UAV based soil moisture remote sensing in a karst mountainous catchment
UAV based soil moisture remote sensing in a karst mountainous catchment
复制标题
基于无人机的喀斯特山区土壤湿度遥感
DOI：
10.1016/j.catena.2018.11.017
复制
发表时间：
2019
期刊：
影响因子：
6.2
通讯作者：
Zhang Rongfei
中科院分区：
文献类型：
--
作者：
Luo Wei;Xu Xianli;Liu Wen;Liu Meixian;Li Zhenwei;Peng Tao;Xu Chaohao;Zhang Yaohua;Zhang Rongfei
土壤水分的空间分布是农业和生态学研究和管理的前提。然而，它仍然是一个挑战，以检索SM数据在高度异质性的景观。通过调查环境因素（土壤、植被和地形）和比较不同的遥感源（Landsat-8，Radarsat-2，ASTER全球数字高程模型V002（ASTGTM 2），无人机）本研究确定了高度异质性景观中土壤有机质空间分布的关键控制因子，并建立了土壤有机质遥感估算模型。结果表明，植被类型（35.7%）、坡向（7.7%）、高度指数（4.2%）、土壤容重（3.3%）、土壤全氮（3.1%）、坡向与植被类型交互作用（3.4%）和土壤全磷（1.3%）对土壤SM变异的解释率为58.8%。SM和地形导数之间的相关性随DEM分辨率（1-50 m）的不同而不同，高度指数、坡度和坡向在7 m时达到最大值，流量累积和地形湿度指数在16 m时达到最大值，曲率在43 m时达到最大值。偏最小二乘回归分析表明，光学和红外波段从Landsat-8和地形导数无人机摄影测量DEM与SM比其他数据集更强的相关性。提出了一个经验模型（SM= 9.27 <$10 − 2 HI − 1.82 <$10 − 5 B5 + 0.519），其中仅使用高度指数和Landsat-8的B5波段作为输入，因为它显示了可接受的性能（R2= 0.36; RMSE = 0.076）。研究结果为喀斯特山区及类似异质景观的SM遥感提供了有用的信息。
Spatial distribution of soil moisture (SM) is a prerequisite for research and management of agriculture and ecology. However, it is still a challenge to retrieve SM data in highly heterogeneous landscapes. By investigating environmental factors (soil, vegetation and topography) and comparing different remote sensing sources (Landsat-8, Radarsat-2, ASTER Global Digital Elevation Model (DEM) V002 (ASTGTM2), unmanned aerial vehicle (UAV)) for karst mountainous catchments of southwest China, this study identified key controlling factors on the spatial distribution of SM and built a remote sensing model for SM estimation in highly heterogeneous landscapes. Results showed that vegetation type (35.7%), aspect (7.7%), height index (4.2%), soil bulk density (3.3%), soil total nitrogen (3.1%), aspect interact with vegetation type (3.4%) and soil total phosphorous (1.3%) totally explained 58.8% of the SM variability. The correlations between SM and topographic derivatives varied with DEM resolutions (1–50 m), and generally reached their highest values at 7 m for height index, slope gradient, and aspect, 16 m for flow accumulation and topographic wetness index, and 43 m for curvature. Partial least-squares regression analysis showed that optical and infrared bands from Landsat-8 and topographic derivatives from UAV photogrammetry DEM were more strongly correlated with SM than other datasets. An empirical model (SM= 9.27 ∗ 10−2HI− 1.82 ∗ 10−5B5 + 0.519) with only height index and B5 band from Landsat-8 as inputs is proposed, as it shows acceptable performance (R2= 0.36; RMSE = 0.076). The results of this study provide useful information for SM remote sensing in karst mountainous area and similar heterogeneous landscapes.