9-2. Key concepts

Context

In the tutorials so far, we've largely looked at the outputs of Earth Observation programmes - classifications of land use, such as World Cover, or other observations of the earth, such as Above Ground Biomass, Soil Water Indexes (tutorial 6), Wind Power Density (tutorial 7) and so on. In this exercise we look more at the actual observation data that is used to produce these.

Multispectral and Hyperspectral data

Multispectral data is imagery captured across several distinct wavelength bands of the electromagnetic spectrum, including visible and non-visible light, enabling the identification and analysis of features that may not be distinguishable in ordinary photographs. Sentinel-2 and Landsat data are examples from ESA and NASA/USGS missions respectively.

Hyperspectral data extends this concept by capturing information across hundreds of narrow, closely spaced wavelength bands, allowing more detailed identification of materials and their properties based on their unique spectral signatures. EnMAP (Germany) and PRISMA (Italy) are examples of satellite missions providing hyperspectral imagery.

There are also commercial suppliers of both multispectral and hyperspectral data.

In these exercises we will work with some multi-spectral data but the same principles apply to hyperspectral data.

Multispectral / hyperspectral data in COGs

Whilst multispectral data is often supplied in Cloud Optimised GeoTIFFs, the structure of these can vary. On some occasions you may encounter a single COG with different bands of the COG representing the different wavelengths. On other occasions you may find each wavelength is supplied as its own separate COG.