Being able to gauge the health of a forest isn’t easy. Each one is a delicately balanced ecosystem, shaped by its location, climate, harvesting, planting, pests and disease. It’s why the people who manage the world’s forests are turning to satellites and big data.
Healthy trees are incredibly important for the entire planet. But quickly spotting when their health is starting to fade can be difficult, especially considering the vast areas of land they cover. In order to spot problems before they get out of hand, one company is taking an innovative, data-driven approach.
Using satellite imagery and other earth observation data, Rezatec has developed a process that can provides scalable, accurate and regularly updated forest landscape intelligence. Added to this, it has developed data analysis techniques and machine learning algorithms in order to produce powerful analytical insights for owners and managers of the world’s forests.
Rezatec’s approach provides a level of insight into forest assets that is not achievable or cost effective when compared to more traditional land surveying techniques.
In one of the first projects of its kind, Rezatec worked with the Department for Environment, Food & Rural Affairs (Defra) and the UK Space Agency to establish how common threats to forests can be detected and addressed using satellite imagery to mitigate the risks to valuable tree stock.
According to Defra Plant Health, “the ability to map ash and oak at remarkably high-levels of accuracy has been an outstanding achievement, which has not only raised awareness of potential future applications of Earth Observational data within Defra, but which could also potentially revolutionise Defra’s response to quarantine pests and diseases in the wider environment.”
Rezatec is also working closely with Forestry Corporation New South Wales (FCNSW) – which manages 2.2 million hectares of forest in Australia – to add to its existing forestry datasets. By using derived tree type mapping, FCNSW hopes to gain a more detailed overview of its forested land assets and to improve its decision making.
“Remotely sensed data is an important forest management tool and combining this with accurately located plot information provides an opportunity to add value to this data,” explains Mike Sutton, Manager, Forest Information and Planning at FCNSW. “Given the complexity of eucalyptus forest [in Australia] this is a difficult task.”
This is all achieved by inputting information from optical and radar satellite image data, using a variety of different ranges and mapping frequencies.
The multiple datasets obtained are then processed and analysed in order to derive data regarding the woodland structure and volume. The key parameters in understanding the value of forestry assets are the height Synthetic Aperture Radar (SAR) phase data from different viewing angles, which is processed to generate interferometric SAR (InSAR) coherence imagery.
This data is fed into a machine learning model alongside environmental variables (including ‘aspect’ and ‘slope’), species specific information (tree ‘class’ and ‘age’), and spectral indices associated with vegetation density. The parameterised model is then applied to pixelated inputs to produce a mapped distribution of top tree height.
Multiple datasets obtained are analysed in order to derive data regarding the woodland structure and volume
Volume Allometric equations that use established scientific relationships of tree proportions are then incorporated with the species distribution, age of trees, and their modelled height to calculate the estimated timber volume (kg/m2) per pixel.
These results are aggregated to understand the full value (in stored matter) obtainable from the woodland. Without modern data analytics, it would be a slow and laborious process.
So that all this information can be easily digested, it is visualised as maps, with pixels representing the various aspects of the data. These can represent the mean top height of the trees per pixel area, the estimated number of trees per pixel area, or the estimated total volume (kg/m2) in each 10m pixel, representing the estimated volume of timber quantity present within each of the pixel areas.
“FCNSW is watching, with interest, the modelling approach being trialled by Rezatec, and has provided plot data to assist with the model development phase, which is an important first step in getting cost effective, current and consistent information about the forest estate,” Mike Sutton adds.
This space-based, data-driven approach could be used in other areas too, such as pipeline leakage detection, peatland management, crop mapping, agricultural diffuse pollution risk, crop yield analysis, harvest optimisation and land use monitoring.
You wouldn’t expect to be able to save something on earth from outer space. Yet the combination of satellite imagery and big data analysis is now helping us safeguard the forests that are so important to the health and wellbeing of our planet.