GIS data

Rhode Island Geographic Information System's (RIGIS) 1995 statewide land use coverage and the United States Geological Survey (USGS) 1997 Digital Elevation Model (DEM) will be used for reference data and as for input into the FLAMMAP fire simulation model. The land use coverage is in vector format, in NAD 83, State-Plane coordinate system and were created at a 1:24,000 scale by aerial photo interpretation.

The interpretation of the land use data was based on the USGS classification system (Anderson, et. al, 1976).In addition, RIGIS Land use data set provides basic forest information, specifically, the breakdown of forest into deciduous, coniferous and mixed forest types. The forest type breakdown will serve as a benchmark mask for the fuels analysis, since it can be assumed that no new forests have grown in the state since the land use coverage was produced.

Satellite data

One satellite image was purchased from the United States Geological Survey (USGS) Earth Resource Observation System (EROS) for this study: Landsat Enhanced Thematic Mapper (ETM+) digital data, Path 12/Row 31, with 30 by 30 meter spatial resolution, was acquired April 10, 2002. The subset area of 3144 km 2 (Rhode Island) was extracted from the image data set using GIS data from Rhode Island Geographic Information System (RIGIS) and reprojected to NAD 83, State-Plane coordinate system.

1:5000 Digital Orthophotographs

1:5000 Digital Orthophotographs can be used to assess the accuracy of the fuel-type classification products. The images, derived from 1997 black and white aerial photographs, are available though RIGIS and are in NAD83, State Plane coordinate system with a spatial resolution of .18 square meters.

Orthophoto depicting pre-burn location in grass fuel type

Field Observations and Fuel Data

Collection of field data is the most important task in this research effort. 3024 field locations have been GPS surveyed, digitally photographed and assigned values for fuel loading, according to NFFL fire models. Most remotely sensed products, such as aerial photographs and satellite images are suitable only for very coarse level extraction of fuel types because the surface fuels, which govern the spread of most timber stand fires, are often obscured from view by the forest canopy (Elvidge 1988, Lachowski et al, 1995). However, knowledge of the forest canopy can provide some insight into surface fuels, especially when extensive ground truthing is conducted and used in concert with existing GIS layers. Also, the most important FARSITE layer, the fire behavior fuel model (FBFM) layer, is based more on expected fire behavior for a stand than characterization of fuel loadings (Anderson 1982, Bradshaw 1978, Burgan 1988). Therefore, in-the-field collaboration with experienced RIDEM foresters is necessary to assist in assigning values to the FBFM layer. Finally, the characterization of FBFM is very difficult to extract from remotely sensed imagery alone. This is due to the fact that wildland fire propagates mainly through fine and dead fuel classes (grasses, pine needles and fine woody materials less than 1" in diameter (Fahnestock, 1970) and fuel loadings for these fuel classes are intrinsically difficult to extract from remotely sensed imagery in timber stands, since dead fuels do not have spectral signatures that can be correlated with tree species or forest community types (Jensen 1986, Jensen and Cowen 1999 ). Given the limitations of remote sensing data to accurately reflect fuel loading, field reference data is used as the primary method to assign fuel classes and assess the accuracy of fuel-type classification products. Landsat remote sensing data is used as a link to impute values for un-sampled locations in the study area. Reference data was collected during the summer and fall of 2002 and 2003 at locations throughout the state. A Garmin etrex Global Positioning System (GPS) unit was used to collect reference points while a Canon Field Imaging System (FIS) was used to capture photographs of various land-cover types at the ground-truthing sites. These data form the core of a virtual field reference database (VFRDB), which allows on-the-ground field observations to be rapidly associated with remote sensing data. Additionally, three RIDEM forestry personnel have been trained in GPS/FIS use aided in the collection of VFRDB data.

Historical Data

When compiling a fire hazard survey and resultant map, historical fire data are normally referenced to determine areas that have been prone to fires in the past, and areas that have not burned in several years. In Rhode Island, this data is mostly unavailable. Through interviews with several RIDEM forest rangers, information regarding burn history throughout the state was compiled. It is known that a series of large 500-acre fires have burned in the southern portion of the state after the 1938 hurricane, but little is known of the exact coordinates or nature of the fire. Currently, RIDEM forest rangers attempt to map the location and extent of fires throughout the state. This program was unofficially implemented in the late 1990's and continues intermittently to date. Since the wildland fires that have been mapped are small in quantity and spatial extent, they will not be of significant use in correlating fire potential with remotely sensed data.