NIFC Wildfire Data: The Federal Database Behind a Century of US Fire Statistics

From its campus in Boise, Idaho, the National Interagency Fire Center coordinates the entire federal wildland firefighting apparatus and maintains the most comprehensive wildfire statistical record in the United States — annual fire counts and acreage back to 1926, individual fire records for 2.3 million fires since 1992, satellite-derived burn severity maps for every significant fire since 1984, and real-time active fire perimeters updated multiple times daily during the fire season. The data tells a story of structural change: acreage burned has roughly tripled from the 1980s to the 2010s–2020s, suppression costs have climbed into the billions annually, and the Wildland-Urban Interface now places 43 million American homes in direct contact with fire-prone landscape. Understanding the federal data infrastructure behind these numbers is prerequisite to any serious analysis of wildfire risk.

NIFC: coordination hub and statistical clearinghouse

The National Interagency Fire Center was established in 1965 as a shared operations base for coordinating federal wildland firefighting resources. Its Boise campus houses representatives from five principal federal land management and firefighting agencies: the USDA Forest Service (USFS), the Bureau of Land Management (BLM), the National Park Service (NPS), the Bureau of Indian Affairs (BIA), and the Fish and Wildlife Service (FWS). The National Weather Service provides on-site fire weather forecasting support. State forestry agencies participate through interagency agreements that give the national coordination system access to state resources and provide states with access to federal assets.

The operational core of NIFC is the National Interagency Coordination Center (NICC), which functions as the national dispatch hub when fires overwhelm local and regional resources. Requests for additional crews, aircraft, and equipment flow upward through eleven Geographic Area Coordination Centers (GACCs) covering the continental US, Alaska, and Hawaii, then to NICC when resources must cross regional boundaries. NIFC tracks the national resource pool — airtankers, helicopters, hotshot crews, type 2 crews, engines, dozers — and manages the repositioning of assets as fire weather and ignition risk shift across the country through a season.

NIFC also serves as the statistical clearinghouse for US wildfire data. Its annual publication, the National Report of Wildland Fires and Acres Burned, is the authoritative source for fire counts and acreage by year and by agency jurisdiction. The historical tables at nifc.gov/fire-information/statistics extend to 1926, making it the longest continuous federal record of fire occurrence in the United States. In 2023 — a relatively moderate fire year — NIFC reported approximately 56,580 fires burning roughly 2.7 million acres nationally. The 10-year historical average runs approximately 7 million acres per year.

The historical acreage trend: a century of data

The NIFC annual statistics database begins in 1926 with the earliest reasonably consistent records of federal fire reporting. The pre-1983 data comes with substantial caveats: reporting standards varied widely by agency and region, many remote fires went unreported or were estimated with large uncertainty margins, and the geographic scope of federal reporting expanded over time as land management infrastructure grew. The record from the early 1980s onward is considered largely comparable across years and forms the basis for modern trend analyses.

The trend in the modern era is unambiguous. The 1980s decade averaged roughly 2–3 million acres burned per year. The 1990s saw higher variability, with the 1996 season producing nearly 6.1 million acres. The 2000s marked a structural shift: 2000 (7.4M acres), 2005 (8.7M), and 2006 (9.9M) all exceeded anything seen since the era of reliable reporting. The 2010s produced three seasons exceeding 10 million acres: 2015 (10.1M), 2017 (10.0M), and 2020 (10.1M). The 2010s–2020s average runs approximately 7–8 million acres per year — two to three times the 1980s baseline.

The record years tell their own stories. The 2015 season burned 10.13 million acres, driven by an exceptional drought year in Washington State and across the Northern Rockies. The 2017 season (10.0M acres) was notable for the Thomas Fire in California (281,893 acres, the largest in California history at the time) and the catastrophic wine country fires of October 2017. The 2020 season (10.1M acres) saw California alone burn 4.2 million acres — more than any prior full California fire season — including the August Complex, the first individual fire in California history to exceed one million acres. Oregon and Colorado also recorded their worst fire years in 2020. The 2021 season produced 7.1 million acres nationally, headlined by the Dixie Fire (963,310 acres, the largest single non-complex fire in California history) and the Bootleg Fire in Oregon (413,748 acres).

