*.shp` (step 3)	`intermediate.osm.hpc_output_dir/nbi_lines_.shp`, `nbi_polygons_.shp`, `combined_nbi_lines.shp`, `combined_nbi_polygons.shp`
3b	`divide_into_segments`	`intermediate.osm.combined_lines`, `combined_nbi_polygons`	`nbi_segments.shp`, `nbi_segments_segment_*.shp`
4	`displacement_fix_nodata`	`external.displacement.vlm_raster`	`intermediate.displacement.fixed_raster`
—	QGIS 2px interpolation (manual)	`intermediate.displacement.fixed_raster`	`intermediate.displacement.susceptibility_raster`
5	`extract_bridge_raster_stats`	`external.ps_density.`, `intermediate.displacement.susceptibility_raster`, `intermediate.osm.`	`intermediate.bridge_lines.*`
6	`nbi_enrich`	`external.auxiliary.*`, `intermediate.nbi.bridges_geo_csv`, `intermediate.bridge_lines.displacement_csv`	`intermediate.nbi.bridges_enriched_csv`
7	`build_bvi`	`intermediate.nbi.bridges_enriched_csv`, `intermediate.bridge_lines.combined_csv`	`intermediate.index/*`, `outputs.tables/table7–9`, `outputs.figures/fig4–5`
8	`vulnerability_maps`	`intermediate.index/`, `external.auxiliary.`	`outputs.figures/fig3, fig6–7, fig10–14`; `intermediate.index/weighted_subindicators.csv`, `bivariate_bins.csv`
9	`bridge_count_maps`	`intermediate.nbi.bridges_enriched_csv`, `external.auxiliary.counties_shp`	`outputs.figures/fig2_*`
10	`monte_carlo` (run twice: `analysis_type` `"threshold"` then `"pca"`; `--plot-only` to skip sampling)	enriched NBI + `intermediate.index.weighted_subindicators_csv` (from step 8)	`intermediate.sensitivity/_rankings.csv`; `outputs.figures/fig8–9`
11	`benchmark_poor_bridges`	`intermediate.index.bivariate_bins_csv`, `external.auxiliary.county_gdp_xlsx`	`outputs.tables.table10_csv`

Paper figures

Paper figures live in outputs/figures/; tables in outputs/tables/. Each figure is a PNG with a sibling CSV of the plotted values. Minimum paper-only sequence: steps 7 → 8 → 9 → 10 → 11 (step 9 does not need step 7).

Paper figure / table	Output file	Folder	Step	Script
Fig. 2	`fig2_Number of Bridges_map.png`	`outputs.figures.dir`	9	`bridge_count_maps`
Fig. 3a–i	`fig3a_Traffic load.png` … `fig3i_Lack of monitoring.png`	`outputs.figures.dir`	8	`vulnerability_maps`
Fig. 4	`fig4_correlation_matrix_scaled.png`	`outputs.figures.dir`	7	`build_bvi`
Fig. 5	`fig5_residual_correlations.png`	`outputs.figures.dir`	7	`build_bvi`
Fig. 6	`fig6_infrastructure_vulnerability.png`	`outputs.figures.dir`	8	`vulnerability_maps`
Fig. 7	`fig7_dominant_sub_indicator_group.png`	`outputs.figures.dir`	8	`vulnerability_maps`
Fig. 8a–c	`fig8a_rankings_uncertainty_threshold.png`, `fig8b_sobol_absolute_variance_decomposition_threshold.png`, `fig8c_sobol_total_effects_threshold.png`	`outputs.figures.dir`	10	`monte_carlo --analysis-type threshold`
Fig. 9a–c	`fig9a_rankings_uncertainty_pca.png`, `fig9b_sobol_absolute_variance_decomposition_pca.png`, `fig9c_sobol_total_effects_pca.png`	`outputs.figures.dir`	10	`monte_carlo --analysis-type pca`
Fig. 10a–d	`fig10a_social_vulnerability_Socioeconomic status.png` … `fig10d_social_vulnerability_Housing type & transportation.png`	`outputs.figures.dir`	8	`vulnerability_maps`
Fig. 11	`fig11_social_vulnerability_Social Vulnerability Index (SVI).png`	`outputs.figures.dir`	8	`vulnerability_maps`
Fig. 12	`fig12_scatter_social_vs_bridge_vulnerability_subplots.png`	`outputs.figures.dir`	8	`vulnerability_maps`
Fig. 13	`fig13_scatter_matrix_bridge_vs_social.png`	`outputs.figures.dir`	8	`vulnerability_maps`
Fig. 14	`fig14_bivariate.png`	`outputs.figures.dir`	8	`vulnerability_maps`
Table 7	`table7_pca_features_components.csv`	`outputs.tables.dir`	7	`build_bvi`
Table 8	`table8_pcfa_eigenvalues.csv`	`outputs.tables.dir`	7	`build_bvi`
Table 9	`table9_pcfa_factor_loadings_rotated.csv`	`outputs.tables.dir`	7	`build_bvi`
Table 10	`table10_bvi_poorbrdgs_benchmarking.csv`	`outputs.tables.dir`	11	`benchmark_poor_bridges`

CSV columns

Each fig*.png has a sibling fig*.csv of the plotted values (not the full pipeline dumps). Values are rounded in tables 7–9 to two decimals.

