The first photograph of earth from space was taken more than 70 years ago, by a camera placed in the nose of a captured V-2 rocket. Today, about 300 of the more than 1,000 satellites in orbit are there to observe our Earth and investment in space programs is changing. Most earth observation (EO) missions are commercially owned and operated with the majority launched in the last 18 months.
Countries have been running global, civil EO missions for decades generating publicly available data and images. Established missions, like NASA’s Landsat, have been joined more recently by the European Space Agency’s Copernicus Programme, with its Sentinel satellites and contributing missions.
The initial objective of these systems was to assist in the scientific understanding of the earth and the management of and response to its powerful forces. Today, because of this, we have a much better understanding of weather patterns, natural disasters and climate change and it has allowed society to respond to these challenges in a quicker and more coordinated manner than had been possible before.
And increased frequency of recording the data associated with these global events has reduced the guesswork involved in predicting, anticipating and judging the likely impact of storm, earthquake or drought.
These publicly funded initiatives have also resulted in other benefits like the development of commercial spin-offs, such as Glasgow-based Global Surface Intelligence (GSI) who uses its next generation Artificial Intelligence (AI) platform to look inside the world’s biological assets, facilitating deep inspection into areas such as forest inventory, landmass change detection and agriculture, in ways and to depths and scales previously unimaginable.
Time-series analysis of historical record made by satellite systems can be used in combination with contemporaneous data to train AI engines. By processing and analysing data from satellites, using both optical and radar data and combined with input from field-based sensors, such as drones and LiDAR, environmental, geological and meteorological data, GSI can collate the richest and most valuable data sources available anywhere.
The introduction of the Copernicus Programme has provided increased resolution of both optical and radar-based data which, when enhanced and processed by the AI in GSI’s platform, can provide answers with a high degree of accuracy about the current inventory and performance of land assets, from forestry to rice, from grass to wheat crops.
“Our AI system is trained to recognise attributes such as plant & tree species, height and volume, as well as the extent of any disease or damage to the entire stock of a forest. Using machine learning to identify similarities in other plots we can create highly accurate predictions across millions of hectares of land within days and present final results within a few weeks."
"This is genuinely a revolution in land use decision making.”
comments Gavin Tweedie, CEO of GSI.
LiDAR and ground survey methods can take months or even years to complete due to weather dependency, according to Tweedie, whereas GSI can process satellite data anytime and continuously refresh it, providing ongoing change detection.
But Tweedie considers GSI’s analytics as complementary to existing methods of data collection, rather than alternative or competitive.
“Some 70% of the earth’s surface is obscured by cloud on any given day. Obtaining usable optical images in parts of the world with constant cloud cover makes continuous monitoring a challenge. SAR (Radar) data, while excellent at seeing through cloud, can be coarse and can contain contamination, noise and distortion making it difficult to use in sophisticated earth observation analysis,” says Tweedie
The GSI platform automatically captures the latest satellite data and merges it with previous ones to effectively remove cloud cover and other contaminants and leave a new image ready for deep inspection. GSI combines this with other sources of information such as LiDAR and ground surveys providing the richest and most valuable results for clients.
GSI’s process, using one of the world’s most powerful computers, applies machine learning to a configuration of pixels on a single image, allowing it to make predictions about likely patterns of behaviour in the surrounding area with a high degree of accuracy.
The method is particularly effective in monitoring and measuring agriculture and forest assets, land- use change, and land-cover change.
Increasingly, other groups with an interest in trees also want to measure how the environment, our forests, are encroaching on critical infrastructure. Knowing there are tall or dead trees near power lines or wind turbines and being alerted to this reduces the risk to capital infrastructure.
“If you look at a satellite image of an area of forestry, the human eye cannot distinguish the species, height or other forest attributes,”
says Tweedie. “LiDAR can provide height and slope data and, if GSI trains its AI platform on that image, we can be very specific about the assets contained in it across a range of factors, such as what species the trees are, how tall they are, whether they have been damaged by pests or disease.”
GSI is currently advising a land-based wind energy company, about the likely impact of trees on wind turbine performance. “Trees over 10m tall, more than a kilometre away have potential to disrupt airflow which can cause reduced energy production,” says Tweedie.
“Renewable energy companies need a reliable and inexpensive way of monitoring forest growth in areas where they have wind farms and we are helping to provide that.”
Comments