The fire suppression paradox underlies much of this trend. Beginning with the 10 AM policy of the early 20th century — requiring all fires to be suppressed by 10 AM the following morning — and reinforced by the Smokey Bear campaign launched in 1944, federal fire policy pursued near-total suppression of wildland fire for decades. Suppression was effective enough at reducing annual burn area that fuel loads (dead wood, accumulated brush, and dense understory) accumulated across tens of millions of acres of western forests. Fire ecologists now understand that many western forest types are fire-adapted ecosystems maintained by periodic low-to-moderate severity fire. Excluding fire for 50–80 years produced unprecedented fuel accumulations that now drive larger, more intense, and harder-to-suppress fires when ignitions occur. The fires of the 2010s and 2020s are in part burning the backlog of fuels that 20th-century suppression policy created.

NIFC annual statistics database

NIFC publishes its historical fire statistics at nifc.gov/fire-information/statistics as annual Excel spreadsheets, PDF tables, and downloadable CSV files. The core data table covers fires and acres burned from 1926 to the current season, broken out by federal agency jurisdiction (USFS, BLM, NPS, BIA, FWS, BIA, other federal) and by state and other land jurisdiction (state, local, private). This jurisdictional breakdown is analytically important: federal land accounts for approximately 640 million acres (28% of US land area), concentrated in the western states. In a severe western fire year, federal land acreage dominates the national total.

California's share of national fire acreage has grown substantially. Historically, California represented roughly 5–10% of national acreage in most years. In the 2017–2021 period, California consistently contributed 15–40% of national totals. This reflects multiple converging factors: a multi-year drought that peaked in 2021–2022 at levels not seen in at least 1,200 years (per tree-ring reconstruction), increasing vapor pressure deficit, the invasion of annual grasses into intermontane valleys (particularly cheatgrass in the Great Basin, which creates continuous fine-fuel carpets that dry earlier and carry fire more readily than native perennial shrubs), and the continued expansion of residential development into fire-prone foothill and mountain landscapes.

Federal fire suppression expenditures tracked by NIFC document the financial consequence of the acreage trend. In typical years of the 1990s, total federal suppression costs ran in the range of $500 million to $1 billion. By the 2010s, $2–4 billion became common in active years. The USFS alone spent approximately $2.5 billion on suppression in 2017. The cumulative cost of federal wildland fire suppression across 2000–2023 exceeded $40 billion — costs borne by the USFS, Department of the Interior bureaus, and FEMA disaster declarations that fund state and local response.

The “fire borrowing” problem plagued USFS budgets for decades before a legislative fix in 2018. When suppression costs exceeded the Congressional appropriation for fire, the Forest Service was authorized to transfer funds from non-fire accounts — reforestation, trail maintenance, timber management, watershed restoration — to cover the shortfall. This mechanism cannibalized the forest health programs that reduce future fire risk in order to pay for fighting fires that inadequate forest health management had made more severe. The 2018 Consolidated Appropriations Act created a fire suppression cap mechanism that allows agencies to draw from an emergency reserve when actual suppression costs exceed 70% of the 10-year average appropriation, ending the practice of borrowing from non-fire accounts.

USFS Fire Occurrence Database

For individual fire records with geographic precision, the authoritative federal source is the US Forest Service Fire Occurrence Database (FOD), developed by Karen Short of the USFS Rocky Mountain Research Station and published through the USFS Research Data Archive (study RDS-2013-0009). The FOD aggregates wildfire occurrence records from federal, state, and local reporting systems for 1992 through the most recent update, and as of current releases contains approximately 2.3 million individual fire records. The database is available in SQLite format at the USFS Research Data Archive at fs.usda.gov/rds/archive.

