A1. The only data generally available used in FARSITE are DEMs (Digital Elevation Model) from the USGS. A GIS can create the slope, aspect, and elevation themes from a DEM. All fuel-related data must be produced using a GIS - essentially mapping the different fuel layers over the landscape. Many federal land management agencies are developing fuel data for use in FARSITE but data do not exist for most of the US yet. Typical data sources are 1) existing vegetation maps that can be cross-walked to fuels, 2) satellite image analysis along with ground-truth data, 3) aerial photography and manual interpretation.

Q2. Has FARSITE been validated?

A2. The individual sub-models that calculate separate fire behaviors have each had varying degrees of testing by their respective developers. FARSITE includes these models for surface fire spread, crown fire spread and transition, fire acceleration, spotting from torching trees, and dead fuel moisture. FARSITE itself has undergone several years of testing and comparison with observed fire perimeters (see Limitations and Assumptions) but comprehensive validation has not been finished. This is largely a consequence of having little control over error in the input data (poorly defined winds and unknown precision of fuel maps) and data used for comparison, specifically the observed fire progression maps (fire locations often contain considerable error in space and time, for example, the fire position was guessed because it could not be seen through smoke, and the time on the fire position was noted only to the nearest day, not hour). These sources of imprecision make it impossible to distinguish model error from data error. Many people have shown that, like all models, a user who understands how the models work and their limitations can obtain useful results from FARSITE.

Q3. Will FARSITE work for my local area?

A3. This question could probably be re-phrased as "what does it take to make it work for my local area?". First, it requires you to understand the fire behavior in relation to fuels and weather in your particular setting. Given this knowledge and understanding the limitations of the individual fire behavior models in FARSITE, you can deduce what will be required in terms of fuel mapping, weather conditions, and expected accuracy of the simulation results. If you cannot get acceptable results using the individual fire behavior models (i.e. surface fire spread in BEHAVE), then it is unlikely that FARSITE will do any better.

Q4. FARSITE seems to give me different results than BEHAVE, why?

A4. Although the exact same surface fire model is used in both systems, there are several reasons why your surface fire behavior produced by FARSITE might seem to be different from what you are comfortable calculating with BEHAVE. Most of the difficulty seems to be related to the fact that the wind speed input is a midflame value in BEHAVE but an open wind speed (20') in FARSITE. The midflame wind speed is calculated automatically in FARSITE using tree height and canopy cover layers. It often seems that midflame wind speed is overestimated by users of BEHAVE for fires burning in the understory. Wind speed is often adjusted until BEHAVE produces an expected fire behavior characteristic, like flame length. This is useful to account for fuel factors that are not present in the standard surface fuel models, like spatial variation in fuels (jackpots, patches of brush, ladder fuels) that produce important fire behaviors for suppression, effects, or torching trees. However the implicit logic of adjusting midflame wind speed to match fire behavior cannot be translated to the simulation directly. Although a midflame wind speed of 5mph in BEHAVE is not unreasonable in itself, the wind reduction factor for flat land under a forest canopy might range from <0.1 to 0.2, meaning that a 5mph midflame wind would translate to a 25 to >50mph wind at 20' !!! If your surface fuel complex really does produce a higher surface intensity than is produced by FARSITE, then you will have to change some of the following: the surface fuel model, make a custom surface fuel model, increase your 20' wind speed, lower your fuel moistures and relative humidity, decrease stand height and canopy cover.

Q5. If the crown fuels are "optional" in FARSITE, how do I calculate crown fire behavior?

A5. The crown fuels (stand height, crown base height, and crown bulk density) are only optional in the sense of building a Landscape (.LCP) File and getting the simulation to run. They are definitely not optional if you have important spatial variation in these themes on your landscape and need to model crown fire. If you do not have significant spatial variation in these themes, then you can use the landscape wide settings in the "Canopy Characteristics" dialog box. These themes are called "optional" because they are sometimes difficult to create - and should not be an obstacle to running FARSITE in areas with mainly surface fire behavior.

Q6. How can I calculate fire effects with FARSITE?

