G2G: from galaxies to the ground

Research into galaxy formation and evolution involves attempting to reconstruct the detailed properties of extremely distant galaxies using a very small amount of data. For example, even in the best imaging possible with the Hubble Space Telescope, the most distant galaxies appear as faint smudges, comprised of just a handful of pixels, scarcely brighter than the random noise in the image. Yet with knowledge of the basic constituents of galaxies, we can accurately model their properties: the total mass in stars, the rate at which they are forming new stars, their ages, and so-on. These are vital for our understanding of how galaxies formed and evolved in the Universe.

One of the key techniques is the use of stellar population synthesis models that predict the amount of light emitted by individual stars of different types in galaxies. Even though we only observe the integrated starlight from distant galaxies, and cannot resolve the individual stars, we can derive likelihood distributions for the key underlying physical properties that give rise to those few bright pixels. Effectively we are measuring 'sub-grid' properties, often pushing the data to its limit and extracting useful information even from below the noise.

The similarities between the problem of modelling distant galaxies and that of modelling the properties of a small portion of land are striking: in both cases we are trying to understand a complex system (a disc of stars or a field of maize, for example) from a very limited set of data. A small, contaminating patch of diseased or unhealthy crop, mixed in with healthy crop, could be blended below the resolution limit of the data. Nevertheless, this mixture is expected to leave a tell-tale spectral signature in the imaging. This is akin to detecting the presence of dust in distant galaxies, that subtly alters the colour of the starlight. In the case of earth observation, continuous monitoring satellites relevant to crop surveillance have a spatial resolution of order 10-metres, such that a small field of crops could be represented by just a few pixels in an image.

G2G will develop and apply astrophysical techniques to continuous-monitoring earth observation satellite imaging, with a specific focus on extracting useful information from images down to the pixel scale. The aim will be to provide spatially pin-pointed information on the health of crops growing within small (10s of metres) parcels of land, measured simultaneously over swaths of hundreds of kilometres. The goal will be to rapidly identify subtle - otherwise undetected - signatures of crop stress or yield threats, with information delivered efficiently to large numbers of farmers and land users, even those with very small agricultural footprints.

This proposal aims to translate astrophysical research (specifically techniques in observational cosmology) to precision agriculture, by applying new data analysis techniques to satellite earth observation data. The goal is to contribute to precision agriculture through the efficient pin-pointing of disease or crop stress, and to improve general monitoring of agricultural land usage for farmers. This project is impactful because it aims to deliver on the UK Government's Industrial Strategy (IS). One of the IS 'Grand Challenges' is to put 'the UK at the forefront of the AI and data revolution', with a goal that data science be adopted into agricultural technology explicitly highlighted in the White Paper. Furthermore, the project also aligns with the 'Clean Growth' theme of the IS (e.g. developing strategies to reduce water usage in agriculture) through the development of precision and high efficiency agriculture - exactly the aims of G2G. Although this project is focused on UK agriculture, there is no reason why these techniques could not be applied elsewhere, particularly in the developing world, where efficient identification of crop disease is acutely critical. We envision future developments aligned with investment in the Global Challenges Research Fund, where we see scope within the themes of 'Equitable Access to Sustainable Development' and 'Sustainable Economies and Societies'.