This file was created by scanning the printed publication. Errors identified by the software have been corrected; however, some errors may remain. FIREMAP: Simulation of Fire Behavior - A GIS Supported System 1 MariQ J. Vasconcelos, D. Phillip Guertin, Malcolm J. Zwolinski2 Forest management and planning is a multiobjective task. Fire will impact the potential resource outputs and is one of the many factors that should be addressed, depending on context and goals. From a planning viewpoint, we must be able to predict the consequences of site-specific management actions on the fire characteristics (areas burned and intensity) for the area of concern. From the literature, fire behavior prediction and modeling are limited to homogeneous cover types (Rothermel1972,1983), where the other driving variables (weather and topography) are assumed to be uniform over the same area. These models are, therefore, unsuitable for dealing with the different spatial combinations of vegetation which a management alternative might consider and cannot be readily applied to patchy enviroments. The FIREMAP system attempts to address the problem of spatial/ temporal variability of the driving variables which has not yet been adequatelly considered in fire modeling. The FIREMAP System The objective of FIREMAP is to simulate the consequences of hypothesized changes in vegetation 'Poster paper presented at the conference. Effects of Rre in Management of Southwestern Natural Resources (Tucson. AZ. November 14-17. 1988). 'Division of Forest-Watershed Resources. School of Renewable Natural Resources. University of Arizona. composition and density on the fire characteristics (area burned and fire intensity) in well known ecosystems. This system estimates fire characteristics taking spatial and temporal variability into account and simulating the spread of fire in discrete time steps. FIREMAP is mainly applicable as a prescriptive tool but it can also be used for predicting fire behavior in "on site" situations when time effects have to be analyzed. Mapped outputs can also provide a basis for better communication. Model Development Tools The development of thi~ simulation tool consists of integrating a fire behavior modeling system- DIRECT, a module from the BEHAVE system (Andrews 1986), with a Geographic Information System-MAP (Tomlin 1985). DIRECT uses Rothermel's fire spread model (1972) to predict fire characteristics for a given continuous and relatively homogeneous area. Andrews (1980) reports favorable statistical comparisons between model predicted and observed rates of fire spread when burning conditions are uniform. To deal with spatial nonuniformity of fuels, weather and topography the field must first be partitioned into homogeneous units. This partitioning allows the use of the spread model within each unit (Fujioka 1985). 217 MAP is a raster based GIS designed to run on IBM compatible microcomputers. It has a grid cell data structure in which map information is stored as numeric values in arrays, each cell representing an uniform parcel of land located within the overall rectangular grid. MAP provides for storage, processing and display of cartographic data allowing for input in the form of grid· cells, digitized points, lines or polygons. The processing capabilities consist of spatial data base management, spatial statistics and cartographic modeling that use sequential processing of mathematical operations plus maps and a common database to store intermediate results (Berry and Reed 1987). The simulation of fire spread is based on the "distance function" spread. This distance accumulating process can be limited to upward or downward directions over a 3-dimensional surface (Tomlin 1985). FIREMAP Structure and Interface There are three main sections in FIREMAP. The first section generates the input overlays required to run the fire model which are based on meteorological data, time of the day, month of the year, an altitude overlay, a vegetation overlay, and a stream channels overlay, by following the framework described in Rothermel (1983). The second section consists of a program, written in FOR- TRAN77, that reads the maps as arrays, runs the fire model and creates the output overlays that store the values describing fire characteristics for each cell in the data base (fig.l). The third section consists of the actual simulation of the fire spread for the given conditions. A source (or sources) of ignition, is selected. It is assumed that the wind is consistently blowing from the same direction, and an overlay with a constant inclination in the direction of the wind is created (WIND). The spread operation previously described, is used over this surface, through an overlay, FRICTION, that has assigned to each grid cell the number of time units it takes the cell to be consumed by a fire, with the given conditions. This calculation is based on the rate of spread (feet/ min) and size of each cell (feet). The value of friction assigned to stream channels is the result of a calibration done for this particular situation. Using spread operation in the direction of the wind, fire spreads preferentially through the path of least resistance, or the cells taking the least time to be consumed by the same fire, up to the point where the predefined simulation time is reached. The other output overlays (heat per unit area, fireline intensity, effective windspeed, flame length, reaction intensity and direction of maximum spread) give useful mapped information about the characteristics of the fire in each of the grid cell units, for the time interval on which the weather conditions utilized apply. Because the fire model predicts characteristics of the fire in the flaming front, they are valid only when the fire is still burning on that cell. Three-dimensional displays of the areas burning can be included. Mountain area, on the Fort Apache Indian Reservation in east-central Arizona. It corresponds to an area 9 square miles in size with elevations ranging from 5800 to 7000 feet, on steep slopes. The vegetative cover consists