Wildfire Size Disparity on Tribal Lands in California and Associated PM2.5 Burden, 2000–2018

Version 7.0 — June 2026 Integrates boundary stability tests (Section 16d) into the canonical deduplicated pipeline (FIRE_v7.ipynb). All boundary stability outputs now generated from the correct deduplicated dataset (111 tribal fires, 6.4× disparity). Supersedes v6, which contained boundary stability outputs generated from the pre-deduplication dataset and should be disregarded. Study Overview This dataset supports a quantitative investigation of wildfire size disparity between federally recognized tribal lands and non-tribal lands in California from 2000 to 2018, and the associated PM2.5 air quality burden borne by tribal communities. Using spatial analysis, panel econometrics, and four convergent causal identification strategies, the study documents that fires on tribal trust land are on average 6.4 times larger than fires on non-tribal land — a disparity that persists at 3.1× after excluding boundary-straddling fires and at 4.8× after excluding the catastrophic 2003 and 2007 fire years. All specifications are statistically significant (Mann-Whitney p < 0.0001). Tribal fire years are associated with a statistically significant increase in county mean PM2.5 of 0.845 μg/m³ (two-way fixed effects OLS, county + year FE, clustered SE, p = 0.0087). The Bureau of Indian Affairs — the only federal agency with statutory trust responsibility toward tribal nations — managed 7.0% of fires on tribal trust land during the study period. CAL FIRE and state agencies combined managed 54.6%, operating without trust obligations or mandates to integrate Traditional Ecological Knowledge. Data Contents This archive contains all outputs necessary to replicate the analysis. The complete replication pipeline (FIRE_v7.ipynb) reproduces all results in a single execution and downloads source data directly from public APIs. Files included: FIRE_v7.ipynb — Full integrated analysis pipeline (Sections 1–22 including Section 16d) ca_fires_tribal_classified.csv — 4,496 California fire perimeters with tribal land classification (111 tribal after deduplication) tribal_fire_pm25_panel.csv — 816 county-year observations (PM2.5 + fire variables) tribal_nation_fire_burden.csv — Fire burden summary by tribal nation (top 20) event_study_coefficients.csv — Event study coefficients (t = −5 to +5) with 95% CI regression_results.txt — Full two-way FE OLS output including all fixed effects boundary_stability_results.txt — Section 16d boundary stability test results (NEW v7) boundary_stability_table.csv — Covariate balance results at tribal boundary (NEW v7) placebo_rd_distribution.csv — 500 permutation placebo RD estimates (NEW v7) figure1_tribal_fire_analysis.png — Mean fire size by year; size distribution; top nations figure2_cedar_fire_aqi.png — San Diego County AQI 2000–2018; Cedar Fire 2003 figure3_regression_discontinuity.png — RD scatter plots at 10, 25, 50, 100 km bandwidths figure4_psm_overlap.png — Propensity score overlap before and after matching figure5_psm_differences.png — PSM matched pair differences distribution figure6_event_study.png — Event study coefficients with pre-trend test figure7_causal_identification.png — Convergent causal identification evidence figure_s1_boundary_stability.png — Placebo RD distribution and covariate balance (NEW v7) wildfire_replication_v7.zip — Full replication bundle README.md — Full documentation including revision notes Methods Summary Fire perimeter data were obtained from the NIFC Historical GeoMAC Perimeters archive (2000–2018) via ArcGIS REST API. Tribal boundaries are from the Census TIGER/Line AIANNH 2025 vintage (185 California areas). PM2.5 concentrations are from EPA AQS daily FRM/FEM monitoring data (parameter 88101; 580,066 California daily observations). All spatial operations use the California Albers Equal Area projection (EPSG:3310) for area and centroid calculations. Spatial join deduplication removed 12 NIFC duplicate polygon records and resolved 8 boundary-straddling fires to their largest-intersection tribal match, yielding 111 deduplicated tribal fires from 143 raw matches. Causal identification employs four strategies: (1) regression