The FOD schema is rich. Key fields include: FOD_ID and FPA_ID (unique identifiers); SOURCE_SYSTEM_TYPE and SOURCE_SYSTEM (which reporting system the record came from); NWCG_REPORTING_AGENCY and NWCG_REPORTING_UNIT_ID (the National Wildfire Coordinating Group agency and unit); FIRE_NAME; FIRE_YEAR; DISCOVERY_DATE and DISCOVERY_DOY (day of year of discovery); NWCG_CAUSE_CLASSIFICATION (Human, Lightning, or Unknown); NWCG_GENERAL_CAUSE and NWCG_SPECIFIC_CAUSE (debris burning, equipment, fireworks, lightning, and so on); CONT_DATE and CONT_DOY (containment date and day of year); FIRE_SIZE in acres; FIRE_SIZE_CLASS (A through G); LATITUDE and LONGITUDE of the fire discovery point; OWNER_DESCR (federal agency, state, private); STATE; COUNTY; and FIPS_CODE.

The size class system follows NWCG conventions. Class A covers fires under 0.25 acres; Class B, 0.26–9.9 acres; Class C, 10–99.9 acres; Class D, 100–299 acres; Class E, 300–999 acres; Class F, 1,000–4,999 acres; and Class G covers fires of 5,000 or more acres. Class G fires represent less than 2% of fire counts in most years but account for 95% or more of total acreage burned. The FOD also includes MTBS_ID and MTBS_FIRE_NAME fields linking FOD records to the MTBS burn severity database for fires that meet the MTBS mapping threshold.

Cause distribution in the FOD reveals geographic and jurisdictional patterns. Lightning is the dominant cause of Class G fires in the remote interior West — ignitions that can smolder for days before detection and where suppression response time is measured in hours to days rather than minutes. Human-caused fires dominate ignition counts in populated areas and tend to occur closer to communities, increasing wildland-urban interface damage probability. Debris burning, equipment use, and campfires each account for significant human-caused ignition shares in different regions. California's electrical infrastructure represents a distinct ignition pathway: PG&E's equipment ignited the Camp Fire in 2018 and multiple other large fires, leading to a $13.5 billion wildfire fund settlement and the utility's 2019 bankruptcy.

MTBS: Monitoring Trends in Burn Severity

Monitoring Trends in Burn Severity (MTBS) is a joint USGS–USFS program producing Landsat-derived burn severity maps for every fire exceeding 1,000 acres in the western United States (500 acres in the eastern US) from 1984 to the present. The program applies pre-fire and post-fire Landsat imagery to compute the differenced Normalized Burn Ratio (dNBR), a spectral index sensitive to char, fire-killed vegetation, and soil exposure. The dNBR is classified into five severity categories: unburned, low severity, moderate severity, high severity, and increased greenness (rare — vigorous post-fire vegetation flush). Fire perimeters are delivered as polygon shapefiles. The MTBS archive at mtbs.gov provides burn severity rasters and perimeter shapefiles for more than 25,000 fires across the four-decade record.

The high-severity fraction of individual fires carries significant ecological implications. High-severity burning kills the entire above-ground plant community and typically the seed bank in the upper soil layer. In fire-adapted conifer forests, post-fire regeneration depends on seed dispersal from surviving trees at the fire perimeter or from seed sources within unburned patches inside the fire perimeter. When high-severity patches exceed the seed dispersal distance of the dominant conifers — typically 50–200 meters for most western pines and firs — natural regeneration fails and the burned area may convert to shrubland or grassland rather than returning to forest. Research on MTBS data shows that the size and spatial continuity of high-severity patches have both increased with the warming and drying trend, creating a post-fire regeneration crisis in some areas of the Sierra Nevada, Cascades, and Northern Rockies.

