In the summer of 2017, the Ministry of Forestries at the Government of Saskatchewan, challenged GSI to deliver a trial inventory in an area of provincial forest, covering some 67,000 hectares, in its attempt to increase accuracy, efficiency and lower costs.
The Glasgow-based geospatial AI specialist was pitted against six other companies providing a variety of conventional and modern, forest inventory methods.
Forestry is the largest industry in the area represented Saskatchewan, covering some 34 million hectares. The commercial forest zone makes up around a third of that, of which 5.3 million hectares is classified as productive forest land, available for commercial timber harvesting.
With 10 major manufacturers producing lumber, pulp and panels, and over 170 small forest businesses producing a variety of products, the success of forestry is essential to the local economy.
The Saskatchewan Government’s challenge was to evaluate each system across a range of criteria and to establish which, if any, provided a compelling case to assert that theirs was a clear winner in terms of speed, accuracy and cost.
Officials wanted to be open-minded, particularly about the potential of new technologies and to separate the technology from the vendor and to this end, the trial was done on a blind basis, with those reviewing the results unaware of the identity of the companies involved, nor their methods of measurement.
They wanted to be able to use the findings, potentially to advise industry stakeholders of inventory assessment; to help guide the development of new inventory standards; and to provide a proving ground for promising methods.
GSI emerged as one of the most accurate & the quickest of the six companies taking part in the trial – including providers of satellite analytics, ground surveys and LiDAR.
A Forestry Service spokesman said:
“GSI was the most robust and fastest turnaround of all methods tested.”
Officials had previously used standard inventory procedures to calculate softwood stocks in terms of species and volume distribution. These procedures require frequent and expensive boat, plane and helicopter site visits which financially constrained the government’s ability to conduct regular and timely analyses. Furthermore, the extended time required to analyse the data collected resulted in the data regularly being out of date by the time it was available for decision making purposes.
With the forest industry generating around $1billion in annual revenues and supporting 8,000 direct and indirect jobs, forests represent very valuable investment commodities and growers, insurers, banks and pension funds need accurate, real-time information about the value of their timber stocks to help them make properly informed commercial decisions one, five or ten years down the line.
The Saskatchewan Government needed quickly to determine the volume and value of their forest inventory and within four weeks GSI were able to provide an inventory of all their growing assets broken down by species, volume and area.
Traditional procedures for analysing forests include “timber cruising” to physically count trees as well as Light Detection and Ranging (LiDAR), a remote-sensing method which uses fixed wing aircraft to fly over an area and collect data, using laser scans.
While these methods offer accurate inventory analysis, they are time consuming and expensive requiring highly qualified and consistent interpreters, and invariably a long lead-in time is needed.
GSI complements those methods by ingesting ground sampling, LiDAR and satellite based data into their machine learning platform and applying patented algorithms to produce highly accurate, insightful and valuable decision-making information, in near real-time serving strategic and operational forest management objectives.
The challenge demonstrated that GSI scored among the highest overall for accuracy and was able to best identify the coverage of each species & where they were located.
GSI demonstrated it matched or outperformed other methods for stand by stand metrics for less than $1/hectare across millions of hectares which is estimated to be in the region of 30 times cheaper than other survey methods.
