1°30′0″ W 1°15′0″ W0°15′0″ E 1°0′0″ W 0°45′0″ W 0°30′0″ W 0°15′0″ W 0°0′0″ N 05 10 20 km 1°30′0″ W Land Use Pasture With Trees Citrus Olive Grove Pasture With Shrubs (a) 1°30′0″ W 1°15′0″ W 1°0′0″ W 0°45′0″ W 0°30′0″ W0°15′0″ W 0°0′0″ 0°15′0″ E Probability (%) 100 Fruit Trees Arable Land Pasture Land Dried Fruit Vineyard Forest 1°15′0″ W 1°0′0″ W 0°45′0″ W 0°30′0″ W0°15′0″ W 0°0′0″ 50 N 0 510 20 km 1°30′0″ W0°30′0″ W 0°15′0″ W 0°0′0″ (b) 1°15′0″ W 1°0′0″ W 0°45′0″ W FIGURE 4. An example of spatiotemporal DL for crop monitoring from high-resolution images. A bi-LSTM deep network model was used to exploit the spatial regularities and temporal dimension of a sequence of Sentinel-2 time series to predict the land use over (a) a number of classes of interest and (b) the pixel probability map. (Adapted from [121]; used with permission.) 180 IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE JUNE 2022 39°0′0″ N 39°15′0″ N 39°30′0″ N 39°0′0″ N 39°15′0″ N 39°30′0″ N 38°45′0″ N 39°0′0″ N 39°15′0″ N39°30′0″ N 38°45′0″ N 39°0′0″ N 39°15′0″ N 39°30′0″ N