of three types: ponderosa pine stands (Pinus ponderosa) with variable crown and understory densities, pine-Douglas fir stands (Pinus ponderosa-Pseudotsuga menziesii), and pinyon-juniper. The vegetation management alternatives considered are no intervention and harvesting by selective cutting (fig. 2). To analyze the influence of one variable in the system, all other variables have to remain constant. Here spatial arrangement of vegetation, under harvest or non-harvest conditions is the driving variable under consideration; therefore, weather conditions and source of the fire are the same for the two simulations. The weather data utilized were taken at Ivins Canyon on June 11, 1988 at 3:00p.m.: Dry bulb temperature- 82° F; Relative humidity -12%; Windspeed at 6ft- 12 m.p.h.; Wind direction - South. Simulations are performed for a period of three hours, whe:re each time step is one hour long. Another example is given Results The final result of the simulation is shown on figure 3. Figure 4 shows how a change in the wind direction (accounted for between time steps) can affect the area and shape of the burned area, illustrating the utilization of FIREMAP for a known weather situation. Figure 5 displays the expected flame lengths for the no intervention alternative during the three hours of simulation. Conclusions The results of the FIREMAP application described here indicate that the approach followed works reasonably well. The integration of a fire model like BEHAVE with a GIS may be an efficient way of accounting for spatial variability when attempting to predict fire behavior. FIREMAP, in the same way as BEHAVE, is a direct implementation FUELS SLOPE WIND SPEED WIND DIRECTION MOISTURE FIRE MODEL RATE OF SPREAD FLAME LENGTH HEAT/UNIT AREA Application of Firemap FIRELINE INTENSITY Problem Description REACTION INTENSITY The area considered in this example is located in the Spotted on how a change in wind direction from south to west, at the end of the second hour would affect the area burned. Figure 1.-Using GIS as a data base. 218 COUER TYPES Ponde~osa Pin• Pinuon-Ju~ipep II II k:::·J Pine-Douglas Ji~ [ ] Ha~u@sted A~eas Figure 2.-Vegetation: left - no Intervention, right - harvested. AIEAt SOU¥-Ct Art~r on~ hou~ BURt~D II II II [] Arter t1111o llours Afte~ th~ee ~ou~s Figure 3.-Simulation results. 219 of Rothermel's fire spread model, and the predictions it makes are subject to the limitations and assumptions of the same model. However there are significant differences between this system and the BEHAVE system. The capabilities of FIREMAP for spatial/temporal simulation of fire behavior make it a useful tool that goes beyond the simple display of spatially summarized, rapidly available information. FIREMAP simulates the actual spread of fire predicting its varying intensity and showing areas burned for chosen time intervals. It can be used for "on site fighting," thanks to MAP capabilities for quick information update (like a change in fuel types due to clear cutting) predicting the extent and intensity of a fire for a certain period of time. In order to choose a fire management program it is necessary to consider not only ranges of fire behavior under various management alternatives, but also to assess the relative probability of occurrence of certain fire events. FIREMAP does not consider this latter aspect, a point that should not be overlooked in the decision-making process. It might be more practical to direct more attention on planning for those less probable ignitions that are likely to escape and cause extensive damage when they do occur (Salazar and Bradshaw 1986). Future work with FIREMAP should include validation and sensitivity analysis, and fine tuning of the module presently running. The prediction capabilities can be greatly increased by addition of other modules, either from BEHAVE or new ones. For example, a module to simulate spotting, or a module to compute scorch height can be easily integrated. The use of a more sophisticated GIS with flexible command language and built-in modelling modules, real number processing capabilities, and larger memory, will also help in AREAS BURNED (wind c~ange f~cM ar tel' one FLAME LENGTH S to W> J'ou~ h~u~s aft•~ t~~@@ hou~s art.P tMo •• • II ~ • 15ij5D Figure 4.-Area burned with a wind change. overcoming some of the present limitations of FIREMAP. Aknowledgements Information for this study was provided by the USDI Bureau of Indian Affairs. Uterature Cited Andrews, P. L. 1980. Testing the fire behavior model Proceedings, Sixth Conference on Fire and Forest Me- <£eet) Q 4 4 8 8 11 > 11 Figure 5.-Expected flame lengths. teorology [Seatle, Wash., Apr. 2224,1980]: Soc. Am. For., Washington, D.C. 70-77 Andrews, P. L. 1986. BEHAVE: Fire behavior prediction and fuel modeling system - BURN subsystem, part 1. USDA Forest Service, General Technical Report, INT-194, Intermountain Forest and Range Experiment Station. 130 p. -Berry, J.K. and Reed, K.L. 1987. Computer assisted map analysis: a set of primitive operators for a flexible approach. Paper presented at the 1987 ASPRS-ACSM Conven- 220 tion Baltimore, Maryland, March 29-April3, 1987. 7 p. Fujioka, F. M. 1985. Estimating wildland fire rate of spread in a spatially nonuniform environment. Forest Science, 31 (1): 21-29 Rothermel, R.C. 1972. A mathematical model for predicting fire spread in wildland fuels. USDA Forest Service, Research Paper INT-115, Intermountain Forest and Range Experiment Station. 40 p. Rothermel, R. C. 1983. How to predict the spread and intensity of forest and range fires. USDA Forest Service, General Technical Report, INT-143,Intermountain Forest and Range Experiment Station. 61 p. Salazar, L.A. and Bradshaw, L. S. 1986. Display and interpretation of fire behavior probabilities for long-term planning. Environmental Management 10 (3): 393402 Tomlin, C. Dana 1986 The IBM personal computer version of the Map Analysis Package. GSD /IBM AdS Project, Report No. LCGSA85-16 Laboratory for Computer Graphics and Spatial Analysis. Graduate School of Design, Harvard University. 60 p. 221