Data from: Multi-temporal analysis reveals that predictors of mountain pine beetle infestation change during outbreak cycles

Over the past two decades, severe mounta in pine beetle (MPB) outbreaks have affected several million hectares of forest in western North America. The extensive ecological and economic damage caused by widespread insect infestations make understanding the development and spread of MPB outbreaks critical. This study uses a time series of Landsat5 TM and Landsat7 ETM + images to map the spread of mortality due to MPB infestation in Arapaho–Roosevelt National Forest, Colorado, between 2003 and 2010. The Normalized Difference Vegetation Index (NDVI) and change in the Normalized Difference Moisture Index (NDMI) were used to classify red attack and non-red attack stands based on a maximum likelihood algorithm with manually selected training classes. The classification was validated by comparison with independent interpretations of aerial photography and high-resolution satellite imagery. The classification had good agreement (84.5–90.5% total accuracy). Cluster analysis for time series showed infestations originating in several different locations on the landscape early in the time series and subsequent infestations likely represent a combination of dispersal from outbreak populations and independent population growth. Analysis using conditional inference trees suggested that a combination of forest composition, topography, and dispersal predicted the distribution of MPB infestation on the landscape and that the importance of these variables changed as the outbreak developed. In early years, red attack was associated with forest and topographic characteristics known to influence susceptibility to MPB. Over time, beetle pressure became an increasingly important predictor of red attack, but in later years host tree availability played an important role in outbreak spread. If this pattern occurs consistently in MPB outbreaks, knowledge of these patterns could aid managers in targeting their efforts to reduce damage resulting from MPB outbreaks.

在过去的二十年里，严重的松树甲虫（MPB）爆发已影响北美西部数百万公顷的森林。广泛的昆虫侵害所造成的广泛生态和经济损害，使得理解MPB爆发的发生和发展至关重要。本研究利用Landsat5 TM和Landsat7 ETM+图像的时间序列，绘制了2003年至2010年间科罗拉多州阿拉帕霍-罗斯福国家公园内因MPB侵害而导致的死亡率蔓延图。通过最大似然算法，基于人工选择的训练类别，利用归一化植被指数（NDVI）和归一化差分湿度指数（NDMI）的变化，对红攻击和非红攻击林分进行分类。该分类通过与独立的航空摄影和高分辨率卫星图像的解释进行比较进行验证，分类结果具有良好的一致性（总准确率84.5-90.5%）。时间序列的聚类分析表明，在时间序列早期，景观上有几个不同位置的侵害起源，后续的侵害可能代表爆发种群扩散和独立种群增长的组合。条件推断树分析表明，森林组成、地形和扩散的组合预测了MPB侵害在景观上的分布，并且随着爆发的进展，这些变量的重要性发生了变化。在早期年份，红攻击与已知会影响MPB易感性的森林和地形特征相关。随着时间的推移，甲虫的压力成为红攻击日益重要的预测因素，但在后期年份，宿主树木的可用性在爆发蔓延中发挥了重要作用。如果这种模式在MPB爆发中持续出现，对这些模式的认识可以帮助管理者将他们的努力集中在减少MPB爆发造成的损害上。