Three MTBS analyses illustrate the pattern. The Thomas Fire of December 2017 in Ventura and Santa Barbara counties, California, burned 281,893 acres with approximately 25% classified as high severity — a significant proportion for a chaparral fire driven by extreme Santa Ana winds. The Camp Fire of November 2018 in Butte County — 153,336 acres, 85 deaths, and 18,804 structures destroyed, making it the deadliest and most destructive California wildfire in recorded history — showed an extreme high-severity core in the MTBS mapping, concentrated in the area that consumed the town of Paradise. The King Fire of 2014 in El Dorado County burned 97,717 acres with approximately 47% classified as high severity, one of the highest proportions recorded in the Sierra Nevada, indicating that much of the burned area was unlikely to regenerate to conifer forest within decades without active replanting.

ICS-209 incident status reports

For fires requiring extended suppression operations — typically fires exceeding 100 acres on federal land or any fire with significant threat to structures or public safety — incident commanders file daily Interagency Situation Reports in the ICS-209 format to NIFC. These reports provide the operational and financial accounting for each significant incident throughout its life cycle.

ICS-209 fields cover: incident name and incident number; reporting period date; jurisdiction agency; total acres; percent contained; resources currently assigned (overhead personnel, engines, water tenders, dozers, hand crews, helicopters); structures threatened and structures destroyed; civilian injuries and fatalities; firefighter injuries and fatalities; total suppression cost to date; current weather conditions including temperature, relative humidity, and wind speed and direction; fire behavior observations; and cause. NIFC hosts the ICS-209 archive through the FAMWEB (Fire and Aviation Management Web) system at famweb.nwcg.gov. The ICS-209-PLUS database, a structured version of the archive, is available for research use and links to FOD records through the ICS_209_PLUS_INCIDENT_JOIN_ID field in the USFS Fire Occurrence Database.

InciWeb (inciweb.wildfire.gov) is the public-facing incident information system that aggregates ICS-209 data and other agency communications into a public-accessible website for current and recently concluded fires. InciWeb provides fire perimeter maps, evacuation information, air quality alerts, and crew assignment information for significant fires. Historical InciWeb records provide a narrative archive of fire progression that complements the structured data in the ICS-209-PLUS database.

The Wildland-Urban Interface

The Wildland-Urban Interface (WUI) designates areas where structures and human infrastructure directly intermingle with or abut undeveloped wildland vegetation. The WUI is the geographic zone where wildfire transitions from a land management problem to a public safety and property loss disaster. Research by Radeloff et al. published in 2018 in the Proceedings of the National Academy of Sciences found that 43 million homes were located in the WUI as of 2010, making it the fastest-growing land-use type in the United States. WUI acreage increased 33% between 1990 and 2010, driven by suburban and exurban residential development patterns that placed new housing in foothill, mountain, and intermontane landscape settings with high fire frequency and ignition probability.

The states with the largest WUI areas — California, Montana, Colorado, Wyoming, and Texas — are also among the states with the highest fire activity. California presents the most acute convergence: the state has by far the largest WUI housing stock, the densest population in fire-prone foothill landscapes, and a fire environment that has grown dramatically more severe with climate change and fuel accumulation. The result is an escalating structure loss rate that was not anticipated in pre-2000 risk assessments.

The Camp Fire of November 2018 is the defining WUI disaster in US history. The fire ignited from PG&E electrical equipment near the community of Pulga and spread at exceptional speed under extreme Diablo wind conditions into Paradise, a city of approximately 27,000 residents in Butte County. The fire overran the community faster than evacuation could be completed. Eighty-five people died, most of them elderly residents unable to evacuate. The entire community of Paradise was destroyed, along with the communities of Concow and Magalia. The event exposed fundamental inadequacies in evacuation route planning, fire-resistant building code adoption, and utility infrastructure safety standards in high-risk WUI communities.

