干旱区稀疏芦苇盖度遥感信息提取  

Extraction of vegetation fraction over the sparse reed in arid area

作　　者：古丽.加帕尔[1] 陈曦[1] 马忠国[1,2] 

机构地区：[1]中国科学院新疆生态与地理研究所,新疆乌鲁木齐830011 [2]中国科学院研究生院,北京100039

出　　处：《干旱区地理》2010年第6期988-996,共9页Arid Land Geography

基　　金：国家自然科学基金(No.40801146);中科院"西部博士"项目(XBBS200801);中国科学院"优秀博士学位论文;院长奖获得者科研启动专项资金"共同资助

摘　　要：选择对角线法、之字型法、随机采用法及全采样法提取干旱区稀疏芦苇覆盖度,对比分析不同采样方法获取参数的精度,同时结合遥感影像,采用线性混合像元分解模型、亚像元变密度分解模型、三波段最大梯度差模型提取样地覆盖度信息,与地面实测覆盖度参量信息进行对比分析,探讨适宜的干旱区植被盖度野外监测方法及遥感模型。研究表明:对角线法及之字型法所获取参数可以满足样地总体植被覆盖度参数精度要求;地面验证结果表明:2006年线性混合像元分解模型所提取的覆盖度精度最高,为19.72%,与地面实测值20%最为接近,证明该模型可有效提取干旱区低覆盖度植被信息,但端元的正确选取较难,从而影响其运用;亚像元分解模型预测值为22.30%,高于实际覆盖度值,绝对误差为11.5%;而三波段最大梯度差法模型与实测值相差最大,绝对误差达到了-75%,说明该模型对于极端干旱区稀疏植被敏感度低。Taking the sparse reeds as the research object, sample plot of 30 m×30 m were set, and separate it to thirty six small foursquare plots, vegetation fraction is derived by means of diagonal sampling, zigzag sampling, ran- dom sampling and full network sampling. The accuracies of different monitoring method mentioned above were paired for analysis. Three kinds of remote sensing inversion models,i, e. the linear spectral un-mixing model, sub- pixel un-mixing model, maximal gradient difference model, were used to derive vegetation fraction from Landsat TM remote sensing data, and the results are compared with the variables that measured in suit to determine the appropri- ate model for deriving the data of the coverage fraction of sparse desert reed in arid area. The results indicate as fol- lows ： Percent cover by means of diagonal and zigzag sampling could characterize the parameter of sparse reed in si- tu; It was also showed that linear un-mixing spectral model had a highest precision, the model predicted was 19.72% matching well the field measured value 20%, being applicable for extract the coverage of sparse desert reed using remote sensing, but the selection of end member is so difficult, and affects the application of the model; The results by the sub-pixel un-mixing model showed that the predicted and absolvable errors are 22.30% and 11.5 %, separately, higher than that in situ measured value, high precision could be obtained based on finely model variable LAI or extinction coefficient for example, that needed to measure lots of vegetation variable ; The results pre- dicted by the maximal gradient difference model were underestimated, absolvable error is up to -75%, indicating that this model was not sensitivity for sparse vegetation in arid area. Limited by the arid environment, the sparse vegetation in arid area lacks of the spectrum characters of the vegetation at bands 560 nm,660 nm and 830 nm that were selected in maximal gradient difference model are sensitivity to the health vegetation. It is the ma

关 键 词：塔里木河 稀疏植被 覆盖度 适宜遥感模型 

分 类 号：P237[天文地球—摄影测量与遥感]