CAJ | 학술논문

遥感影像可以同时获取地物波谱及空间位置信息，为作物长势的空间变异研究提供新的技术手段，该文利用高空间分辨率遥感影像开展了冬小麦地块内长势空间异质性特征的提取及分析。研究基于小麦挑旗期QuickBird遥感影像，选取不同长势冬小麦地块，计算地块内冬小麦归一化植被指数（normalized difference vegetation index，NDVI）的经验半方差函数，采用最小二乘算法进行模型拟合，得到冬小麦地块 NDVI 最优半方差模型及参数（基台值、变程、块金值）。结果表明，不同冬小麦地块NDVI经验半方差图呈现明显的有基台模式，冬小麦NDVI 表现出明显的带状异向性特征。在垂直及平行垄向上，基台值与地块内 NDVI 纹理值域范围、纹理方差均极显著正相关（P<0.01），且受方向影响不大；变程在垂直垄向上与地块内部作物 NDVI 均值呈极显著负相关（P<0.01），与变异系数及NDVI小于0.40的像元覆盖度极显著正相关（P<0.01）；块金值垂直垄向上与作物NDVI均值、变异系数、小于0.40的像元覆盖度有极显著关系（P<0.01），变程、块金值在平行垄向与各个因子无相关性。该研究为利用遥感影像揭示作物长势的空间变异提供了参考。
요감영상가이동시획취지물파보급공간위치신식，위작물장세적공간변이연구제공신적기술수단，해문이용고공간분변솔요감영상개전료동소맥지괴내장세공간이질성특정적제취급분석。연구기우소맥도기기QuickBird요감영상，선취불동장세동소맥지괴，계산지괴내동소맥귀일화식피지수（normalized difference vegetation index，NDVI）적경험반방차함수，채용최소이승산법진행모형의합，득도동소맥지괴 NDVI 최우반방차모형급삼수（기태치、변정、괴금치）。결과표명，불동동소맥지괴NDVI경험반방차도정현명현적유기태모식，동소맥NDVI 표현출명현적대상이향성특정。재수직급평행롱향상，기태치여지괴내 NDVI 문리치역범위、문리방차균겁현저정상관（P<0.01），차수방향영향불대；변정재수직롱향상여지괴내부작물 NDVI 균치정겁현저부상관（P<0.01），여변이계수급NDVI소우0.40적상원복개도겁현저정상관（P<0.01）；괴금치수직롱향상여작물NDVI균치、변이계수、소우0.40적상원복개도유겁현저관계（P<0.01），변정、괴금치재평행롱향여각개인자무상관성。해연구위이용요감영상게시작물장세적공간변이제공료삼고。
Crop growth diagnosis or evaluation mainly relies on field survey, manual sampling and biochemical analysis in the laboratory. It is difficult to master the real spatial variance characteristics for the whole crop field because sample collection is restricted to the manpower and material resources, as well as the time consumption of data analysis in the laboratory. Remote sensing technology provides an opportunity for spatial variance monitoring of crop growth with its rapid development in recently years. In this study, the remote sensing image of the QuickBird with a high spatial resolution acquired on May 2nd, 2006 was used to analyze the spatial heterogeneity characteristics of winter wheat from different fields. Firstly, the coarse and precise geometric corrections were carried out by ground control points (GCP) and difference global positioning system (DGPS), respectively. Then, atmospheric correction was processed using the ‘empirical line method’ (ELM) based on ground spectral measurements. After the geometric and atmospheric corrections, a pan-sharpening process was applied to the QuickBird’s four multi-spectral bands by using the pan band. Then the normalized difference vegetation index (NDVI) image was calculated based on the QuickBird images in Band 3 and Band 4. Six winter wheat fields were selected for the spatial heterogeneity analysis through the geo-statistics method. The empirical semi-variance function was established based on the NDVI values of the pairs of pixels within the range of 0.6 meter to 27 meters in the directions vertical to ridge and parallel to ridge in all six fields. Then semi-variograms were fitted with Spherical model, Exponential model and Gaussian model, respectively. The optimization model was then selected after evaluated by the SSE (sum of squares due to error) and R2 (determination coefficient) through the method of maximum likelihood. Three parameters for semi-variogram model, sill, range and nugget were calculated for all six fields in two directions using the least squares algorithm. Meanwhile, the statistical parameters for winter wheat’s NDVI image, including the values of minimum, maximum, mean, standard deviation and coefficient of variance (CV), as well as the image texture parameters, including the data range, data variance and data entropy were calculated for all the fields. The NDVI coverage information with different value range was also used in this study. After that, the relationships between NDVI semi-variogram parameters and NDVI statistical information, texture information, and NDVI coverage information were analyzed, respectively. The results indicated that NDVI’s spatial semi-variogram showed an obvious sill pattern for wheat field. The value of sill in the direction vertical to ridge was higher than that in the direction parallel to ridge for the same field. And the range and nugget values in the two directions were also different for the same field. It can be concluded that the wheat growth shows the zonal anisotropy. The results revealed that the sill values in the directions vertical to ridge and parallel to ridge were both related to the NDVI texture range, texture variance and the NDVI coverage value. While the NDVI range was related to the NDVI mean value, CV value and the coverage of pixels with NDVI value less than 0.30 and 0.40 in the field in the direction vertical to ridge. The NDVI nugget was related to the NDVI mean value, CV value and the coverage of pixels with NDVI value less than 0.30 and 0.40 in the field in the direction vertical to ridge. But the range and nugget were irrelevant to any factor in the direction parallel to ridge. This study indicates that remote sensing technique can provide an effective new method for the study on spatial heterogeneity of crop growth.