Surface and Canopy Heat Fluxes Parameterization Improvements in WRF-SFIRE

Coupled atmospheric fire models are considered one of the best ways to simulate large wildfires.

They describe not only the progression of the fire front but also the modification of the atmospheric dynamics caused by the fire's intense heat flux. In this approach, the fire is represented within the atmospheric model through the sensible and latent heat fluxes produced at the fire front. Therefore, the computation of sensible and latent heat fluxes is a critical framework component.

In WRF-SFIRE, the heat fluxes are proportional to the surface fuel consumption, represented as an exponential decay with one parameter that controls the characteristic time of consumption. Thus, it is not possible to represent the complex dynamics of the rapid combustion of fine elements and the slow consumption of large elements, as well as the contribution of the canopy when the latter ignites.

This work proposes to improve the ability of the WRF-SFIRE model to calculate fire-related heat fluxes in two respects: i) to make the time profile of surface fuel consumption more complex by independently calculating consumption for several sizes of fuel elements, and ii) to take into account the influence of the canopy on the rate of spread and heat fluxes.

The first step is to divide the surface fuel into four size classes (<0.25in, <1in, <3in, >3in), otherwise known as 1h, 10h, 100h, and 1000h fuels. As this categorization is already employed by the fuel moisture parameterization, the fuel load per element size is already known for the Anderson fuel model. Each element burns independently, and the contributions are aggregated to calculate surface heat fluxes. To get away from exponential profiles, which greatly simplify real consumption dynamics, we will use a novel multi-dimensional combustion kinetics solver. This model describes the different phases of thermal degradation of fuel components, characterized by their chemical composition. This provides a consumption time profile for each element size, considering detailed chemical kinetics like pyrolysis, gas-phase reactions, and char oxidation. The model has been validated at various scales by comparison with thermogravimetric experiments as well as pyrolysis and combustion experiments carried out at the particle scale.

The obtained profiles are added to WRF-SFIRE, and a linear interpolation is used to determine a consumption value for each element size at a given time. The four profiles reflect the rapid combustion of fine elements, which burn off in a few seconds, and the slow consumption of large elements, which degrade over several hours. As a result, the finer elements will give a short-lasting intense flux (flaming phase), while the coarser elements will produce a moderate or weak heat flux for long periods (smoldering). This representation, called the Multi-Layers Heat Fluxes (MLHFX) scheme, is currently only used for surface fuels.

The second improvement is the integration of a crown fire model in WRF-SFIRE. Scott's parameterization, using one of the five rate-of-spread models available in WRF-SFIRE, is implemented to compute the rate of spread in three different situations: surface fire, passive crown fire, and active crown fire. In addition, it provides an estimate of the crown fraction burning, a variable used to calculate the heat fluxes associated with crown fire. The canopy heat fluxes are calculated the same way as for the surface with the MLHFX scheme, but using a single layer of fuel element and assuming that the crown fuel is homogeneous in size. This requires one consumption profile for the canopy fuel. For the latter, we use experimental profiles on familiar trees.

The heat fluxes calculated for the surface and the canopy are separated to manage their vertical distribution independently. Indeed, the current vertical distribution uses an exponential profile with a ground-level reference. The latter can still be used for surface fluxes calculated with the MLHFX scheme. A second exponential profile is added with a reference height at the canopy bottom for the vertical distribution of canopy heat fluxes. The contributions are then averaged over the 3D atmospheric grid. In addition, an “Exponential-Linear Modulated” profile has been added to study the impact of vertical heat flux distribution. This profile allows the maximum heat flux intensity to be shifted to a height greater than the reference height (the ground or canopy bottom). The parameter controlling this peak flux height can be user-defined, as is the parameter for the exponential distribution currently in use, or can be calculated from the fire line intensity.

To evaluate the benefits of these methods, we first compare their effects independently on idealized simulations. The latter highlights the interest in using the MLHFX scheme and accounting for crown fire in plume dynamics. Indeed, the more intense fluxes due to the fast-burning fine elements and the addition of canopy fuel promote a more intense plume with greater vertical velocities and a higher plume height, especially for fuel classes that contain coarse fuel elements.

To evaluate the methods in a more realistic context, the Williams Flats Fire is simulated. The MLHFX scheme is used alone and then in combination with the crown fire scheme in order to attribute the observed improvements to one or the other.

Finally, the integration of a new surface heat flux scheme, using multiple fuel-size elements that are consumed independently following a time profile derived from a detailed kinetics model, and the addition of a crown fire model improves the rate of spread estimations as well as the plume dynamics compared to observations.