OpenGeoHub EO-council Summer School 2025

Training points (ground observations and measurements of environmental variables) that are referenced in space and time, and are available for longer periods, can be used to build spatiotemporal Machine Learning models (stmlm). Such stmlm's can then be used to generate time-series of predictions, which can then be used to run time-series analysis. Spatiotemporal modeling is different from purely spatial mapping is in the following three aspects: (1) points and covariate layers are matched in spacetime (usually a day or month-year period of ground observations or at least the year of ground observations); (2) covariate layers are based on time-series of usually EO-based images (spatiotemporal data cubes) and include also accumulative indices (e.g. cumulative rainfall, cumulative snow cover, cumulative cropping fraction, and similar) and derivatives; (3) during model training and validation, points are subset in both spacetime to avoid overfitting and bias in predictions. This talk will address 15 practical lessons from running stml including how to organize monitoring networks, how to prevent overfitting, how to derive prediction errors in spacetime, how to use time-series of predictions to detect changes and similar.