abigailhaddad/federal-public-lands-spending

Federal Public Lands Spending

Contract and grant transaction data from the USAspending Award Data Archive for federal agencies that manage public lands.

Interactive demo:

Pipeline code: github.com/abigailhaddad/federal-public-lands-contracting

Agencies covered

Department of the Interior (014): Bureau of Land Management, Bureau of Reclamation, Bureau of Safety and Environmental Enforcement, U.S. Geological Survey, National Park Service, Office of Surface Mining, U.S. Fish and Wildlife Service, Office of the Inspector General

Department of Agriculture (012): Forest Service

Dataset structure

Transaction files

4 files per fiscal year, split by department and award type:

File pattern	Description
transactions/doi_contracts_fyYYYY.parquet	Interior Department contract actions
transactions/doi_grants_fyYYYY.parquet	Interior Department grant awards
transactions/usda_contracts_fyYYYY.parquet	USDA Forest Service contract actions
transactions/usda_grants_fyYYYY.parquet	USDA Forest Service grant awards

Columns in each file:

• federal_action_obligation — dollar amount of the transaction
• fiscal_year — federal fiscal year (October–September)
• funding_sub_agency_name — which agency funded it
• recipient_parent_name / recipient_name — who received the money
• product_or_service_code_description — what was purchased (contracts)
• cfda_title — assistance program name (grants)
• county_fips — recipient county FIPS code (5-digit, zero-padded)
• recipient_state_code — recipient state

Aggregate files

Pre-computed annual obligation totals:

• aggregates/annual_by_psc.parquet — by product/service code (contracts)
• aggregates/annual_by_cfda.parquet — by assistance listing (grants)
• aggregates/annual_by_recipient.parquet — by recipient organization
• aggregates/annual_by_county.parquet — by county FIPS code

Lookup files

Historical statistics for t-test analysis:

• lookups/historical_lookup.parquet — wide format: one row per group, columns for each FY, plus hist_mean, hist_sd, hist_se, n_years
• lookups/historical_annual_detail.parquet — same data in long format

How to use

With DuckDB (no download needed)

import duckdb

HF = https://huggingface.co/datasets/abigailhaddad/federal-public-lands-spending/resolve/main

# Top 10 recipients in FY2023
duckdb.sql(f"""
    SELECT recipient_parent_name, SUM(federal_action_obligation) AS total
    FROM read_parquet([
        '{HF}/transactions/doi_contracts_fy2023.parquet',
        '{HF}/transactions/usda_contracts_fy2023.parquet'
    ])
    GROUP BY recipient_parent_name
    ORDER BY total DESC
    LIMIT 10
""").show()

With pandas

from huggingface_hub import hf_hub_download
import pandas as pd

path = hf_hub_download(
    repo_id=abigailhaddad/federal-public-lands-spending,
    filename=transactions/doi_contracts_fy2023.parquet,
    repo_type=dataset,
)
df = pd.read_parquet(path)

With R

library(arrow)
url <- https://huggingface.co/datasets/abigailhaddad/federal-public-lands-spending/resolve/main/transactions/doi_contracts_fy2023.parquet
download.file(url, doi_contracts_fy2023.parquet, mode = wb)
df <- read_parquet(doi_contracts_fy2023.parquet)

Update schedule

A GitHub Action runs daily and checks if USAspending has published a new monthly snapshot of the Award Data Archive. When a new snapshot is detected, it re-downloads all fiscal years one at a time and uploads any changes. On days when the Archive hasn't changed, nothing happens.

The Archive is typically updated once per month by USAspending.

Data source

All data comes from the USAspending Award Data Archive — the same source used in the original manual R-based analysis. Files are downloaded from https://files.usaspending.gov/award_data_archive/, filtered to the public-land agencies listed above, deduplicated, and converted to parquet.