基于高光谱参数的柑橘园土壤水分动态监测

摘要：土壤水分是提高柑橘产量和品质的关键因素，为了高效、无损、精准的获取柑橘园土壤水分动态变化，利用ASD光谱仪选取了适宜的响应波段（350~1075nm）的数据作为试验光谱反射率，采用多元线性逐步回归分析（SMLR）对提取的特征波段反射率及9种光谱的转换形式的数据进行计算和分析，并利用实测柑橘根系0~60cm的土壤含水率进行验证，建立了预测柑橘园土壤含水率的高光谱模型。结果表明：土壤含水率在0~20cm的深度条件下变化最为明显，有助于提高模型的预测精度；光谱的微分处理较非微分处理，与波长的关系曲线波动更大且反演精度显著上升；柑橘园的试验样本水分的特征波段在700~760nm以及950nm左右是进行建模优先考虑的特征波段；光谱对数（1gR）的一阶导数和倒数的对数（lgR-1）的一阶导数对土壤水分的拟合精度较高，两种拟合方式的决定系数（R2）均达到0.876以上，均方根误差（RMSE）均达2.19%，相对分析误差（RPD）均达7.107；其中光谱对数（1gR）的一阶导数为预测柑橘园土壤含水率的最优模型，在进一步验证中实测值和模型计算值拟合的相关系数高达0.992。因此，基于光谱对数（1gR）的一阶导数构建的模型可实现对柑橘园土壤水分的精确监测。

Soil moisture dynamics monitoring in citrus orchards based on UAV spectral remote sensing

Abstract: Soil moisture is the key factor to improve yield and quality of Citrus, in order to obtain soil moisture in formation of citrus orchard efficiently, nondestructively and accurately. ASD spectrometer was used to select the appropriate response band (350~1075nm) data as the test spectral reflectance. Stepwise multiple linear regression (SMLR) was used to calculate and analyze the extracted characteristic band reflectance and the data of 9 spectral conversion forms. The measured soil moisture content of citrus roots in 0~60cm was used to verify and establish a hyperspectral model for predicting soil moisture content in citrus orchards. The results show that the variation of soil moisture content is the most obvious at 0~20cm, which is helpful to improve the prediction accuracy of the model; Compared with non-differential treatment, the relation curve between spectrum and wavelength fluctuates more greatly and the inversion accuracy in creases significantly; The first derivative of the logarithm of the spectrum (1gR) and the first derivative of the reciprocal of the logarithm of the spectrum (lgR-1) had high fitting precision for soil moisture, the determination coefficient (R2) of the two fitting methods was above 0.876, the root mean square error (RMSE) was 2.19%, and residual predictive deviation (RPD) was 7.107; The first derivative of the logarithm of the spectrum(1gR) is the optimal model for predicting the soil moisture content of the citrus orchard. In further verification, the correlation coefficient between the measured value and the calculated value of the model was as high as 0.992. Therefore, the model based on the first derivative of the logarithm of the spectrum (1gR) can achieve accurate monitoring of soil moisture in citrus orchard.