discontinuity at tribal land boundaries (four bandwidth specifications, 10–100 km); (2) propensity score matching with caliper 0.05 (93 matched fire pairs); (3) event study with county and year fixed effects and pre-trend joint F-test (F = 1.720, p = 0.206 — no pre-trend); (4) two-way fixed effects DiD (county + year FE, clustered SE). Boundary Stability Tests (Section 16d — NEW in v7) Three tests establish that the regression discontinuity design satisfies the boundary stability assumption: Test 1 — Boundary vintage check: All 185 California tribal boundaries carry Census TIGER MTFCC codes G2101 (American Indian Reservation), G2102 (Off-Reservation Trust Land), or G2160 — federally recognized designations established by Executive Order and treaty, primarily predating the 1934 Indian Reorganization Act. No California reservation boundaries in the study sample were administratively adjusted during the 2000–2018 study period. Boundary endogeneity is ruled out by legal structure. Test 2 — Placebo boundary RD (label permutation): Treatment labels were randomly permuted 500 times within the 25 km bandwidth window (n = 2,085 fires). The null distribution has a mean effect of +13.4% and standard deviation of 58.4%. The real boundary effect of +283% exceeds 498 of 500 permuted estimates (permutation p = 0.004), confirming the discontinuity is a statistically extreme outlier inconsistent with a geographic or statistical artifact. Test 3 — Covariate balance at boundary: Fire centroid latitude and longitude exhibit discontinuities at the tribal boundary, reflecting the geographic concentration of California reservations in distinct ecological zones rather than boundary endogeneity. Propensity score matching (Section 17) directly rebuts geographic confounding: standardized mean differences on latitude, longitude, and distance-to-boundary fall below 0.05 after caliper matching (93 matched pairs). Key Results Fire Size Disparity Specification Tribal mean (acres) Non-tribal mean (acres) Ratio p-value Full dataset (deduplicated) 17,265 2,706 6.4× < 0.0001 Excl. 2003 14,845 2,782 5.3× < 0.0001 Excl. 2003 + 2007 13,545 2,808 4.8× < 0.0001 Excl. boundary-straddlers 8,253 2,706 3.1× < 0.0001 PM2.5 Health Burden Estimator Coefficient p-value Two-way FE OLS (primary) 0.845 μg/m³ 0.0087 Event study post-period (F-test) F = 3.350 0.0320 Cedar Fire 2003 San Diego max AQI 328 (Hazardous) +124% spike Boundary Stability Test Result Vintage check (TIGER MTFCC) 185/185 federally recognized (G2101/G2102/G2160) Placebo RD permutation p 0.004 (2/500 placebos exceed real estimate) Covariate balance Lat/lon discontinuity explained by geographic clustering; PSM SMD < 0.05 Managing Agency Distribution (Tribal Trust Lands) Agency Fires (N) Share CAL FIRE (CDF) 39 27.3% State Agency 39 27.3% U.S. Forest Service 33 23.1% Bureau of Indian Affairs ✦ 10 7.0% Bureau of Land Management 6 4.2% Other / Unknown 16 11.2% ✦ Only agency with statutory trust responsibility toward tribal nations. Revision History v7.0 (June 2026): Integrates boundary stability tests (Section 16d) into canonical deduplicated pipeline. Three tests conducted: TIGER vintage check (185/185 G2101/G2102/G2160), label permutation placebo RD (permutation p = 0.004), and covariate balance at boundary. All tests run on correct 111-fire deduplicated dataset. Supersedes v6 boundary stability outputs. v6.0 (June 2026): Boundary stability outputs generated from pre-deduplication dataset — superseded by v7. v5.0 (June 2026): Zenodo archive with complete replication pipeline. v2.0 (June 2026): Critical methodological corrections — spatial join deduplication (143 → 111 tribal fires), two-way fixed effects added to Section 15, 2SLS removed (weak instruments), boundary-straddler sensitivity documented. v1.0 (June 2026): Initial deposit. Computational Environment Python 3.12.13 · geopandas 1.1.3 · pandas 2.2.2 · statsmodels 0.14.6 · scipy 1.16.3 · scikit-learn 1.6.1 · netCDF4 1.7.4 · Executed on Google Colab CPU runtime · Approximate runtime: 25–35 minutes License Creative Commons Attribution 4.0 International (CC BY 4.0). Users are free to share and adapt the material for any purpose provided appropriate credit is given to the authors.