S. H. Paul1, J. P. Dash1 and M. B. Scott2
1Scion, 49 Sala Street, Private Bag 3020, Rotorua 3046, New Zealand
2Scion, P.O. Box 29237, Fendalton, Christchurch 8041, New Zealand
Northern hemisphere Pinaceae are used very successfully as plantation trees in many southern hemisphere countries and provide significant economic benefits, but are also recognised for their invasiveness, and as transformational invaders in often highly valued environments such as the high country of New Zealand. As part of a successful management strategy, the level of invasiveness of certain species and their dispersal dynamics at specific sites, needs to be understood to predict the risk that existing populations create. Knowledge on the presence, abundance and demographics of invasive Pinaceae across the land is required for efficient control operations and long term successful management. To gain such knowledge, robust detection of infestations is required, whether it is obvious (e.g. dense and large trees) or difficult to spot (e.g. scattered small trees or in rarely visited sites) to inform targeted control of invasive conifers and to protect valuable environments. Our research aims to deliver methods for detection and mapping of invasive conifers at various spatial scales throughout New Zealand’s varied landscapes and across a range of ecosystems, and at the same time inform research on understanding the ecology of invasive tree species better. As part of such a large project, we will present the results of a detailed study in an area known for its vulnerability to pine invasions.
Our study site is a large local Pinaceae infestation which originated from a 25 year old shelterbelt in the South Island high country of New Zealand. A comprehensive field survey was undertaken to provide ground truthing for remote detection and detailed information on structure and composition of the infestation representing a “first order” dispersal of trees, spreading from a known source (no secondary spread from newly established trees has yet occurred in the infestation). Multi-scale remote sensing data (airborne LiDAR sensor, ALS; and spectral imagery) were acquired from an unmanned aerial vehicle (UAV) and a fixed wing manned aircraft. Data fusion methods were developed to enhance the detection of invasive conifers and analysis techniques were deployed which enabled us to define the detection threshold for various sensors.
Our ground survey expanded across more than 800 ha. More than 17,000 trees were measured (Height range = 1 cm to 476 cm) and georeferenced providing spatially accurate information on abundance, maturity (coning) and size distributions. The full census allowed us to test the detection ability of our remote sensing approaches and we found that data from both platforms and using both logistic regression and random forests for classification provided highly accurate (kappa < 0:996) detection of invasive conifers. Our study showed that the data from both UAV and manned aircraft was useful for detecting trees as small as 1 m in height and enabled us to detect nearly every coning individual in the surveyed area. It also allowed us to describe the spread pattern accurately and mapping the densities across the topography of the study site. Both will enable us to test various hypotheses on wind dispersal patterns.
The results of the remote sensing work allowed us to remotely map, with a given accuracy, the full extent of the infestation as the basis for future control. Furthermore, the results provide confidence on the usability and cost efficiency of remote sensing approaches to detect invasive tree species early enough for effective control and management in New Zealand’s High country.