A6. There are some fire effects that can be related to the fire behavior outputs of FARSITE (see Output > Export & Output). For example, through the relationship between fire line intensity and scorch height, tree mortality rates can be calculated with additional information on tree species and size. These calculations need to be performed in a GIS.

Q7. What is the difference between "canopy cover" and "crown closure"?

A7. These terms are often used interchangeably, but unfortunately, they are not the same at all. Canopy cover is a physical measurement of the average fraction of the horizontal ground surface occupied by the horizontal projection of the tree crowns. It is largely unrelated to "crown closure" which is an ecological term describing the development of a forest stand. A stand can reach 100% crown closure (where no other individual trees can squeeze into the canopy) and still remain at 50% canopy cover. It would be extremely rare for a stand to have 100% canopy cover. Canopy cover is the measurement needed for FARSITE because it is used to compute wind reduction (via filling fraction of the crown space) and shading. You will get too much wind reduction and shading if you base your canopy cover input on crown closure.

Q8. Can I use FARSITE to estimate smoke production?

A8. Yes, but you must provide two additional spatial data themes to the Landscape (.LCP) File (duff loading and a coarse woody fuel profile index (index is just a number for reference to a description of the profile, like a fuel model number relates to the description of a custom fuel model). You must provide the loadings of the large woody fuels associated with the coarse woody fuel profile number. Then, run your simulation with the post-frontal combustion enabled (Model > Post-frontal Combustion). Tables, graphs, and maps can be produced to show you amounts and rates of emissions, heat, and fuel weight loss rates.

Q9. How does FARSITE handle winds in complex terrain?

A9. Not very well at present, mainly because we do not have access to data that represent spatial variation in winds or weather. FARSITE uses winds as a stream of data describing temporal variations in speed and direction. This is a non-spatial wind stream, and you must use more than one weather and wind stream to approximate a limited amount of spatial variation in winds or weather variables (see Simulate > Modify Map > Weather Monitoring Grid). FARSITE does allow you to use spatial weather and wind grids that are produced externally to FARSITE (by some atmospheric simulation, see Gridded Weather). However, most people to not yet have access to these kinds of models.

Q10. How does FARSITE deal with fuel moisture variation, dead and live fuels over time and space?

A10. Dead fuel moisture is computed by the simulation over time and space using the initial fuel moisture file (by surface fuel model) and the weather and wind streams you provide. The fuel moisture model is the same one used for fine fuel moisture calculation as used in BEHAVE along with some from the NFDRS. Every time FARSITE needs to make a fire behavior calculation, it calculates the current values for dead fuel moistures starting from their initial condition. The weather data (temperature and humidity at a given elevation) are adjusted for the elevation of the local calculation. Solar irradiance effects on fuel moisture are calculated for each fuel location based on slope, aspect, canopy cover, and cloud cover. This gives a degree of sensitivity in fuel moisture at each cell on the landscape to topography and overstory vegetation structure. The influence of the initial fuel moisture condition on fire behavior is often not noticeable after a few hours of simulation time for the finest fuels because these have short time-lags and respond quickly to new weather conditions. Live fuel moistures specified for a given surface fuel model remain constant throughout the entire simulation.

Q11. Why does FARSITE seem to over predict fire growth rates?

A11. The general reason is that the scale of the input data that describe the fire environment to the models is very coarse compared the scale of the real fire environment that affects the actual fire. This means that the fire behavior models are fed very homogenous data (spatial and temporal) that produce equilibrium values that are too fast for the real heterogeneous environment. For example, we input winds at an hourly interval but know that in reality the winds are highly variable. The nonlinear response of the fire behavior models to wind means that the average wind value does NOT produce the average fire behavior value. Another example is the assumed fuel homogeneity at the resolution of the input raster data (say 30m). The raster landscape data contain no variation in the fuel data finer than 30m, but in reality most fuels are more heterogeneous, having areas within each 30m cell that have more, less, faster, and slower fuel types. This heterogeneity keeps the actual fire changing in response to the different fuels, accelerating and decelerating, and produces a resulting fire spread rate through the 30m cell that will be slower than calculated for an equilibrium condition using the input data. This heterogeneity seems to be increasingly ignored by the fire as the burning conditions become more extreme. This is reflected by the experience that under very dry and windy conditions, the adjustment factors in FARSITE rise to about 1.0, meaning the models need no adjustment, because the fire is not sensitive to much of the temporal or spatial heterogeneity in the real environment.

