摘要
植被覆盖度测度的准确性很大程度上影响着研究结论是否科学合理。在干旱半干旱退化草原区,尤其是受采矿剧烈扰动的矿区,发育的生物土壤结皮(Biological soil crust,BSC)由于其颜色和光谱同绿色植被具有相似性,导致对植被覆盖度的测量存在一定的影响。以伊敏露天矿区为研究区,在西排土场和内排土场采集了含苔藓结皮、地衣结皮和藻结皮的样方相片各四组(每组中包含样方喷水前和喷水后的相片各一张),并采集了一组不含结皮的样方相片作为对照组,运用数码照相法提取植被覆盖度,通过不同的数据处理方法(最大似然分类法及RGB阈值法)进行植被覆盖度提取,设立对比试验,分析BSC对于植被覆盖度测度是否有影响,其影响大小如何,影响程度是否受BSC含水量大小的影响,并对比各常规处理方法的优劣,研究能否通过结合纹理特征与色彩信息剔除BSC对植被覆盖度提取值的影响。研究结论:1)基于照相法的常规数据处理方法提取植被覆盖度时,BSC的存在导致测得的植被覆盖度值偏高,且苔藓结皮、地衣结皮吸水后比吸水前影响更显著,藻结皮相反;2)3个演替阶段的BSC中,尤以含苔藓结皮的样方植被覆盖度高估最为明显,其次为地衣,而含藻结皮样方规律不明显;3)样方内BSC覆盖度越高,植被覆盖度越低,其植被覆盖度测度越不准确,因此在研究草原矿区这类草本植物覆盖度较低、结皮发育的区域时,应当注意BSC的影响;4)试通过应用纹理信息提出改进的提取方法,发现单纯的纹理分类精度极低,而结合了纹理信息与RGB色彩信息的分类精度较高;5)对两种常规分类方法的精度进行比较,RGB阈值法较最大似然分类法更为不准确,对植被覆盖度的高估接近最大似然分类法的2倍。对两种改进的提取方法的精度进行比较,二者都可以有效提高测量精度,基于波段合成的纹理分类方法最佳。四种方法精度由高到低的顺序为:纹理结合RGB法>考虑生物土壤结皮的最大似然分类法>普通最大似然分类法>RGB阈值法。
Vegetation coverage is of great significance to regional and global issues in the field of hydrology, meteorology, and ecology, among others. The accuracy of its estimation can profoundly affect research conclusions. With the popularization of high-precision digital cameras, photographs are widely used for the estimation of vegetation coverage in micro-regions in ecological research because of its advantages, such as objectivity and high precision. In arid and semi-arid degraded steppe, and especially in severely disturbed mining areas, the development of biological soil crust （BSC） can affect the photographically measured value of vegetation coverage because its spectral signature is similar to that of vegetation. In this study, the Yimin open-pit mine area was chosen as the research area, and four groups of photographs （before and after sprinkling water） containing moss crust, lichen crust, algae crust and no BSC （control） were taken as samples. Then, vegetation coverage was extracted using the digital photographic method and the data were processed by different methods （maximum likelihood classification and RGB threshold method ） to establish a comparative test. The extracted values were compared with the truth-value, acquired by manually outlining the vegetation coverage, to analyze the effect of BSC on the vegetation coverage measurement, evaluate the extent of the effect, and determine whether the effect was related to BSC water content. Moreover, based on the comparison of conventional methods, a more accurate method that would eliminate the effect of BSC on estimated vegetation coverage was proposed by combining texture and color information. The main conclusions were as follows ： 1 ） The existence of BSC led to the overestimation of vegetation coverage when using conventional methods, and those of moss crust and lichen crust were more significant after watering, whereas the results for algae crust were opposite. 2） In the three succession stages of BSC, the samples containing moss crust led to a substantial overestimation of vegetation coverage （the results were up to ten times the truth-value using the RGB threshold method and up to six times using the maximum likelihood classification）, followed by lichen crust （ the result were up to four times the truth-value using the RGB threshold method and up to two times using the maximum likelihood classification）, but the algae crust was not significant because the variance was too large. 3 ） When BSC coverage increased or vegetation coverage decreased, the accuracy of the estimated vegetation coverage decreased, which suggested that the effect of BSC cannot be ignored in low vegetation coverage mining areas on the steppe. 4） In an attempt to improve the estimation method using texture information, the accuracy of texture classification was very low, while combining texture information with RGB color information resulted in high accuracy. 5） For the two conventional classification methods, the RGB threshold method led to vegetation coverage overestimation as large as twice that of the maximum likelihood classification. For the two proposed methods, both effectively increased accuracy and texture classification, although the method based on band stacking was better. Comparing the four estimation methods, the accuracy was ranked as follows： texture combined with RGB method 〉 maximum likelihood classification considering BSC〉 maximum likelihood classification 〉 RGB threshold method.
出处
《生态学报》
CAS
CSCD
北大核心
2018年第4期1272-1283,共12页
Acta Ecologica Sinica
基金
东部草原区大型煤电基地生态修复与综合整治技术及示范
关键词
植被覆盖度
照相法
生物土壤结皮
露天矿区
纹理分析
vegetation cover
photographic method
biological soil crust
open-pit mine
texture analysis
作者简介
通讯作者:白中科,E-mail:baizk@cugb.edu.cn
