The SOPPHY vLab consists of several workflow components for processing high resolution and multi-spectral satellite imagery (see Figure 1). Users can perform state-of-the-art pre-processing of satellite remote sensing imagery to produce atmospherically corrected reflectance maps. Together with model emulators and neural networks, created from radiative transfer models using machine learning, users can retrieve plant traits (leaf area index, foliar chlorophyll and nitrogen) on a pixel basis. By ingesting species cover maps acquired by an AI spectral classification on high resolution satellite observations, users can disaggregate these into species traits maps.
Users can select various software tools within the different workflow components, depending on their requirements.
- Satellite remote sensing pre-processing – Standardised Atmospheric correction (SIAC)
- Machine Learning – training model emulators and neural networks from 1D (PROSAIL) and 3D (DART) radiative transfer models
- AI spectral classification. Geographical mapping of plant species (through DeepForest & Ailanthus software tools)
- Retrieval of plant traits. Inversion of radiative transfer models to retrieve leaf area index, foliar chlorophyll and nitrogen as species traits (through the MULTIPLY framework) from multiple satellite observations.