全球变化知识资源中心

Accurate observations on multiple cropping practices are required to better understand the status and potential of cropland use intensity. However, previous studies largely relied on multiple cropping index (MCI), which only measures the average state for an administrative unit. In this study we use various satellite images to observe the multiple cropping frequency (MCF), in order to know how MCF improves the observation of multiple cropping practices as an alternative of MCI, and how temporal and spatial resolution affect the observation of MCF. We apply the NDVI time-series curve to observe cropland phenology, and subsequently to estimate the MCF by counting the number of peaks. Three MCF maps are developed for the experimental region (Jinxian County, Poyang Lake Plain, South China) in the year 2015 based on MODIS, GF (GF-1/WFV) and GF-MODIS fusion, which represent a character of higher temporal resolution, higher spatial resolution, and higher temporal-spatial resolution, respectively. All these maps have detected multiple cropping patterns, including various single-, double-and triple-cropping pixels: 90.38%, 9.54%, and 0.08% from the MODIS map, 70.32%, 29.27%, and 0.41% from the GF map, and 64.85%, 33.62%, and 1.53% from the GF-MODIS map. The confusion matrix containing 161 field samples shows that the overall accuracy and Kappa coefficient are 62.11% and 0.34, 78.88% and 0.65, and 90.06% and 0.84, for the MODIS, GF, and GF-MODIS maps, respectively. Moreover, the statistics show that the county-level MCI is 1.42, while the aggregated MCI for these maps are 1.10, 1.30, and 1.37, respectively. Our study indicates that the GF-MODIS map not only has highest accuracy but also has a closest estimation on MCI. It implies that higher spatial resolution is the first necessity for mapping the MCF in the landscape fragmented region. Higher temporal resolution is also important to distinguish the nuances on MCF induced by crop rotation.