The dataset is organised as a multimodal collection around LUCAS survey points. For each point, the ground photographic bundle is linked to LUCAS tabular metadata and to co-located Earth-observation time series derived from Sentinel-2 and Sentinel-1. Sentinel-2 provides optical multispectral observations, useful for vegetation state, phenology, crop development, bare soil, water, and built-up surfaces; Sentinel-1 provides C-band SAR observations, useful because it is sensitive to structure and moisture and can acquire data under cloud cover and at night. Sentinel-2 carries 13 spectral bands at 10, 20, and 60 m spatial resolution, while Sentinel-1 provides all-weather radar imagery. For each LUCAS point, the dataset therefore contains not only the field photographs, but also temporally ordered satellite observations around the survey year, for example Sentinel-2 reflectance time series and Sentinel-1 backscatter time series extracted over a spatial neighbourhood of the point. These time series provide the seasonal and phenological context that is not visible from a single ground photograph. The accompanying metadata include the LUCAS point identifier, survey year, geographic coordinates, country or administrative information where available, land-cover and land-use labels, photo-view identifiers, and links between the image files and the corresponding LUCAS observation. For the satellite component, metadata should also describe the sensor, acquisition dates, spatial extraction window, coordinate reference system, preprocessing level, cloud or quality masks for Sentinel-2, and relevant SAR acquisition/preprocessing information for Sentinel-1. It combines in-situ field evidence, ground-level visual context, satellite optical and radar time series, and structured metadata for multimodal learning, land-cover classification, validation, cross-modal retrieval, and analysis of complementarity between ground and satellite observations.