A 30m African Cropland Layer for 2016 by Integrating Multiple Remote sensing, Crowdsource, and Auxiliary Datasets

In order to produce the most accurate cropland layer at 30 m spatial resolution for Africa, the cropland layers extracted from four remote sensing land cover datasets were integrated. The four datasets covered the period 2015 to 2017. Hence, the constructed cropland layer was produced for the nominal year 2016. To build the final layer, the cropland mapping accuracies of the four cropland layers were firstly investigated at the units of Agro-ecological zones. Then, the best cropland layers for all zones were spatially joined. The resulted cropland layer is a binary mask with higher overall accuracy than individual layers and more consistent with FAO official statistics.

为制备空间分辨率为30米的高精度非洲耕地图层，本研究整合了从4套遥感土地覆盖数据集提取的耕地图层。上述4套数据集的覆盖时段为2015年至2017年，因此最终构建的耕地图层对应基准年2016年。为生成最终图层，研究首先以农业生态区（Agro-ecological zones）为单元，评估了4套耕地图层的耕地制图精度；随后将各分区的最优耕地图层进行空间拼接。所得耕地图层为二值掩膜（binary mask），其整体精度高于单套数据集图层，且与联合国粮食及农业组织（Food and Agriculture Organization，FAO）官方统计数据更为契合。