Q12. Why do we need adjustment factors in FARSITE and how do I know what values to use?

A12. See the answer to the previous question regarding over prediction. The adjustment factors are there to make the fire spread rate correspond to the observed rate of progress over the landscape. Under mild burning conditions, more adjustment seems to be necessary. To obtain adjustment factors for a given area you will have to compare FARSITE projections with actual fire progression data. Keep in mind that the adjustment factors assume that the fuel model is correctly specified and will not be useful if your fuel model map is inaccurate. In that case you will need to make changes to the map before doing any adjustment.

Q13. Why does spotting have a stochastic component to it in the simulation?

A13. There are many things about spotting that are unknown and probably unknowable. First, the number of embers is not calculated; experience has suggested that there are so many embers produced by torching trees that their numbers are not the limiting factor in producing spot fires. Embers can be assumed to be produced in sufficient quantities to effectively cause spotting if conditions of the receptive fuels permit ignition by the brands. The model calculates lofting and flight for embers of different sizes which have different initial heights, trajectories, and burning times. Which of these embers that arrives alight will land on receptive fuel is not known. Ignition and growth of a spot fire to a self-sustaining fire is dependent on fuel conditions that vary at a fine scale compared to the input data (say 30m). These receptive fuels are often rotten wood, stumps etc. that cannot be described quantitatively at the scale of the input data and therefore, must be randomized.

Q14. Is the "ignition frequency" setting in FARSITE the same as "probability of ignition"? If not, then how do I know what to use for the ignition frequency?

A14. Nope. Probability of ignition will be much higher than a practical setting for ignition frequency. The setting you use for ignition frequency is typically very low (like about 1-5%). You need to set ignition frequency so that FARSITE is producing a number of spot fires that are both practical (in terms of computing power required) and realistic (in simulating the frequency and importance of spotting to the actual fire). As conditions become more extreme, and torching and crowning become more prevalent, you might need to lower the ignition frequency because there will be so many embers produced that you don't need that many to simulate the fire behavior that is being observed.

Q15. My fire simulations don't seem sensitive to rivers that I bring in as vectors, why?

A15. FARSITE only uses raster data for fire behavior calculations. The vector data that you bring in are only for orientation and getting a clearer picture of where the fire is burning relative to features that are important to you. This means that any feature smaller than the raster resolution will not be "visible" to the simulation. You must use other information to make the fire stop at real barriers, like a large river or a trail used as a control line. You can import vector data for road segments, or stream segments as "barriers" to surface or crown fire (but not spotting) with the Simulate > Modify Map > Import Barrier File command.

Q16. Where do I get weather and wind data to put into FARSITE?

A16. This kind of depends on what you are using FARSITE for. There are 3 general uses for FARSITE: 1) fire reconstruction (past fires), 2) fire projection (active fires), and 3) fire planning (potential fires). You should use observed weather data for #1, either from field observers or from a representative RAWS station. For #2, we need to generate them by hand -- just like a horse race, we can't get the results ahead of time, only an expectation or forecast. So, for a short term fire projection, we will probably want to use base our Weather (.WTR) and Wind (.WND) Files on a real forecast for the fire locality. For medium to long term projections (beyond the forecast) we will need to create weather and wind scenarios based on the concerns of the decision makers on the fire, run all of them and compare them to see what the differences are. For #3 we can use climatological information to devise conditions and scenarios that are of interest to the project. This might mean that we use NFDRS stations or RAWS stations and summary statistics from each for percentile weather conditions or the frequency of certain conditions that are of special concern.

Q17. Can I bring GPS data on fire position into FARSITE?

A17. Yep. All GPS systems have software that can output or transform data to a given format. If you can get the data into ARC/INFO or ArcView, then you can get them into FARSITE. FARSITE allows data in ASCII vector format or in the 2D ArcView Shapefile format. For example a fire perimeter that has been surveyed by GPS can be imported as the ignition for a simulation.