The Lahaina fire of August 2023 on Maui, Hawaii, demonstrated that catastrophic WUI fire is not confined to the western mountain states. Hurricane-force Kona winds from a distant storm — sustained at 60–80 mph with higher gusts — combined with exceptional drought and the presence of invasive African buffelgrass and Guinea grass in the fire pathway to produce a fire that destroyed more than 2,200 structures in Lahaina within hours and killed more than 100 people, making it the deadliest US wildfire in more than 100 years. The Lahaina event highlighted how invasive annual grasses can create continuous fine-fuel matrices in landscapes previously considered lower fire risk and how extreme wind events outside traditional fire weather parameters can compress the time available for evacuation to near zero.

Structure loss statistics reflect the WUI expansion trend. Federal estimates of average annual structure loss from wildfire ran approximately 3,000 structures per year through the 2000s. The 2010s and 2020s have seen routine years of 10,000 or more structures destroyed, with individual fires like the Camp Fire (18,804 structures), the 2017 Tubbs Fire (5,636 structures), and the 2020 Creek Fire (855 structures plus 234 outbuildings) driving single-event losses that previously occurred only in the most catastrophic years.

Active fire data and satellite detection

During the fire season, NIFC publishes daily active fire situation reports and maintains GIS data feeds for current large fire perimeters. The NIFC ArcGIS Feature Service at services3.arcgis.com/T4QMspbfLg3qTGWY/arcgis/rest/services/Active_Fires/FeatureServer provides GeoJSON of current active large fires, updated continuously as incident commanders submit perimeter mapping from aircraft, ground crews, and satellite sources. Fields include incident name, GIS-derived acreage, percent contained, point of origin state, fire cause, and modification timestamp. The endpoint accepts standard ArcGIS REST query parameters for filtering, spatial queries, and output format specification.

NASA's Fire Information for Resource Management System (FIRMS) provides satellite-detected active fire point data from the MODIS and VIIRS instruments, updated within hours of satellite overpass. FIRMS data is accessible at firms.modaps.eosdis.nasa.gov and includes latitude/longitude of fire detections, brightness temperature, fire radiative power, acquisition time, satellite, and confidence score. FIRMS' near-real-time coverage complements the perimeter-based NIFC data by detecting new ignitions and fire spread into areas not yet covered by ground-based perimeter mapping. The two data sources are used in combination by operational fire managers and researchers: FIRMS for initial detection and spread monitoring; NIFC perimeters for definitive containment boundaries and resource deployment planning.

LANDFIRE (landfirereview.org) provides the national geospatial reference data on vegetation cover, fuel models, and historical fire regime for the continental US, Alaska, and Hawaii. LANDFIRE 13 custom fuel model rasters and existing vegetation type data are the standard inputs to operational fire behavior modeling systems including FARSITE, FlamMap, and the Prometheus fire growth model. LANDFIRE is updated approximately every three to five years to account for post-fire vegetation change, timber harvest, and invasive species spread. Researchers combining LANDFIRE fuel data with MTBS historical fire patterns can identify landscapes where fuel accumulation and fuel type change create elevated risk of future high-severity fire.

The climate change signal in wildfire data

The scientific consensus linking climate change to increased US wildfire risk rests on multiple independent lines of evidence visible in the federal data record. Westerling et al. (2006, Science) showed that large wildfire frequency and duration increased suddenly and markedly in the mid-1980s across western US forests, correlated with earlier spring snowmelt and longer, hotter summers. Earlier snowmelt extends the period of low soil moisture and dry fine fuels into late spring, effectively lengthening the fire season. The shift was detectable against the NIFC historical record and against the fire occurrence and containment date data in the USFS FOD.

