Portier Jeanne, Gauthier Sylvie, Robitaille André, Bergeron Yves
1Département des Sciences Biologiques, Centre for Forest Research (CFR), Université du Québec à Montréal, Case postale 8888, Succursale Centre-ville, Montreal, QC H3C 3P8 Canada.
2Natural Resources Canada, Canadian Forest Service, Laurentian Forestry Centre, P.O. Box 10380, 1055 du PEPS, Stn. Sainte-Foy, Quebec, QC G1V 4C7 Canada.
Landsc Ecol. 2018;33(1):19-34. doi: 10.1007/s10980-017-0578-8. Epub 2017 Oct 24.
Wildfires play a crucial role in maintaining ecological and societal functions of North American boreal forests. Because of their contagious way of spreading, using statistical methods dealing with spatial autocorrelation has become a major challenge in fire studies analyzing how environmental factors affect their spatial variability.
We aimed to demonstrate the performance of a spatially explicit method accounting for spatial autocorrelation in burn rates modelling, and to use this method to determine the relative contribution of climate, physical environment and vegetation to the spatial variability of burn rates between 1972 and 2015.
Using a 482,000 km territory located in the coniferous boreal forest of eastern Canada, we built and compared burn rates models with and without accounting for spatial autocorrelation. The relative contribution of climate, physical environment and vegetation to the burn rates variability was identified with variance partitioning.
Accounting for spatial autocorrelation improved the models' performance by a factor of 1.5. Our method allowed the unadulterated extraction of the contribution of climate, physical environment and vegetation to the spatial variability of burn rates. This contribution was similar for the three groups of factors. The spatial autocorrelation extent was linked to the fire size distribution.
Accounting for spatial autocorrelation can highly improve models and avoids biased results and misinterpretation. Considering climate, physical environment and vegetation altogether is essential, especially when attempting to predict future area burned. In addition to the direct effect of climate, changes in vegetation could have important impacts on future burn rates.
野火在维持北美北方森林的生态和社会功能方面发挥着关键作用。由于其蔓延方式具有传染性，在分析环境因素如何影响野火空间变异性的火灾研究中，使用处理空间自相关的统计方法已成为一项重大挑战。
我们旨在展示一种在燃烧率建模中考虑空间自相关的空间明确方法的性能，并使用该方法确定1972年至2015年间气候、物理环境和植被对燃烧率空间变异性的相对贡献。
利用加拿大东部针叶林地区48.2万平方公里的区域，我们构建并比较了考虑和不考虑空间自相关的燃烧率模型。通过方差分解确定了气候、物理环境和植被对燃烧率变异性的相对贡献。
考虑空间自相关使模型性能提高了1.5倍。我们的方法能够纯净地提取气候、物理环境和植被对燃烧率空间变异性的贡献。这三组因素的贡献相似。空间自相关程度与火灾规模分布有关。
考虑空间自相关可以显著改进模型，避免有偏差的结果和错误解读。综合考虑气候、物理环境和植被至关重要，尤其是在试图预测未来火烧面积时。除了气候的直接影响外，植被变化可能对未来燃烧率产生重要影响。