Q18. I'm having trouble getting the simulation to produce the fire growth and behavior that we observed, what can I do to get better simulations?

A18. There are 3 sources of error that we look for in simulations. Data, User, and Model. Because we scrutinize the errors in that order, the acronym for these is DUM. First we look for data errors which can be lurking in the fuels or weather data. Are the fuel models mapped as accurately as possible? What about the crown fuels? Are the weather and wind data reliable and reasonable? Ultimately, you want the fire behavior to be correct over a range in burning conditions, so you need to check the fuel and weather inputs that dominate a given type of fire behavior that you are not presently seeing in the simulation. Second, user error is a constant problem, even for the most experienced modelers. We all make mistakes, thinking we've made a change and haven't or are using one data set versus another. One of the most common problems is with data units in the Landscape (.LCP) File. Use the Input > Landscape Utilities > File Information command to look at the contents and units of a Landscape (.LCP) File. We need to double- and triple-check our assumptions and inputs before asserting that the model doesn't work. However, all models have limitations, and at some point, if the Data and User problems are fixed, we need to assess whether the fire behavior meets the assumptions and limitations of the fire behavior models in FARSITE - maybe or maybe not.

Q19. It takes me too long to get the data organized to run the simulation when we get fires, what strategies are there for data preparation to make this process smoother?

A19. You need to get organized before fire season begins - namely divide your area up by useful sub-units, like USGS 7.5 minute quadrangles. You can make Landscape (.LCP) Files for each quad or unit ahead of time, either one at a time in FARSITE or using the LCPMAKE.EXE utility in the FARSITE directory. You will need to get your data into the proper units however, before you use LCPMAKE. For a given fire you can use the Input > Landscape Utilities > Merge Landscape Files and Input > Landscape Utilities > Extract Portion of LCP File commands to get the dimensions of the landscape to an appropriate size. You can also collect and analyze weather records and come up with some standard weather/wind scenarios that are of typical interest during your fire season. These can then become the starting point for rapid modifications when a fire occurs and you have specific conditions that need to be modeled. Lastly, you can and should do some simulations for past fires and learn what it takes to use your fuel maps and weather data to make FARSITE given you reasonable results.

Q20. I noticed several executable files also come with FARSITE, (RAWS2FAR.EXE, LCPMAKE.EXE, V2SHAPE.EXE, CONCAT.EXE) what are these for?

A20. They are command-line utility programs that must be run in a DOS window and might be useful for certain tasks that you encounter. On the command line, just type the name of the .exe file and it will give you some instructions on how to use it. LCPMAKE.EXE requires a number of switches and arguments and will make a Landscape (.LCP) File from ASCII GRID files. It is useful for making many Landscape (.LCP) Files and can be run from a batch file (.BAT). RAWS2FAR.EXE converts an ASCII RAWS file into 2 FARSITE format files for weather (WTR) and wind (WND). V2SHAPE.EXE converts an ARC ASCII vector file into an ArcView Shapefile. CONCAT.EXE concatenates two ARC ASCII vector files into a single file. This is useful if you have your vector files clipped to individual quadrangles and want to merge them.

Q21. When running a simulation, how can I tell what the fire behavior is doing at a given spot on the landscape?

A21. There are several ways. The first indication is the space between the time contours of the fire perimeter (white lines). Faster spread is indicated by wider spacing between the lines. Second, using the Output > Fire Perimeter Colors command you can specify that the fire perimeter be color coded to reflect one of the frontal fire behavior characteristics, like flame length. Be careful with the colors of the background landscape so that the different colors on the perimeter will not blend in with the background. Third, you can suspend the simulation and use the Fire Data Query feature to click inside the perimeter near the edge and see a list of the frontal fire characteristics. Fourth, the most comprehensive way to do this is to select an output raster file to be made as you are simulating (Output > Export and Output). You can then view this raster file at any time (View > Raster File) and query the landscape and look at the "overlay value" in the land data window.