Figshare

Part of the dataset

CIMMYT Ethiopia

Our study region is located in the northwest of Mexico, in the Yaqui Valley, where most farmers predominantly grow crops under irrigated conditions during the winter months. Sowing typically starts in late October. Dry bean is one of the first crops to be sown (Table 1). Wheat, the dominant crop, is usually sown between mid-November and mid-December, however, some fields are sown as late as early January. Among the other eight crops that will be referred to as minority crops in this study, maize and chickpea were the most important ones. The last crop to be sown during the winter months is safflower. It is typically sown in March or April, after field pea or fallow. Sen2-Agri allows for the identification of only one crop per field and season, we therefore did not include it in the study. The Yaqui Valley also is an important producer of various types of vegetables. Their production is quite dynamic. The growth cycle of vegetables tends to be quite short and often, they do not have a distinct seasonality. Broccoli and different types of tomatoes are the most important ones. We also included some permanent crops such as asparagus, alfalfa and pasture (grassland), as well as tree fruit and nuts, categorized as orchard. Alfalfa and pasture were categorized as forage crop. The planners of the Yaqui Valley irrigation scheme had divided the land into blocks, measuring 2 by 2 km. The blocks were then further subdivided into 40 lots, each measuring 10 ha. The blocks and lots were numbered consecutively. At the beginning of the winter growing season, the irrigation district, called Distrito del Riego del Rio Yaqui, requires each farmer to declare the type of crop they plan to grow on each irrigated lot. The irrigation district kindly shared those data with us. Most farmers do not follow the initial lot boundaries anymore. Some lots got split up, whereas in the majority of cases, lots were merged. If farmers had merged several lots, they would use the number of their first lot as an anchor and also report the area of the entire field, i.e., the merged lots, that was planted with the same crop. This then allowed us to visually match the reported data with the crop fields. Based on the farmer's declarations, which include the crop type, block, lot and field size, the crop types were then assigned to the field boundaries which had been manually drawn beforehand, using a Sentinel-2 image from March 13, 2017 as a background. This resulted in 6048 labeled fields. The average area of a field was 11.5 ha.