华北平原冬小麦面积遥感提取及时空变化研究

多年作物种植面积的时空变化可以反映种植结构的调整结果,并可用于进行驱动力分析。为解决连续遥感监测作物种植面积变化过程中,不同的训练样本或分类规则不能较好地反映作物种植面积时空格局变化的问题,本文首先根据研究区(华北平原)农业气象观测站提供的主要农作物物候观测资料获得主要农作物典型物候期特征,结合HANTS滤波后的NDVI时间序列数据提取不同地物类型的NDVI时序曲线,引入复种指数,探讨了CART算法在提取华北平原冬小麦种植面积的可行性,最后提取了2000—2013年华北平原冬小麦种植面积,并参考市级的农业统计数据进行精度评价。经检验,近13年的遥感监测种植面积与农业统计面积相关系数达到0.94(置信水平为95%),且各市13年面积一致性小于40%的概率仅为15%。利用遥感监测多年冬小麦空间分布信息获得其空间种植概率,能较好地反映研究区冬小麦的主要种植区,该方法可为大范围、连续年份冬小麦种植面积时空格局的遥感监测提供参考。

Using remote sensing to determine spatio-temporal variations in winter wheat growing area in the North China Plain

Agricultural land uses affect land surface energy and water balance. The North China Plain (NCP) is one of the most important agricultural regions in China and is experiencing a severe water shortage due to decades of excessive extraction of groundwater for crop irrigation. The precise determination of the distribution of the land areas under different crops in NCP does not only provide the basic information needed for evaluating agricultural water consumption in space but also improve agricultural planning for sustainable water resources management. In this study, we established a method for agricultural land use classification using MODIS NDVI data time-series. Raw 16-day composite NDVI data were first processed using HANTS filtering and then sampled the time-series pattern of NDVI for different croplands based on ground truth data. The typical phenological characteristics of the main crops were determined based on field-monitored phenological data in agro-meteorological stations in the study area. Then multiple cropping index (MCI) was induced based on the NDVI time-series to distinguish land areas under wheat-corn double cropping system. Finally, the cultivated area under winter wheat in NCP for the period of 2000-2013 was extracted by using the CART algorithm. The result was validated with county agricultural statistics data, which showed a statistically significant correlation for the 13-year period with correlation coefficient of 0.94 at 95% confidence level. The probability of consistency less than 40% between the remote-sensing derived cultivated area and agricultural statistics data for the study period was less than 15% for the municipalities in the study area. Agricultural land use frequently changed (about yearly) due to crop rotation, price fluctuation, water limitation, etc. This made it difficult to evaluate agricultural land use change from only comparison of agricultural land use maps for any set of years. Thus in this study, the land use change for different crops was evaluated based on multi-year cropping probability comparison, which composed of actual counts of number of years of wheat cultivation in the study period, with high cultivation probability reflecting dominant crop distribution. Through comparison of the differences in wheat cultivated probability maps for the periods of 2000-2006 and 2007-2013, it was noted that wheat cultivated areas significantly increased in Henan and Shandong Provinces, but significantly decreased in the north part and Heilonggang region of Hebei Plain. These changes were mainly considered to be driven by groundwater conditions and national policies for increased grain production, leading to extensive land reclamation in Henan and Shandong.