Pontus Olofsson co-authored a paper titled "A global land cover training dataset from 1984 to 2020" that presents a global dataset for training machine learning algorithms in remote sensing applications. The dataset contains nearly two million training units spanning the period from 1984 to 2020 for seven primary and nine secondary land cover classes. Training data were collected across global ecoregions by sampling the spectral-temporal feature space from Landsat imagery. In addition, training data were added from publicly available and collaborator-provided datasets.
The training database is relevant for a wide array of studies such as land cover change, agriculture, forestry, hydrology, urban development, among many others. The effort was funded by a NASA MEaSUREs grant to Boston University. The open-access paper is published in Nature Scientific Data and available for download here: https://www.nature.com/articles/s41597-023-02798-5.
