Integrating fire spread patterns in fire modelling at landscape scale

Highlights

• We used fire spread patterns to simulate fire spread in landscape succession models.

• Modelling fire spread patterns improved simulations of fire propagation.

• Factors governing fire spread differed among topographic, convective and wind fires.

• Synoptic weather situations can populate fire spread modelling at large spatial scales.

Abstract

Fire spread modelling in landscape fire succession models needs to improve to handle uncertainty under global change processes and the resulting impact on forest systems. Linking fire spread patterns to synoptic-scale weather situations are a promising approach to simulating fire spread without fine-grained weather data. Here we present MedSpread—a model that evaluates the weights of five landscape factors in fire spread performance. We readjusted the factor weights for convective, topography-driven and wind-driven fires (n = 123) and re-assessed each fire spread group's performance against seven other control simulations. Results show that for each of the three fire spread patterns, some landscape factors exert a higher influence on fire spread simulation than others. We also found strong evidence that separating fires by fire spread pattern improves model performances. This study shows a promising link between relevant fire weather information, fire spread and fire regime simulation under global change processes.

Introduction

Fire models are designed to reproduce essential fire regime descriptors (Sturtevant et al., 2009, Sullivan, 2009). The last ten years have seen a surge in the development of fire models (Miller and Ager, 2013) as part of a wider effort to capture the essential processes driving fire dynamics in real landscapes. Fire models reproduce specific fire regime attributes that serve to assess fire impacts on different scales of applicability. While some models focus on fuel-heat transfers at small scale, other models are able to simulate fire dynamics and patterns at regional and long-term scales. Models that spatially reproduce fire spread can be collapsed into two groups according to the scale considered and processes modeled: the fire level and the landscape level.

The first group of models working at the fire scale (Keane et al., 2004), known as Fire Growth Models (FGMs), simulate fire spread growth of single events and mainly aim to support operational decision making and assess the effectiveness of different fuel treatments on fire behavior and spread (Duff and Tolhurst, 2015, Stratton, 2004). FGMs use detailed spatial and temporal information on weather, fuel and topography affecting fire behavior and spread to reproduce the potential growth of fires (Albini, 1976, Anderson, 1983, Rothermel, 1983). Farsite (Finney, 2004) and Prometheus (Tymstra et al., 2010), for example, have been used for contrasting purposes in different countries (Salis et al., 2013, Suffling et al., 2008). However, the complexity characterizing fire as a process makes each event highly specific and context-dependent, thus introducing significant constraints on the extrapolation of model results to other contexts or fire events (Andrews and Queen, 2001, Zhou et al., 2005).

The second group of models simulate multiple fire events at the landscape scale and reproduce long-term fire regimes shaped by dynamic interactions between wildfires, vegetation and climate on wide temporal and spatial scales (e.g. Boychuk et al., 1997, Brotons et al., 2013, de Groot et al., 2003, He and Mladenoff, 1999, Keane et al., 2002, Loepfe et al., 2011, Millington et al., 2009). Known as Landscape Fire Succession Models (LFSMs), they simulate specific fire regime properties operating at landscape scale, such as fire occurrence or frequency in large areas. These models are often capable of handling a range of factors influencing forest landscape dynamics and fire regimes, such as climate or land-use management (Keane et al., 2004). However, as LFSMs answer questions that tend to target coarser spatio-temporal scales, fire spread modelling usually passes unnoticed and the underlying physics is largely simplified. Fire spread simulation within LFSMs ranges from predetermined fire shapes (Green et al., 1983) to dynamic lattice or vector spread strategies determined by probabilistic functions or empirically-based equations (Adou et al., 2010, Keane et al., 2004, Sullivan, 2009).

There is a challenge to bridge the current gap between FGMs and LFSMs (Sturtevant et al., 2009). Although the two kinds of models have been designed to achieve different goals, large-scale long-term LFSMs performance could be improved by including key processes that reproduce fire spread in a more reliable way. Improved performance over a wider range of temporal and spatial scales of final fire shapes may eventually lead to a better assessment of several aspects tied to operational suppression needs, effectiveness of vegetation treatments, effects of treatments designed to preclude runoff or post-fire regeneration patterns at these scales (Gil-Tena et al., 2016). The resulting fuel heterogeneity from a simulated fire, in turn, may influence the spatial pattern of subsequent fires (Turner and Romme, 1994, Yang et al., 2008).

Fire spread is determined by weather, topography and fuel (Keane et al., 2004, Parisien and Moritz, 2009). The specific contribution of these factors to fire propagation is still unknown, and several studies have shown that relative influence of weather, topography and fuel can vary (Gardner et al., 1999, Green et al., 1983, Mouillot et al., 2001, Turner et al., 1989). Of these factors, weather conditions present the most variability within and between fires (Rothermel, 1983). The complexity of weather conditions is not easily translated into fire modelling frameworks capable of extrapolating calibration results at local scales from one fire to another (Andrews and Queen, 2001). Furthermore, model requirements to adequately and accurately reproduce fire spread are usually highly complex and reliant on data at fine temporal and spatial resolutions on weather changes during a given fire event (Hargrove et al., 2000). However, fire spread patterns do tend to be repeatable and often predictable in time and space (Duane et al., 2015). In Mediterranean ecosystems, these patterns have been described as convective fires, wind-driven fires and topography-driven fires (Castellnou et al., 2009, Duane et al., 2015). These spread patterns can be related to specific synoptic weather situations, which in turn dictate general weather conditions at a regional-landscape scale. A synoptic weather situation describes general atmospheric characteristics prevailing in a region over a temporal span of hours to days, and defines the relation between general atmospheric circulations and surface conditions (Crimmins, 2006). Synoptic weather situations may therefore be the appropriate factor-weather scale influencing coarse spatial fire patterns (Turner et al., 2001). Fire spread patterns could then be used as a better approach to reliably simulate fire spread without needing detailed weather data, which is difficult to gather over long-term periods in future climate projections without high levels of associated uncertainty.

Once under the influence of a synoptic weather situation, the specific contribution of the multiple drivers governing fire spread (slope, wind, etc.) can be different for each spread type. A fire could therefore become more affected by wind than fuel structure in a windy situation (Jin et al., 2014, Moritz, 2003), whereas vegetation flammability and structure may have a higher influence in other situations (Artès et al., 2015). Thus, under each synoptic weather situation, fire spread drivers could have different roles in determining final fire perimeters, thus offering a promising link between local fire spread patterns and fire regimes at the landscape scale.

The aim of this study was to assess the potential advantages and limitations of incorporating fire spread patterns defined by synoptic weather situations into a fire spread algorithm in a LFSM context. Here we present MedSpread, a landscape fire spread model that reproduces fire spread from the ignition point of a fire of predefined size. By fitting actual fire scars occurred in a Mediterranean area from 1989 to 2012, we attempted to assess the performance of MedSpread when including the main fire spread patterns (i.e. wind-driven, topography-driven and convective fires). First, we assessed the contribution of each driver potentially affecting fire spread for each of the different fire spread patterns and discussed their role through a sensitivity analysis. Second, we calibrated the relative contribution of each of these factors on fire spread for each of the three main fire spread patterns documented in the study area. Third, we attempted to determine the potential improvement of fire spread performance for fires only influenced by one factor alone. Finally, we discuss the incorporation of fire spread patterns into a LFSM as a way to bridge the gap between fire spread and landscape fire models by boosting fire spread model performance.