A New Index for Remote Sensing of Soil Organic Carbon Based Solely on Visible Wavelengths
A New Index for Remote Sensing of Soil Organic Carbon Based Solely on Visible Wavelengths
复制标题
仅基于可见波长的土壤有机碳遥感新指标
DOI：
--
复制
发表时间：
2019
影响因子：
2.9
通讯作者：
Qian Yu
中科院分区：
文献类型：
--
作者：
E. Thaler;I. Larsen;Qian Yu
遥感是绘制土壤属性图的有力手段，例如土壤有机碳(SOC)是土壤质量的关键属性。已开发了依赖于短波红外(SWIR)或近红外(NIR)波长的光谱遥感指数，以量化SOC中的空间模式。然而，由于对高分辨率多光谱或高光谱图像的需求，基于SWIR和NIR的指数在量化SOC精细尺度模式方面的应用受到限制。可见波长几乎是由所有传感器测量的，通常是高分辨率的；因此，开发基于可见波长的指数可以极大地提高远程估计SOC的能力。在这里，我们通过使用7916 SOC和美国农业部全国快速碳评估的高光谱测量来评估实验室测量的SOC和光谱反射率之间的关系，从而开发出这样一个指数。我们的新SOC指数(SOI)预测7916个样本的SOC浓度，RMSE为1.5%，与SWIR/NIR比率(RMSE=1.3%)的预测相当，并优于基于近红外和红光波长的指数预测(RMSE=2.8%)。我们将该指数应用于一张高分辨率的卫星图像，并测试了基于图像的SOI预测爱荷华州一块耕田的SOC测量浓度的能力。无论有无局部校准数据，回归模型都能准确地预测实测有机碳，其均方根误差(∼)为0.5%。鉴于可见波长的光谱数据的图像的广泛可获得性，有可能利用SOCI来解决一系列土壤--农学问题。
Remote sensing is a powerful method for mapping soil properties, such as soil organic carbon (SOC), a key property of soil quality. Spectral remote sensing indices that rely on shortwave-infrared (SWIR) or near-infrared (NIR) wavelengths have been developed to quantify spatial patterns in SOC. However, the application of SWIR- and NIR-based indices for quantifying fine-scale patterns of SOC is limited due to the requirement of high-resolution multispectral or hyperspectral imagery. Visible wavelengths are measured by virtually all sensors, often at high resolution; hence, development of a visible wavelength–based index can greatly increase the ability to remotely estimate SOC. Here we develop such an index by assessing the relationship between laboratory-measured SOC and spectral reflectance using 7916 SOC and hyperspectral measurements from the nationwide USDA Rapid Carbon Assessment. Our new SOC index (SOCI) predicts SOC concentrations for the 7916 samples with a RMSE of 1.5%, which is comparable to predictions from the SWIR/NIR ratio (RMSE = 1.3%) and outperforms the predictions of an index based on NIR and red wavelengths (RMSE = 2.8%). We applied the index to a high-resolution satellite image and tested the ability of the image-based SOCI to predict measured SOC concentrations for a plowed field in Iowa. Regression models with and without local calibration data accurately predict measured SOC, with RMSE values of ∼0.5%. Given the widespread availability of imagery with spectral data in the visible wavelengths, there is potential to use the SOCI to address a range of soil-agronomic problems.