Synergistic Modern Global 1 Km Cropland Dataset Derived from Multi-Sets of Land Cover Products

Abstract

The quality of global cropland products could affect our understanding of the impacts of cropland reclamation on global changes. With the advancement of remote sensing technology, several global land cover products and synergistic datasets have been developed in recent decades. However, there are still some disagreements among the global cropland datasets. In this paper, we proposed a new synergistic method that integrates the reliability of spatial distribution and cropland fraction on a pixel scale, and developed a modern (around 2000 C.E.) fractional cropland dataset with a 1 km × 1 km spatial resolution on the basis of the spatial consistency of cropland reclamation intensity derived from multi-sets of global land cover products. The main conclusions are shown as follows: (1) The accuracy of spatial distribution assessed by validation samples in this synergistic dataset reaches 87.6%, and the dataset also has a moderate amount of cropland pixels when compared with other products. (2) The reliability of cropland fraction on the pixel scale had been highly improved, and most cropland pixel has a higher fraction (over 90%) in this dataset. The “L” shape of the histogram of pixel numbers with different reclamation intensities is reasonable because it is consistent with the up-scaling results derived from satellite-derived products with high spatial resolutions and the expert knowledge on cultivation. (3) The cropland areas in this non-calibrated result are generally closer to that of FAOSTAT on scales from global to national when compared to other non-calibrated synergistic datasets and original satellite-derived products. (4) The reliability of the synergistic result developed by this method might be decreased to some degree in the regions with high discrepancies among the original multi-sets of cropland datasets.

Coincidence Analysis of the Cropland Distribution of Multi-Sets of Global Land Cover Products

“…However, the cropland distribution pattern indicated by the original satellite-derived products had been slightly changed in these fusion datasets due to the synergizing and quantitatively calibration process. The accuracy of some fusion fractional datasets is not as high as the publisher claimed [52].…”

“…Second, the cropland subsets were extracted from these Boolean products by selecting the cropland-related classes (cropland class and mixed-cropland class) directly. For GlobeLand30 and ESA-CCI-LC, we generated the 1 km × 1 km fishnet for the whole world land area and converted the original Boolean products to the upscaling results with 30 × 30 resolution by using the zonal statistics tool in ArcGIS instead of resampling into 1 km × 1 km directly [52]. The newly generated results are fractional type.…”

Study on the Classification and Change Detection Methods of Drylands in Arid and Semi-Arid Regions

“…Such data always lacked real-time information and accurate spatial information. Some studies aimed to synthesize various data to obtain more accurate datasets [12][13][14], which showed higher consistency in large irrigated areas and lower consistency in small irrigated fields.…”

Dataset of 1 km cropland cover from 1690 to 1999 in Scandinavia

“…Many regional land use reconstructions illustrate that global datasets have nonnegligible discrepancies in reflecting regional spatial land use patterns historically, especially for cropland. Historical document-based reconstructions conclude that the SAGE, HYDE, and PJ datasets have drawbacks in capturing the spatial distribution of historical cropland change in China (Li et al, 2010;Zhang et al, 2013;Li et al, 2016Li et al, , 2019Wei et al, 2019). In the USA, the HYDE maps substantially underestimate crop density in high-cropland-coverage regions but overestimate it in the low-density areas for 1850-2016 (Yu and Lu, 2018).…”