Williams et al. (2019, PNAS) demonstrated that increasing vapor pressure deficit (VPD) across the western United States is the strongest statistical predictor of year-to-year variation in burned area. VPD — the difference between the moisture-holding capacity of air at a given temperature and the actual moisture content — captures the atmospheric drying demand on vegetation. Because warmer air holds more moisture at saturation, temperature increases alone drive VPD upward even without any change in relative humidity. VPD increases since the 1970s explain approximately 55% of the increase in western US forest area burned in the NIFC record, independent of changes in land management. Climate models project that VPD will continue increasing across the West under all emissions scenarios, implying continued expansion of fire weather conditions.

Projected fire weather trends are stark. Multiple modeling studies project 3–4x increases in the number of extreme fire weather days by the end of the 21st century under high-emissions scenarios. Lightning-caused fire ignitions are expected to increase with warming, as convective storm intensity increases in a warmer atmosphere. The combination of more ignitions, longer fire seasons, higher VPD, and continued WUI expansion creates a compound risk trajectory that fire policy, land management, and building codes have not yet adjusted to at scale.

California presents the sharpest case study of the climate–fire nexus. The state's wildfire crisis reflects the convergence of: prolonged and intensifying drought driven by both reduced precipitation and higher evaporative demand; VPD increase from warming temperatures; invasive annual grasses (primarily cheatgrass and wild oat) replacing native perennial shrublands in intermontane valleys, creating more flammable continuous fuel carpets; continued residential development in high-risk landscapes; and electrical transmission infrastructure that creates ignition risk during extreme wind events. PG&E's transmission lines caused the Camp Fire and a significant fraction of the other largest California fires of the 2017–2021 period, ultimately resulting in a $13.5 billion wildfire fund settlement with fire victims and the utility's reorganization under Chapter 11 bankruptcy.

Python: NIFC historical trends and active fire data

The script below performs two tasks. The first section computes decade-by-decade average fires and acres burned from NIFC historical data (1983–2023), identifies the peak years by acreage, and calculates the pre-1990 versus post-2010 ratio that captures the scale of the long-term increase. The data is hardcoded from the published NIFC tables as a robust fallback, though the commented URL points to the remote CSV that NIFC publishes at their statistics page. The second section queries the NIFC ArcGIS GeoJSON endpoint for currently active large fires, parsing the feature collection to rank fires by GIS-derived acreage and display containment status. The ArcGIS endpoint returns an empty feature collection outside the active fire season, so the script handles that case gracefully.

import json
import urllib.request
import pandas as pd
import numpy as np

# =============================================================================
# NIFC Wildfire Data Analysis
# =============================================================================
#
# Part 1: NIFC Historical Annual Fire Statistics
#   Source: https://www.nifc.gov/fire-information/statistics
#   The canonical CSV is available at the NIFC statistics page.
#   We use inline fallback data drawn from published NIFC tables (1926-2023).
#
# Part 2: Active large fires from NIFC ArcGIS GeoJSON endpoint
#   Endpoint: https://services3.arcgis.com/T4QMspbfLg3qTGWY/arcgis/rest/
#             services/Active_Fires/FeatureServer/0/query
#   Returns GeoJSON of current large fires with acreage and containment.
# =============================================================================

# ---------------------------------------------------------------------------
# Part 1: Historical annual statistics (decade-by-decade analysis)
# ---------------------------------------------------------------------------

