Plant survey of current vegetation of Sonoran desert plant community distribution in mountain parks in central Arizona-Phoenix, August 1999.

This study represents an effort to map the distribution of plant community types across the Central Arizona - Phoenix Long Term Ecological Research (CAP-LTER) site centered in metropolitan Phoenix using Landsat ETM data. Vegetation classification was carried out using field data collected from within the study area describing woody plant species. A system was devised which represented a compromise between providing floristic information and enabling maximum spectral discrimination between community types. Image classification used reference spectra derived from training sites in the field and was carried out on subsets defined by soil surface texture in order to control for the strong background soil signature inherent to arid regions. While groundtruthing revealed that vegetation on clayey soils was mapped to 91% accuracy, other sections produced maps with less accuracy. The results of this study demonstrate that image classification of desert vegetation using only Landsat ETM data is problematic and may not be practical without other supporting data, such as radar imaging.本研究旨在利用陆地卫星ETM（Landsat ETM）数据，绘制以凤凰城大都会区为核心的亚利桑那州中部-凤凰城长期生态研究（CAP-LTER）站点内的植物群落类型分布图谱。研究基于研究区域内采集的木本植物物种野外调查数据开展植被分类工作。本研究设计了一套兼顾植物区系信息提取与最大化区分群落类型光谱辨识度的折中方案。图像分类采用野外训练样地获取的参考光谱，并基于土壤表面纹理划分的影像子集进行处理，以消除干旱区域固有的强烈背景土壤信号干扰。野外验证结果显示，黏土区域的植被制图准确率达91%，其余区域的制图精度相对较低。本研究结果表明，仅依靠陆地卫星ETM数据开展荒漠植被图像分类存在局限性，若缺乏雷达成像等其他辅助数据支撑，该方法可能难以实际应用。