Training and testing data within the context of the "Handful of Pixels" course, and in particular the Land-Use and Land-Cover mapping chapter. The data included samples 300 random locations across 10 land cover classes from the data described in Fritz et al. 2017, and downloaded through the AppEEARS API using the {appeears} R package (Hufkens et al. 2023). A 80% split is executed on this larger dataset with 240 locations retained for training, while the remainder is used for testing purposes. For the testing data the input data is shared, the labels are withheld (stored in a closed release of this archive, and accessible on reasonable request). This data can be used within the context of small demonstration machine learning exercises or competitions. Data structure The data contains all seven (7) bands of the MODIS MCD43A4 data product for the year 2012. Band names are indicated in full. In addition MODIS MOD11A2 daytime land surface temperature (LST) data is provided, where band names only contain the date (YYYY-MM-DD) of acquisition. Additional indices can be calculated from these band combinations if so desired. Data is provided in compressed serialized R rds files, and can be read into R as follows: df <- readRDS("training_data.rds") References Fritz, Steffen, Linda See, Christoph Perger, Ian McCallum, Christian Schill, Dmitry Schepaschenko, Martina Duerauer, et al. “A Global Dataset of Crowdsourced Land Cover and Land Use Reference Data.” Scientific Data 4, no. 1 (June 13, 2017): 170075. https://doi.org/10.1038/sdata.2017.75. Koen Hufkens. (2023). bluegreen-labs/appeears: appeears: an interface to the NASA AppEEARS API (v1.0). Zenodo. https://doi.org/10.5281/zenodo.7958270