# Inline data derived from the NIFC published historical table.
# Fires = number of wildland fires; Acres = total acres burned.
# Pre-1983 data has significant reporting variability; modern era is 1983+.
ANNUAL_DATA = {
    1983: (18229, 1323666),   1984: (20493, 1148409),
    1985: (82591, 2447296),   1986: (140226, 3928252),
    1987: (72750, 2719162),   1988: (122738, 4621621),
    1989: (85438, 1873199),   1990: (58908, 2237860),
    1991: (79107, 4521000),   1992: (87394, 2069929),
    1993: (58810, 1797574),   1994: (79107, 4073579),
    1995: (82234, 1840546),   1996: (96363, 6065998),
    1997: (66196, 2856959),   1998: (81043, 1329704),
    1999: (92702, 5626093),   2000: (92250, 7393493),
    2001: (84076, 3570911),   2002: (73457, 7184712),
    2003: (63629, 3960842),   2004: (77534, 8097880),
    2005: (66753, 8689389),   2006: (96385, 9873745),
    2007: (85822, 9328045),   2008: (78979, 5292468),
    2009: (78792, 5921786),   2010: (71971, 3422724),
    2011: (74126, 8711367),   2012: (67774, 9326238),
    2013: (47579, 4319546),   2014: (63312, 3595613),
    2015: (68151, 10125149),  2016: (67743, 5509995),
    2017: (71499, 10026086),  2018: (58083, 8767492),
    2019: (50477, 4664364),   2020: (58950, 10338012),
    2021: (58985, 7125643),   2022: (68988, 7577183),
    2023: (56580, 2693910),
}

rows = [
    {"Year": yr, "Fires": fires, "Acres": acres}
    for yr, (fires, acres) in ANNUAL_DATA.items()
]
df = pd.DataFrame(rows).sort_values("Year").reset_index(drop=True)

print("=== NIFC Historical Wildland Fire Statistics (1983-2023) ===")
print(f"Total years:          {len(df)}")
print(f"Mean annual fires:    {df['Fires'].mean():,.0f}")
print(f"Mean annual acres:    {df['Acres'].mean():,.0f}")
print(f"Peak year (acres):    {df.loc[df['Acres'].idxmax(), 'Year']}  "
      f"({df['Acres'].max():,.0f} acres)")
print(f"Peak year (fires):    {df.loc[df['Fires'].idxmax(), 'Year']}  "
      f"({df['Fires'].max():,.0f} fires)\n")

# Decade-by-decade averages
df["Decade"] = (df["Year"] // 10) * 10
decade_summary = (
    df.groupby("Decade")
    .agg(
        Years=("Year", "count"),
        AvgFires=("Fires", "mean"),
        AvgAcres=("Acres", "mean"),
        MaxAcres=("Acres", "max"),
    )
    .reset_index()
)

print("=== Decade-by-Decade Averages ===")
print(f"{'Decade':<8} {'Avg Fires':>10} {'Avg Acres':>12} {'Max Acres':>12}")
print("-" * 46)
for _, row in decade_summary.iterrows():
    print(
        f"{int(row['Decade'])}s   "
        f"{row['AvgFires']:>10,.0f} "
        f"{row['AvgAcres']:>12,.0f} "
        f"{row['MaxAcres']:>12,.0f}"
    )

# Pre-1990 vs post-2010 comparison
pre_1990 = df[df["Year"] < 1990]["Acres"].mean()
post_2010 = df[df["Year"] >= 2010]["Acres"].mean()
print(f"\nPre-1990 mean acres:  {pre_1990:,.0f}")
print(f"Post-2010 mean acres: {post_2010:,.0f}")
print(f"Ratio (post/pre):     {post_2010/pre_1990:.1f}x increase\n")

# Record years ranked by acreage
print("=== Top 10 Years by Acres Burned ===")
top10 = df.nlargest(10, "Acres")[["Year", "Fires", "Acres"]]
for _, row in top10.iterrows():
    print(f"  {int(row['Year'])}: {row['Acres']:>12,.0f} acres  ({row['Fires']:,.0f} fires)")

# ---------------------------------------------------------------------------
# Part 2: Active large fires from NIFC ArcGIS GeoJSON endpoint
# ---------------------------------------------------------------------------

