Gridded 5 arcmin datasets for simultaneously farm-size-specific and crop-specific harvested areas in 56 countries

Farms are not homogeneous. Smaller farms generally have different planted crops, yields, agricultural inputs, and irrigation applications compared to larger farms. However, gridded farm-size-specific data that are moreover crop specific, are currently lacking. This obscures our understanding of differences between smallscale and large-scale farms, e.g., with respect to climate change adaptation and mitigation strategies, contribution to (local) food security, and water consumption patterns. This study fills a significant part of the current data gap, by developing high-resolution gridded, simultaneously farm-size-specific and crop-specific datasets of harvested areas for 56 countries (i.e., covering about half the global cropland). Hereto, we downscaled the most complete global direct measurements of farm size and crop type by compiling state of the art datasets, including crop maps, cropland extent maps, and dominant field size distribution, representative for the year 2010. Using two different crop map sources, we were able to produce two new 5 arcmin gridded datasets on simultaneously derived farm-size-specific and crop-specific harvested areas: one for 11 farm sizes, 27 crops, and 2 farming systems, and one for 11 farm sizes, 42 crops, and 4 farming systems. In line with previous findings, our resulting datasets show major differences in planted crops and irrigated area (%) between farm sizes. Consistency between our resulting datasets and (i) observations from satellite images on farm-size-specific oil palm, (ii) household surveys on the farm-size-specific irrigated area (%), and (iii) previous studies that mapped noncrop-specific farm sizes and support the validity of our datasets. Although at grid level some uncertainties remain to be overcome, particularly those stemming from uncertainties in crop maps, results at country level seem robust. Source data, code, and resulting datasets are open access and freely available at https://doi.org/10.5281/zenodo.6976249 (Su et al., 2022).