CAJ | 학술논문

针对当前遥感抽样估算中分层标志缺乏遥感识别误差描述的问题,该文探讨了基于农作物遥感识别结果的不同分层方法的抽样效率.以江苏省为研究区,采用2阶段分层,采用数字高程模型(digital elevation model, DEM)标准差进行一阶段分层,在一阶段分层的基础上,分别采用农作物识别种植规模、遥感识别破碎度、种植结构以及种植结构与破碎度指标进行二阶段分层.试验结果表明:种植结构与破碎度指标的分层效率最高,相对效率达到5.90,该分层指标融合了遥感分类结果反演出的种植结构和破碎度,不但能够有效地反映出农作物区域的景观特征,同时也较为合理地反映出区域间作物种植的差异性,为提高省级农作物种植面积遥感抽样估算效率提供有力的参考.
침대당전요감추양고산중분층표지결핍요감식별오차묘술적문제,해문탐토료기우농작물요감식별결과적불동분층방법적추양효솔.이강소성위연구구,채용2계단분층,채용수자고정모형(digital elevation model, DEM)표준차진행일계단분층,재일계단분층적기출상,분별채용농작물식별충식규모、요감식별파쇄도、충식결구이급충식결구여파쇄도지표진행이계단분층.시험결과표명:충식결구여파쇄도지표적분층효솔최고,상대효솔체도5.90,해분층지표융합료요감분류결과반연출적충식결구화파쇄도,불단능구유효지반영출농작물구역적경관특정,동시야교위합리지반영출구역간작물충식적차이성,위제고성급농작물충식면적요감추양고산효솔제공유력적삼고.
In current studies on estimation of crop area using remote sensing on a stratified sampling, researchers tend to use the classified images directly as stratification index and do not consider the impacts of classification errors coming from the stratification procedure. In this study, we tried to improve the efficiency and accuracy of crop area estimation at provincial level for solving this problem. Jiangsu Province was chosen as study area, and the crop area was estimated by a stratified sampling. We developed two steps to make stratification. Firstly, taking account the categorization of terrestrial geomorphology and the situation of study area, standard deviation of DEM was used as a stratification index to categorize the whole study area into two different layers. The study area where its standard deviation of DEM is lower than 10 m is classified as flat terrain layer, while the others are classified as alpine terrain layer. Secondly, in each first layer we further stratified it into sub-layer (i.e. classification) by using 4 different indices individually, including crop scale, crop structure, cropland fragmentation, combination of crop structure and fragmentation. The stratification results showed that there were some problems to identify wheat area with directly using remote sensing method. There are low planting structure and fragmented planting areas located in large scale crop layer due to the effect of plant spatial distribution referred to various crops in the same growing period. In order to eliminate the problem of confusion caused by crops in the same growing period, the index of crop structure was introduced in this study. For the purpose of quantitative analysis of the heterogeneity of plant spatial distribution arisen by mixed pixels, the index of cropland fragmentation was introduced. In order to analyze the effect of stratification by different indexes, we calculated the coefficient of relative stratification efficiency, which was defined as the ratio of variance of random sampling to variance of a given stratification. The results showed that the relative efficiencies of above four stratified methods were better than unstratified simple random sampling. Among the single index, the relative stratification efficiency achieved the highest value of 4.96 by using the crop structure. As a whole, the relative stratification efficiency was the highest by using the combination of crop structure and cropland fragmentation as a stratification index. This index can reflect not only the fragmental characteristics of cropland but also the difference among different regions in our study area; therefore, it contributes to the improvement of the accuracy and efficiency of crop area estimation at provincial level.