By Sam Thiele and Rupsa Chakraborty
Google Earth revolutionised the Earth Sciences, geology especially. For the first time, aerial imagery for the whole earth could be interactively viewed across a huge range of scales - from continents, to mountain belts, to individual volcanoes, faults and folds. For many of us this gave a whole new appreciation for the scale and complexity of these geological building blocks. Changes over time also became visually evident - many relating to humanity’s increasing impact on the planet.
But, Google Earth imagery is limited to red, green and blue channel colour images, matching how our eyes perceive the world, but very limited when it comes to identifying rock or vegetation types. Most plants are green, most rocks are grey-brown. Boring.
Several recent satellite launches could revolutionize the geosciences yet again! These hyperspectral satellites can see in many hundreds of colours, most outside the visible range that we perceive. This extra capability allows the identification of individual plant species, rocks and minerals, based on their distinctive ‘spectral fingerprints’.
These satellites have not yet achieved global coverage - but eventually they will. PRISMA, an Italian satellite launched in 2019 has now mapped ~200 million square kilometers. EnMap, a German project, was launched last year and has been a game changer already. These are the pioneers, but many more powerful hyperspectral missions are planned in the near future that will increase this database by leaps and bounds (e.g., the Surface Biology and Geology mission from NASA).
This wealth of data presents a huge opportunity, by providing the information needed to build comprehensive earth system models, monitor changes in the bio- and geo-spheres, and find valuable new deposits of critical raw materials. But the huge volume of data being generated, and its high dimensionality, also make this a daunting challenge. Now we need a hyperspectral Google Earth!
A hyperspectral satellite following a sun-synchronous orbit consistently acquires data at noon, for optimum lighting conditions and data quality. The resulting data cube can be analysed using statistical and machine learning techniques to, for example, map changes in mineralogy or vegetation.
Comentarios