光学与雷达遥感协同的大尺度草地灌丛化监测研究

Study on large scale grassland shrub monitoring based on optical and radar remote sensing

受全球变化及人类利用方式强度影响,草地灌丛化已成为干旱半干旱地区草原生态系统最为突出问题之一。准确掌握大尺度草地灌丛化信息,对草地可持续利用管理及气候变化响应分析具有重要的意义。极化合成孔径雷达在地表粗糙度、灌木生物量等估算上已经展现出了一定潜力,其与多光谱光学影像的结合从理论上来说具备开展草原灌丛化监测的可能,然而目前严重缺乏这方面的研究。本研究选择锡林郭勒草原为研究区,以欧空局Sentinel-1与Sentinel-2时间序列影像及地面草原灌丛化样地观测数据为数据源,通过相关性分析及随机森林等模型方法,开展了锡林郭勒草原的灌丛覆盖度估算研究。研究结果表明:(1)光学数据与草地总覆盖度显著相关,但对草地灌丛化特性不敏感,相对来说3月份的红边植被指数与6月份的光学植被指数与灌丛覆盖度相关性略高。(2)极化合成孔径雷达数据对于灌丛覆盖度有较强敏感性(VH极化好于VV极化),且高相关性贯穿了整个生长期,其中6月份的VH极化数据与灌丛覆盖度相关性可达0.64。(3)联合光学与雷达数据的随机森林模型可以实现较高精度草地灌丛覆盖度估算(R^(2)=0.76,RMSE=0.05),其中雷达数据Sentinel-1的贡献度为71.54%,光学数据Sentinel-2的贡献度为28.47%。

Affected by the global change and the intensity of human utilization,grassland shrub has become one of the most prominent problems of grassland ecosystem in arid and semi-arid areas.It is important for the sustainable utilization of grassland and the analysis of climate change to grasp the information of large-scale grassland shrub accurately.Polarimetric synthetic aperture radar has shown a certain potential in the estimation of surface roughness and shrub biomass.The combination of PolSAR images and multispectral optical images is theoretically possible to carry out grassland shrub monitoring.However,there is a lack of research in this area.In this study,Xilingol grassland was selected as the study area,and the time series images of Sentinel-1 and Sentinel-2 of ESA and the observation data of grassland shrub sample were used as the data sources.The shrub cover of Xilingol grassland was estimated by correlation analysis and random forest model methods.The results show that:(1)The optical data is significantly related to the total grassland vegetation cover,but not sensitive to the characteristics of grassland shrub cover.The red edge vegetation index in March and June have slightly higher correlation with the shrub cover on grassland.(2)PolSAR data have a strong sensitivity to shrub cover(VH polarization is better than VV polarization),also the high correlation runs through the whole growth period,among which the correlation between VH polarization data and shrub cover in June can reach 0.64.(3)The random forest model based on the combination of optical and radar data can achieve a high accuracy estimation of grassland shrub coverage(R^(2)=0.76,RMSE=0.05),in which the contribution of sentinel-1 radar data is 71.54%,and that of sentinel-2 optical data is 28.47%.