A new interactive tool providing the forestry industry with powerful inventory information to make management, harvesting and wood processing decisions easier has been unveiled by Scion in collaboration with Indufor Asia Pacific.
Scientists from the Crown Research Institute outlined Forest Insights’ capabilities and applications in Rotorua, at ForestTECH 2023.
The interactive Forest Insights tool provides forest owners, managers and wood processors with an overview of the changing availability and growth of planted radiata pine over time.
The prototype is currently focused on modelling of east coast pine forests, but Scion has plans to provide the same data and for a wider range of trees for other regions, with Bay of Plenty the next cab off the rank.
Scion portfolio leader for new value from digital forests and the wood sector, Grant Evans, says the prototype will support forestry and wood processing companies to make more informed management decisions.
“Long term, it will help anyone with trees planted know their precise location and ultimately, what they can do with the trees in the future.”
Forest Insights has been built using AI technologies as well as LiDAR, to detect and identify stands of trees to quantify their volume and maturity over time. It outlines the boundaries for each stand of trees and provides essential details, such as age class, area in production, and the number of stems per hectare.
Forest Insights also tracks the history of planting and harvesting, which provides insights into changing inventory levels.
Automatically detecting commercial radiata pine forests using trained Deep Learning Convolutional Neural Networks by their boundaries is a game-changer for forestry companies. What used to be a laborious task of drawing polygons is now replaced with the click of a button.
Evans says Forest Insights also levels the playing field for smaller forest owners.
“These individuals, who own smaller woodlots or stands, can use the tool to see where other small lots in their region are maturing at a similar time and potentially co-operate to negotiate better pricing from mills.”
Scion is already working with the University of Canterbury to identify tree species beyond radiata pine, aligning with the government’s goal of having 20% non-radiata pine forests by 2030.
Currently, such measurements rely on people voluntarily reporting their data, making it difficult to track progress. Forest Insights intends to change that by using satellite imagery and LiDAR data from Toitū Te Whenua Land Information New Zealand to detect different tree species accurately and utilise Indufor’s dashboarding expertise.
Beyond tree species identification, the prototype can automatically track forest activities. This functionality not only helps in tracking inventory but it is hoped, with further training, that it will provide a means to assess forest damage following natural disasters.
“For the east coast, it could also be used as a tool to reveal where planted forests are being abandoned or are no longer being harvested due to concerns relating to planting on erosion-prone land,” Evans says.
“For forestry and wood processing companies, this data offers them a holistic view of their assets and a basis for well-informed decisions.”
The granularity of the information available from Forest Insights offers huge benefits to industry and investors alike, says Dr Pete Watt from Indufor’s Resource Monitoring Team.
“Such information provides the cornerstone for developing wood availability forecasts that underpin investment decisions and support infrastructure planning and policy settings.”
The development of Forest Insights started in 2022 and is a collaborative effort. Scion’s data scientists have supplied all the models and data, working with Indufor Asia Pacific Ltd to create the online tool and dashboard interactivity.
Claire Stewart, programme manager for the FGR-led Precision Silviculture Programme, says inventory management tools are critical to being able to see our national forest scape more holistically and to support foresters manage at a finer grain level of detail.
“There are simple tasks that machine learning models can assist us with like boundary mapping, cut over mapping and post-plant survival assessment,” she says.
“By having a platform that pools data to create robust models we can move a lot quicker.”