ACTIVE_FIRES_URL = (
    "https://services3.arcgis.com/T4QMspbfLg3qTGWY/arcgis/rest/services"
    "/Active_Fires/FeatureServer/0/query"
    "?where=1%3D1"
    "&outFields=IncidentName,GISAcres,PercentContained,POOState,"
    "ModifiedOnDateTime,FireCause"
    "&outSR=4326"
    "&f=geojson"
)

print("\n=== Active Large Fires (NIFC ArcGIS Feed) ===")
try:
    req = urllib.request.Request(
        ACTIVE_FIRES_URL,
        headers={"User-Agent": "Mozilla/5.0 (research/analytics)"},
    )
    with urllib.request.urlopen(req, timeout=20) as resp:
        geojson = json.loads(resp.read().decode("utf-8"))

    features = geojson.get("features", [])
    if not features:
        print("No active large fires at time of query (low-activity period).")
    else:
        fires_list = []
        for feat in features:
            props = feat.get("properties", {})
            fires_list.append({
                "Name": props.get("IncidentName", "Unknown"),
                "State": props.get("POOState", ""),
                "Acres": props.get("GISAcres") or 0,
                "Pct_Contained": props.get("PercentContained") or 0,
                "Cause": props.get("FireCause", ""),
            })
        active = pd.DataFrame(fires_list).sort_values("Acres", ascending=False)
        print(f"Active large fires: {len(active)}")
        print(f"Total active acreage: {active['Acres'].sum():,.0f}\n")
        print(f"{'Name':<30} {'State':<6} {'Acres':>10} {'%Cont':>7} {'Cause'}")
        print("-" * 72)
        for _, row in active.head(15).iterrows():
            print(
                f"{str(row['Name']):<30} "
                f"{str(row['State']):<6} "
                f"{row['Acres']:>10,.0f} "
                f"{row['Pct_Contained']:>6.0f}% "
                f"{row['Cause']}"
            )
except Exception as exc:
    print(f"Active fire feed unavailable: {exc}")
    print("(Outside active fire season, or network issue.)")
    print("Endpoint for manual access:")
    print("  https://services3.arcgis.com/T4QMspbfLg3qTGWY/arcgis/rest/"
          "services/Active_Fires/FeatureServer")

The decade summary output from this script makes the long-term trend immediately legible: 1980s averages in the 2–3 million acre range; 2010s averages exceeding 6–7 million acres; individual years in 2015, 2017, 2020 crossing 10 million acres. The pre-1990 versus post-2010 ratio typically computes at 3x or higher. The ArcGIS endpoint, during an active fire season, will return dozens to hundreds of large fire records with containment percentages and state attribution that allow quick identification of which fires are driving current national acreage.

Data access summary

The primary data sources and access points in the NIFC ecosystem are:

Limitations and analytical considerations

The NIFC historical record before 1983 is best treated as indicative rather than strictly comparable to modern figures. Reporting standards, detection technology, and the geographic scope of federal reporting have all changed across the century-long record. Analyses drawing trend conclusions from the full 1926–present series must account for these structural breaks.

The FOD's detection probability and reporting completeness are not uniform. Small fires on non-federal land, fires in remote areas detected only by lightning detection systems, and fires in states with less-developed reporting infrastructure are underrepresented relative to actual occurrence. Cause code assignments are based on post-fire investigation, which is often incomplete for large fires in remote areas. Arson rates are likely underestimated, as formal arson coding requires an investigation finding rather than suspicion.

MTBS burn severity classifications are sensitive to the timing of post-fire Landsat imagery. Early post-fire images capture thermal damage before recovery; later images in the growing season incorporate initial vegetation flush that may reclassify some high-severity patches downward. MTBS standardizes imagery timing to the extent possible, but users comparing burn severity across fires and years should be aware that some variation reflects image timing differences rather than actual fire effects differences.

The NIFC GIS active fire perimeters reflect the most recent mapping available from incident management teams and are authoritative for operational purposes but carry mapping uncertainty, particularly on fire flanks and in areas of complex fire behavior. GIS acreage from the active fire endpoint may differ from official NIFC reported acreage, which is released through the daily NICC Situation Report and reflects the incident management team's certified figure rather than the automated GIS calculation.