Files	Columns
Figs 2, 3a–i, 6, 10a–d, 11	`County Name` / `County` (Fig. 2), `County_Code`, mapped value, `bin` (colour interval). Fig. 2 also has `over_water_count`.
Fig. 4	`indicator_row`, `indicator_col`, `correlation`, `p_value`, `significant_p05`
Fig. 5	`indicator_row`, `indicator_col`, `residual_correlation`
Fig. 7	County, three group contributions, `dominant_sub_indicator_group` (code), `dominant_group` (label)
Figs 8a / 9a	Per-county rank summary: `original_rank`, `min`, `p05`, `p25`, `median`, `p75`, `p95`, `max`, `mean`
Figs 8b / 9b	Per-county `total_variance`, `unexplained_variance`, plus one column per sampled parameter
Figs 8c / 9c	Per-county Sobol total-effect (`ST`) values by parameter
Figs 12–13	Long form: county, `social_value`, BVI or `bridge_value`, panel `pearson_r` / `p_value`
Fig. 14	Raw BVI and SVI, `final_scores_bin` / `SVI_bin`, `combined`, `BVI Category`, `Social Vulnerability` (High/Medium/Low)
Table 7	Indicator rows × unrotated `PC1`… columns
Table 8	Eigenvalues before extraction and after rotation (blank cells are unretained components)
Table 9	Rotated factor loadings, unity-sum squared loadings, `Communality`; footer rows `Explained variance` and `Explained/total`
Table 10	BVI rank vs poor-bridge replacement-cost rank, `Priority Shift` (Traditional − BVI), `BVI Category`, `Social Vulnerability`

Full Monte Carlo sample dumps (sensitivity_results_*_rankings.csv, sobol_indices_*_rankings_wide.csv) stay under intermediate.sensitivity.dir for --plot-only; they are not the figure CSVs.

External dependencies

PS density maps for California were generated with the PS prediction workflow in SafeStruct/GlobalRiskBridge (src/ps_predictions/predict_PS_dens.ipynb). That notebook currently supports regions within a single 1×1° tile. To recreate the statewide PS prediction raster, generate all tiles separately and merge them (for example in QGIS). The published data bundle includes the merged California PS-density GeoTIFF required as input to step 5 (merged_california.tif), so regenerating the PS density map is optional.

License

MIT

Name		Name	Last commit message	Last commit date
config		config
hpc		hpc
src/bridge_vulnerability		src/bridge_vulnerability
.gitignore		.gitignore
.pre-commit-config.yaml		.pre-commit-config.yaml
CHANGELOG.md		CHANGELOG.md
CITATION.cff		CITATION.cff
LICENSE		LICENSE
README.md		README.md
conda_run_full_pipeline.sh		conda_run_full_pipeline.sh
environment.yml		environment.yml
poetry.lock		poetry.lock
poetry_run_full_pipeline.sh		poetry_run_full_pipeline.sh
pyproject.toml		pyproject.toml

Navigation Menu

Repository navigation

About

Repository files navigation

California Bridge Vulnerability Index (BVI)

How to cite

Data availability

Installation

Configuration

Required inputs

External data (download or obtain)

Bundled intermediate (skip HPC steps 2–3b)

Displacement susceptibility (manual if not bundled)

Run the full pipeline (local)

Running the pipeline

NBI inventory

OSM bridge lines — optional (HPC only; can run in parallel with step 4)

Displacement rasters (can run in parallel with steps 2–3)

Displacement susceptibility raster (required for step 5)

Bridge-level raster extraction

NBI enrichment

Build BVI (PCA/PFA)

Result maps

Sensitivity analysis

Validation

Pipeline overview

Repository structure

Modules that are not run directly

Pipeline I/O reference

Paper figures

CSV columns

External dependencies

License

About

Releases

Packages

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

California Bridge Vulnerability Index (BVI)

How to cite

Data availability

Installation

Configuration

Required inputs

External data (download or obtain)

Bundled intermediate (skip HPC steps 2–3b)

Displacement susceptibility (manual if not bundled)

Run the full pipeline (local)

Running the pipeline

NBI inventory

OSM bridge lines — optional (HPC only; can run in parallel with step 4)

Displacement rasters (can run in parallel with steps 2–3)

Displacement susceptibility raster (required for step 5)

Bridge-level raster extraction

NBI enrichment

Build BVI (PCA/PFA)

Result maps

Sensitivity analysis

Validation

Pipeline overview

Repository structure

Modules that are not run directly

Pipeline I/O reference

Paper figures

CSV columns

External dependencies

License

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages