LBP Model Run Considerations and Interpretation

Landscape Burn Probability (LBP) is a new fire behavior model for many users. Although most of the inputs are similar to Landscape Fire Behavior the outputs are quite different. The information laid out below is very important and helps provide a better understanding of the nuances of selecting inputs, running the model, and interpreting the outputs.

This table summarizes the key information about the Inputs and Outputs. Click on the links for more in depth information about each item.

Run considerations ordered by input and output parameter.
Parameter Consideration
Inputs
Choosing a Landscape Extent Choose a Landscape size that encompasses the entire area within which you want to compare results. In most cases this will be a project area boundary but in some cases can be larger to encompass an entire ownership area such as a district or park.
Inputs such as wind direction or landscape features may affect outputs in unusual ways (see sensitivities) In some model outputs you may encounter areas with high conditional flame lengths and low burn probabilities. Causes for this can include fuels that are located in difficult locations for fire to spread to, such as highly flammable but patchy fuels. (see sensitivities)
Weather (wind and fuel moisture) Use near “Worst Case”: 90th Percentile or greater or “problem fire” scenario for your area. The model is not intended to be used in moderate or moist conditions.
Simulation Time Because inputs should represent “problem fire” conditions, the duration of the burn period should reflect the active burning period of wildfires of the area under extreme conditions.
Spotting Spotting Probability of 20% is the recommended input and the default in IFTDSS, however it can be changed if needed. Entering 100% means that all points on the landscape where crown fire is initiated will launch embers. A value of 0% in essence turns off spotting.
Outputs
Burn Probability and Integrated Hazard are unique to the landscape analyzed and influenced by extent. Compare pixel values ONLY within the same analysis/model run. Additionally, keep in mind these values are dynamic, burn probability class may change if you change the extent from the full landscape to a specific area of interest.
Analysis Maximum (Burn Probability Output)

The legend for Burn Probability output lists the “Analysis Maximum”, the highest burn probability for that analysis and landscape extent. The colored classes in the legend are scaled based on this value. The colored classes are not evenly distributed across all runs nor landscapes. Rather, they are classified in relation to the analysis maximum for the specific outputs and area displayed; some analysis outputs may be heavily skewed to just one or two classes depending on the range of results.

If seeing the analysis maximum displayed as a decimal is confusing to you, move the decimal point two places to the left and think of the value as the percent of times the most frequently burned pixel experienced fire during that analysis extent. (e.g. 0.00619237 implies that the maximum any pixel on that extent burned was in 0.6 % of the total simulated fires during in that particular analysis scenario.)

Conditional Flame Length Represents an estimate of the mean flame length for all the fires that burn a given point on the landscape during a LBP model run. Use the Identify tool in Map Studio to view this mean along with proportion of flame lengths for each class - the percentage of times flame lengths occurred in each class throughout all simulated fires.
Integrated Hazard Takes into consideration both Conditional Flame Length and Burn Probability.

LBP Model Input Sensitivities

Isolating on model sensitives on a particular landscape is the responsibility of the user through trial and error in the model calibration process. It may be important to iterate on a number of model runs changing one input at a time to understand how the features and fuels on your landscape influence the model outputs. This is an important part of the model calibration process. An example of model sensitivity to wind direction and landscape features is included below.

Example: wind direction and landscape features on burn probability

Inputs such as wind direction or landscape features may affect outputs in unusual ways, such as producing areas with high conditional flame lengths but low burn probabilities. Causes for this can include fuels that are located in difficult locations for fire to spread to, such as highly flammable but patchy fuels. Some examples of this are high elevation timber pockets, or shrub islands surrounded by non-burnable fuels (pictured below).

two images, left shows vegetation patches separated by water. The right image shows two stands of trees separated by short sparse grass.

The interaction of wind with landscape features can also contribute to low burn probabilities in seemingly volatile fuels, such as in cases where there are non-burnable or slow-spreading fuels upwind. A visual example of this is displayed below.

in the landscape with SE winds most of the area burned, in the landscape with NW winds, there are patches of unburned area

The two burn probability outputs were generated with identical inputs with the exception of wind direction. The first output was run with a Southeast wind and the second with a Northwest wind. You can see in the example with NW Wind, that the burnable area on the east end of the lake was sheltered by the water and did not burn (black polygon) whereas it did burn in the same scenario under SE winds.

Click the input link HERE to leave feedback about this page.