Product Descriptions
Operational Fire Danger Forecast products can be accessed through searching the interactive viewer; downloading bulk data; or web data services for GIS platforms. Descriptions of the Fire Danger Forecast products are provided below and relevant publications about this project can be found on the Publications page.
NDVI and Relative Greenness
A key component of the Fire Danger Forecast product involves the use of remote sensing to determine vegetation conditions across the country. The Normalized Differenced Vegetation Index (NDVI) is the foundational tool used to estimate the amount and condition of vegetation on the ground. It is calculated using the red and near-infrared (NIR) spectral bands as follows:
\(NDVI = \frac{NIR - Red}{NIR + Red}\)
NDVI ranges from -1.0 to 1.0, with features such as barren, sand, and snow resulting in low values (0.1 or less) and surfaces with sporadic vegetation such as grasslands and senescing vegetation registering moderate NDVI values (0.2 – 0.5). High and dense vegetation coverage on a surface gives higher NDVI values (0.6 – 0.9).
NDVI is available from many platforms, however the Fire Danger Forecast products utilize the eVIIRS system produced by the US Geological Survey. This system produces a 1-kilometer resolution, 7‑day composite of NDVI, using the Visible Infrared Imaging Radiometer Suite (VIIRS) aboard Suomi NPP. NDVI is rescaled by 10000 in eVIIRS, with the composite outputs ranging from -1999 to 10000. This composite is used to calculate Relative Greenness (RG).
RG indicates how green—or how high the NDVI value— each pixel currently is with respect to the historical maximum and minimum NDVI at that location. Specifically, RG is derived using the following equation:
\(RG = \frac{NDVI_{Observed} - NDVI_{mMinimum}}{NDVI_{mMaximum} - NDVI_{mMinimum}} * 100\)
RG values are scaled from 0 to 100%. Higher values of RG indicate more healthy vegetation, and thus a higher proportion of live fuels that are unavailable to burn. Conversely, lower RG values indicate a drought-stressed or dormant vegetation, with a higher proportion of dead fuels that are available to burn. This method is applied to the Wildland Fire Potential Index calculation (see below).
The Wildland Fire Potential Index
The Wildland Fire Potential Index (WFPI) is a numerical rating of fuel availability and ignitability, based on an assessment of the proportion of dead fuel loading and its dryness. It can be used to indicate the “combustibility” of the landscape, with increasing values indicating increasing potential for large fires, defined as fires that burn more than 500 acres.
WFPI has the following assumptions:
- Live fuels are considered unavailable to burn.
- Dead fuels’ availability to burn is primarily a function of 10-hour timelag fuel moisture content.
- Proportions of Live to Dead fuels can be estimated using RG, a function of satellite based NDVI.
- Fire is assumed not to spread in temperatures at or below freezing (32°F, 0°C) and daily rainfall amounts over 0.5 inches (13 mm)
- Wind linearly increases fire potential, doubling every 35 knots at flame height
Input data used for the WFPI system include gridded weather data from the National Digital Forecast Database, 10-hour timelag fuel moisture content from the Wildland Fire Assessment System, and NDVI composite data using the eVIIRS system from USGS. Fuel parameters (extinction moisture, live ratio) are informed by a national, 1-kilometer map of NFDRS 1978 fuel models, which can also be found on WFAS.
The WFPI model is primarily driven by two factors, a dryness factor and a deadness factor, each ranging from 0 to 100. When the vegetation at each pixel is more dry and more dead, the fire potential is higher. Rainfall and low temperatures reduce fire potential, and this is done by reducing the dryness factor in the model. To account for wind, fire potential is then increased such that it doubles every 35 knots of windspeed, measured at flame height:
\(WFPI = Deadness * Dryness * (1 + \frac{Windspeed_{Knots}}{35.0})\)
\(Deadness = (1 - RG * LiveRatio_{Max}) * 100\)
\(Dryness = (1 - \frac{Fuel Moisture}{Extinction Moisture}) * 100\)
where LiveRatioMax is a scaled value indicating the maximum recorded greenness of a pixel based on historical NDVI data, and Fuel Moisture is the 10-hour timelag moisture corrected for precipitation and temperature effects.
High WFPI indicates weather and vegetation conditions are more supportive of large fire activity. WFPI typically ranges from 0 to 150 but can be higher in extreme conditions. WFPI and associated products are calculated daily for the next seven forecast days and these data are freely available to view and download via this website.
WFPI-Based Large Fire Potential and WFPI-Based Fire Spread Potential Products
Two additional products are created that estimate the likelihood of large fire activity, given the fire history and WFPI at a given location and time. This is based on WFPI verification using a national fire occurrence dataset. Forecast skill was determined by grouping WFPI values into one of 12 classes of WFPI (0–10, 10–20, …, 110–120, 120+), estimating the observed proportion of fire occurrence and spread events in each class, and comparing these to the Observed Fire Spread Proportion (OFSP) and Observed Large Fire Proportion (OLFP). OFSP is defined as the ratio of large fires observed across CONUS to all fires greater than one acre, within the given day and WFPI value. OLFP is defined as the ratio of the number of observed large fires to all pixels across CONUS. The results are summarized below in Table 1.
This relationship forms the foundation for the WFPI-based Fire Spread Potential (WFSP) and WFPI-based Large Fire Potential (WLFP) products, using a non-linear regression. WLFP is the probability of the occurrence of a large fire at a pixel, which is displayed as number of fire occurrences per million pixels. WFSP is the probability of the occurrence of a large fire, given the existence of a 1-acre fire at that pixel. WFSP is displayed as the number of existing fires that become large per hundred pixels.
These probabilistic products are designed to be used in conjunction with WFPI to provide a context from which to assess a location’s fire danger. In other words, WLFP measures how likely a large fire is to occur and WFSP measures how likely an existing fire is to grow; both based on the location’s history and the WFPI rating. This is most useful when assessing larger areas to strategize areas that are most prone to large fire activity. WLFP forecasts are used to estimate area forecasts by Geographic Area Coordination Center to provide an estimate of the number of fires each region should expect in the short term with respect to the 0–25th , 25th–75th, and 75th–95th percentiles of historical fire activity throughout the year.
Table 1: Proportion of large fires per pixel (km2) and the proportion of ignitions that spread to become a large fire for the period between 2001–2015. These values may be used as estimates of the expected risk metrics, LFP and FSP, at each WFPI level.
|
WFPI
|
OLFP (Per Million Pixels) | OFSP (Per Hundred Pixels) |
|---|---|---|
| 0-10 | 0.04 | 1.24 |
| 10-20 | 0.08 | 1.59 |
| 20-30 | 0.20 | 1.48 |
| 30-40 | 0.28 | 1.74 |
| 40-50 | 0.47 | 2.12 |
| 50-60 | 0.61 | 2.48 |
| 60-70 | 0.82 | 2.50 |
| 70-80 | 1.12 | 3.22 |
| 80-90 | 1.57 | 4.95 |
| 90-100 | 2.12 | 8.32 |
| 100-120 | 2.88 | 14.47 |
| 120-147 | 8.26 | 32.70 |
