基于多源遥感影像的作物信息提取与需水量时空分析

Crop information extraction and spatial-temporal analysis of water demand based on multi-source remote sensing images

在我国水资源短缺的大背景下,合理规划农业水土资源尤为重要。为实现区域作物种植结构提取,分析作物需水量的时空分布规律,以山东省济宁市重要农业区老万福河流域为研究区,以Landsat 8影像为基础构建作物生育期NDVI时间序列,综合考虑GF-2影像与实地调研数据,结合作物物候期,采用决策树方法提取老万福河流域作物种植结构,并结合Penman-Monteith公式与GIS空间分析功能,使用单作物系数法计算流域作物需水量。结果表明:(1)多时相NDVI序列下,采用决策树方法提取小麦-水稻的平均误差最小,为13.5%,蔬菜的平均误差最大,为23.0%。分类结果显示:小麦-水稻轮作模式分布在流域东部,小麦-玉米轮作模式主要分布在流域西部,经济作物蔬菜主要分布在流域中、西部。(2)作物需水量年内时空分布存在显著差异。时间尺度上,冬季作物需水量明显小于其他季节,夏季作物需水量最大,春夏作物需水量占全年作物需水量的70%左右;空间尺度上,流域东部水稻种植区的作物需水量明显大于其他地区。同一种植模式在不同种植区域的需水量有所差异,其中小麦-玉米年需水量范围在915.6~952.9 mm,不同区域需水量差异达37.3 mm。(3)主要作物水稻、玉米、蔬菜、小麦的生育期需水量存在显著差异,作物需水量分别为846.87、561.7、519.37、415.24 mm。在拔节孕穗(抽薹)期,作物的需水量达到峰值。研究表明,基于Landsat 8和GF-2影像数据构建NDVI时间序列并结合决策树分类的方法可以实现区域作物种植结构的提取。研究结果可为当地灌溉制度的制定及农业水土资源的优化管理和分配提供科学依据。

Under the context of water shortage in China,it is important to rationally plan agricultural soil and water resources.To realize the extraction of regional crop planting structure and analyze the spatial-temporal variations of crop water demand,this study took the Old Wanfu River Basin,an important agricultural area in Jining City,Shandong Province,as the study area.NDVI time series of crop growth period were constructed based on Landsat-8 image,combined with GF-2 image,field survey data,and crop phenology.The decision tree method was used to extract crop planting structure.Furthermore,the single crop coefficient method was used to calculate crop water demand,combined with the Penman-Monteith formula and GIS spatial analysis function.Results showed that the average error of wheat-rice extraction using decision tree method was the smallest(13.5%),and the average error of vegetable extraction was the largest(23.0%).The wheat-rice rotation pattern was distributed in the eastern part of the basin,wheat-corn rotation pattern mainly distributed in the western part of the basin,and vegetables mainly distributed in the middle and west of the basin.There were significant differences in the annual spatial and temporal distribution of crop water demands.On the temporal scale,water demand of winter crops was significantly lower than that of other seasons,water demand of summer crops was the largest,and spring and summer crops accounted for about 70% of the annual crop water demand.On the spatial scale,crop water demand in the eastern rice-growing area was significantly higher than that in other areas.The water demand of the same planting pattern was different in different planting areas.The annual water demand of wheat-corn ranged from 915.6 mm to 952.9 mm,and the difference between different areas was 37.3 mm.There were significant differences in water demands of rice,corn,vegetables,and wheat during the growth period,being 846.87,561.7,519.37,and 415.24 mm,respectively.Furthermore,crop water demand reached its peak at the jointing booting(bolting)stage.The results indicated the method by combining NDVI time series,generated with on Landsat-8 and GF-2 image data,and decision tree classification could realize the extraction of regional crop planting structure.This study could provide a scientific basis for the formulation of local irrigation system and the optimization management and allocation of agricultural soil and water resources.