融合星载LiDAR系统GEDI数据与Sentinel-2影像的长江中游洲滩典型禾本科植物高度动态研究

植物是大型河流生态系统的重要成分。但受气候变化和人类活动影响,洲滩禾本科植物高度不断发生调整,进而影响洲滩生境和河道防洪安全,故需长期监测。近年来,伴随着星载激光雷达(LiDAR)技术的发展,应用LiDAR卫星数据反演洲滩禾本科植物高度成为一种可能。本文融合新一代星载LiDAR系统GEDI数据与Sentinel-2影像,基于XGBoost算法构建了考虑物候、累积温度与光合有效辐射指标的洲滩典型禾本科植物高度外推模型,同时利用Attention-UNet算法搭建了洪淹区域识别模型。随后以长江中游洲滩为例,探明了星载LiDAR技术在获取洲滩植株高度方面的性能,分析了各指标对模型精度的影响,并初步得出了洲滩典型禾本科植物高度对不同淹没条件的响应模式。主要结论包括:(1)星载LiDAR系统GEDI具有准确探测洲滩植物高度的能力,与无人机航测数据相比RMSE=0.43 m;(2)运用GEDI数据构建禾本科植物高度外推模型时,考虑物候和累积温度等指标可有效提升模型精度,提升幅度为6.8%～10.7%;(3)利用无人机航测数据对模型外推植物高度进行评价,RMSE=0.80m。同时从模型外推结果中可知...

Integrating GEDI and Sentinel-2 data for mapping height dynamics of floodplain representative Poaceae vegetation in the Middle Yangtze River

Vegetation plays a vital role in riverine ecosystems. However, the height of Poaceae vegetation on floodplains encountered continuous adjustments under the combined effects of climate change and anthropogenic activities. These adjustments influence the physical habitat and flood risk management within fluvial systems. Therefore, a long-term and large-scale surveillance of floodplain vegetation height is crucial. In recent decades, advancement in spaceborne Light Detection and Ranging (LiDAR) technology has made it feasible to obtain the vegetation height. Here, we integrated the new-generation spaceborne LiDAR system GEDI and Sentinel-2 data to construct a XGBoost-based vegetation height model specified for floodplain poaceous vegetations by adopting phenological metrics, accumulated temperature, and photosynthetically active radiation. Meanwhile, a flood inundation detection model was constructed based on the Attetion-UNet algorithm. Taking the floodplains in the Middle Yangtze River as a representative example, the applicability of spaceborne LiDAR technology in obtaining poaceous vegetation height was examined. Furthermore, the contribution of each indicator to the model and the vegetation height dynamics in response to different inundation patterns was analysed. The main findings are as follows: (1) spaceborne LiDAR system GEDI demonstrates a capability for accurately detecting the Poaceae vegetation height on floodplains, with an RMSE of 0.43 m compared to the UAV-detected data; (2) Incorporating phenological and accumulated indicators into the vegetation height model can effectively improve the model accuracy by 6.8%-10.7%; (3) The vegetation height model yields an RMSE of 0.80 m compared to UAV-detected data. The outcome of vegetation height model demonstrated that the average vegetation height in 2020 decreased by 0.03-0.24 m compared to 2019 as a result of the catchment-level flood in 2020; (4) Affected by the 2020 catchment-level flood, the Poaceae that were submerged <10 days demonstrated an upward trend in their height in the next year, whereas for those submerged >10 days, an average decrease of 2.3% to 3.1% in their height was observed. Furthermore, for floodplains where the daily average submersion depth exceeded 0.95 times plant height, the negative impact of flooding gradually became severe with the increase in the ratio.