Mapping Alien Plants In An Estuarine Wetland With High Spatial Resolution Images

Biological invasion is an important component of global change. Along with the development of globalization, biological invasion has become one of the most cumbersome environmental problems. To challenge biological invasion, land managers and ecologists have to consider how to effectively control alien invasive plants and keep local biodiversity. This needs to know the spatial distribution of invasive plants for better understanding invasion severity and intensity. However, due to rapid expansion and wide range distribution of invasive plants, no longer can traditional methods on field survey meet the demands for effectively monitoring them. As a large scale and timely survey technique, remote sensing can provide answers against such measurement and monitoring limitations. Although, to integrate various remote sensing data resources is the trend of remote sensing technology, images with high spatial resolution play important roles to accurately understand the distribution of single invasive plant. In this study, Spartina altern iflora, an alien plant that has successfully invaded Chongming Dongtan estuarine wetland, was selected as a representative invasive plant to explore how to map single invasive plant more accurately. Firstly, due to wetlands are usually hard to reach and it is sometimes difficult to warrant good weather condition to measure spectra of plants periodically, how to properly use UniSpec-DC to measure spectra of plant canopies in wetlands was studied, preparing for the following spectral discrimination analysis. Second, according to the phenological and spectral characteristics of Spartina alterniflora and another two local dominant plants(Phragmites australis and Scirpus mariqueter), the best and other potential time and spectral bands for monitoring Spartina alterniflora in one year were determined. Finally, referring to the results of spectral discrimination, a proper QuickBird image was used to monitor Spartina alterniflora. Two methods, the pixel-based classification technology and object-oriented classification technology, were compared to see their performance considering algorithms, feature spaces and image process techniques. The merits and demerits of both technologies were revealed. The results show that:1) The two channel spectroscopy UniSpec-DC can effectively weaken the deviation of measured spectra caused by change of solar irradiance, without repeat measurements of standard white panel. In addition, to guarantee the quality of data, the spectroscopy should be adjusted to maximize its signal-to-noise ratio.2) For all the three dominant plants in Chongming Dongtan, their spectral characteristics change with stages of their phenology. The best time to monitor Spartina alterniflora is the period between the late May and early June, with the best spectral bands in near-infrared region. The second best time is the days between the middle of April and the beginning of May as well as the late of December, with the distinguishable bands in visible region. Other time of possibilities to monitor Spartina alterniflora was also indicated. If data sources of different dates were combined, there would be more solutions to monitor Spartina alterniflora.3) Both pixel-based classification technology and object-oriented classification technology acquired high accuracy when the QuickBird image was used to monitor Spartina alterniflora in Chongming Dongtan. At the cost of more time, object-oriented classification technology achieved higher accuracy than pixel-based technology. However, traditional pixel-based unsupervised classification attained high accuracy more than 90% with little time. Therefore, application should depend on case-by-case basis considering both time and accuracy.