Quantifying the risk of non-native conifer establishment across heterogeneous landscapes.
Pines (genus Pinus) are cultivated extensively for forestry purposes, particularly in regions that are outside the genus' native range. The most common forestry species are also typically those most likely to escape cultivation and spread rapidly, and thus pines constitute a substantial weed problem in many regions. However, there is limited knowledge of the factors underlying the spread of pines from plantations. Assessments across heterogeneous landscapes are required to provide rigorous data to support management tools and policies aiming to protect vulnerable ecosystems from pine invasions. We examined the spread of Pinus radiata from forestry plantations over a ~9,000 km2 landscape on Banks Peninsula, New Zealand. We used ground-based surveys from a set of viewpoints to determine tree locations, coupled with geographical information system (GIS) viewsheds to define the area surveyed. We used boosted regression trees to build a habitat model for P. radiata establishment on Banks Peninsula. We surveyed an area approximately 107 km2, recording 470 naturally established P. radiata individuals. Our habitat models suggested that proximity to the nearest plantation forest was the most important variable predicting P. radiata establishment, with individuals most likely to establish within 150 m of a plantation. Individuals were also most likely to establish in early successional shrub communities, proximate to roads, and on steeper topography. Highly grazed habitats were least vulnerable to P. radiata establishment. The slope and aspect of the source plantation influenced the distances from the plantation at which P. radiata individuals were recorded, with individuals recorded furthest away likely to have originated from plantations that were south-facing or on steeper slopes, and therefore most exposed to strong winds. Synthesis and applications. Our findings on distances from plantations at which individuals established, vulnerable habitats, and the interactions we detected among our predictor variables, can be extended to aid management of non-native conifer plantings elsewhere in the Southern Hemisphere. These data can be used to contribute to improvements of decision support systems that assess likely spread risk from non-native conifer plantings. Such tools can reduce the likelihood of future pine establishment, potentially preventing biological invasions.