论文部分内容阅读
在田间利用高光谱技术监测土壤含水率(Soil Moisture Content,SMC)成为精准农业研究的热点之一,但农田原状光谱受到土壤表层属性如表面粗糙度、质地、微聚体和其它环境因素的影响,且小尺度区域SMC空间差异较小,增加了SMC光谱信息的提取难度,导致SMC的估算精度较低;基于实验室内经过筛制备的土壤样品的光谱数据建模,虽然模型精度较高,但人为改变土壤结构和紧实度的预处理方式无法表征农田SMC的实际状况.因此,该文尝试提出一种耦合土样原状光谱数据和标准光谱数据估算农田SMC的新方法.通过获取江汉平原潮土土样的原状光谱反射率(Rund)和烘干光谱反射率(Rdry);基于Rdry确定研究区同一土壤类型在烘干状态下(SMC为0)的标准光谱(Std-R);采用差值、比值、归一化的方法耦合Rund和Std-R,得到耦合光谱(Cpl-RS、CplRD、Cpl-RN);提取耦合光谱中水分敏感波段的光谱(Moe-RS、Moe-RD、Moe-RN),基于偏最小二乘回归方法(PLSR)建立SMC的估算模型.结果表明,标准光谱具有良好的代表性,能够为光谱耦合提供统一且稳定的背景值;耦合土样的原状光谱和标准光谱可以有效地削减土壤水分以外其它因素对土壤高光谱观测的影响;利用耦合光谱的水分敏感波段建立的SMC估算模型相较基于Rund建立的模型,有效降低了模型的复杂度,精度有较大程度地提升,建模集R2c从0.46最高上升至0.61,验证集R2p从0.49最高上升至0.71,RPD值从1.39最高上升至1.72,模型的稳定性、拟合度和预测能力都得到提升.该方法简单、易推广,为快速准确评估农田墒情提供了新途径.
The use of hyperspectral techniques to monitor soil moisture content (SMC) in the field has become one of the hot topics in precision agriculture. However, the original spectrum of farmland is affected by soil surface properties such as surface roughness, texture, micro-aggregates and other environmental factors , And the SMC space difference is small in the small-scale area, which makes it difficult to extract SMC spectral information and leads to low accuracy of SMC estimation. Based on the spectral data modeling of the soil samples prepared in the laboratory through sieving, although the accuracy of the model is high, However, man-made changes in soil structure and compaction of the pretreatment methods can not characterize the actual situation of farmland SMC.Therefore, this paper attempts to propose a new method of coupling soil undisturbed spectral data and standard spectral data to estimate farmland SMC.By obtaining the Jianghan Plain The original spectral reflectance (Rund) and dry spectral reflectance (Rdry) of the soil samples were determined based on Rdry. The standard spectrum (Std-R) of the same soil type in the dry state (SMC = 0) The coupling spectra (Cpl-RS, CplRD, Cpl-RN) were obtained by coupling Rund and Std-R using the difference, ratio and normalized methods. The spectra of Moe-RS and Moe- RD, Moe-RN), the estimation model of SMC was established based on Partial Least Squares Regression Method (PLSR). The results show that the standard spectra are well represented and provide a uniform and stable background for spectral coupling. The original spectrum and the standard spectrum can effectively reduce the influence of soil moisture on soil hyperspectral observation. The SMC estimation model based on coupling moisture spectrum can reduce the complexity of the model compared with the model based on Rund. The accuracy of R2c increased from 0.46 to 0.61, R2p increased from 0.49 to 0.71, and the RPD increased from 1.39 to 1.72. The stability, fitness and predictive ability of the model were all good The method is simple and easy to popularize, which provides a new way for rapid and accurate assessment of soil moisture in farmland.