*2.1. Study Area*

A classification experiment was conducted in the crop cultivation areas of Kansas State, USA, in 2015 (Figure 1). The reason for the choice of the study area was two-fold: Kansas is known as one of the main production areas of winter wheat, in addition to various crops such as corn, sorghum, and soybean [27]. Thus, it was possible to examine how well the self-learning approach of this study could discriminate between complex land-cover types. The second reason was the availability of past time-series land-cover maps. The CDLs, provided by the National Agricultural Statistics Service (NASS) of the United States Department of Agriculture (USDA) [28], were used to both extract the cultivation rules of cropping systems in the study area, and validate the classification results.

**Figure 1.** Location of the study area and data sets: (**a**) Moderate Resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) on 9 May 2015, at a 250 m spatial resolution and (**b**) cropland data layer (CDL) 2015 at a 30 m spatial resolution.
