%0 Journal Article
%T 基于宽范围动态植被指数的高分四土壤湿度反演<br>GF4 Soil Moisture Inversion Based on Wide Range Dynamic Vegetation Index
%A 殷铁耕
%A 张显云
%A 娄远兴
%A 普莉兰
%J Advances in Applied Mathematics
%P 8874-8883
%@ 2324-8009
%D 2022
%I Hans Publishing
%R 10.12677/AAM.2022.1112935
%X 土壤水分(SM)在农业生产、大气科学和环境监测中具有重要作用。为克服植被覆盖对SM探测的影响，众多基于归一化植被指数(NDVI)的干旱指数模型先后被提出。然而，NDVI在中高度植被覆盖区域易发生饱和，从而有可能降低SM的反演精度。为充分发挥高分四高时空分辨率的优势，以高分四红光波段反射率(Rred)、近红外波段反射率(Rnir)和宽范围动态植被指数(WRDVI)为解释变量，SM实测值为因变量，采用多元线性回归建立了贵阳市SM反演模型(下称MLR_W)，并将MLR_W模型与基于垂直干旱指数(PDI)、改进型垂直干旱指数(MPDI)和植被调整垂直干旱指数(VAPDI)的一元线性回归模型(下分别称SM_P、SM_M和SM_V模型)，以及以Rred、Rnir和NDVI为解释变量的多元线性回归模型(下分别称MLR_N)的建模精度和SM反演精度进行了比较分析。实验结果表明：由于顾及了中高度植被覆盖的影响，MLR_W较SM_P、SM_M、SM_V和MLR_N模型取得了更高的模型拟合精度，模型拟合平均绝对误差(MAE)和均方根误差(RMSE)分别为5.05%和6.38%；MLR_W模型SM预测精度高于除SM_M外的其他三种模型，SM反演的MAE和RMSE分别为5.17%和6.29%。但由于MLR_W模型无需确定土壤线，实际工作中较SM_M更易于实施。<br />
Soil moisture (SM) plays an important role in agricultural production, atmospheric science and en-vironmental monitoring. In order to overcome the influence of vegetation coverage on SM detection, many drought index models based on Normalized Difference Vegetation Index (NDVI) have been proposed. However, NDVI is prone to saturation in the medium-high vegetation cover area, which may reduce the inversion accuracy of SM. In order to give full play to the advantage of GF4’s high spatio-temporal resolution, the SM inversion model of Guiyang City (hereinafter referred to as MLR_W) was established by multiple linear regression with GF4’s red spectral reflectance (Rred), near infrared spectral reflectance (Rnir) and Wide Range Dynamic Vegetation Index (WRDVI) as ex-planatory variables and the measured SM value as dependent variable. The MLR_W model is com-bined with a unitary linear regression model based on Perpendicular Drought Index (PDI), Modified Perpendicular Drought Index (MPDI) and Vegetation Adjusted Perpendicular Drought Index (VAPDI) (hereinafter referred to as SM_P, SM_M and SM_V models respectively). The experimental results show that MLR_W has a higher model fitting accuracy than SM_P, SM_M, SM_V and MLR_N, and the mean absolute error (MAE) and root mean square error (RMSE) of the model are 5.05% and 6.38%, respectively. The prediction accuracy of MLR_W model SM is higher than that of the other three models except SM_M. MAE and RMSE of SM inversion are 5.17% and 6.29%, respectively. But due to MLR_The W model does not need to determine the soil line, which is more SM_ M is easier to implement.
%K 高分四号，土壤湿度，PDI，MPDI，VAPDI，WRDVI，多元线性回归<br>GF-4
%K Soil Moisture
%K PDI
%K MPDI
%K VAPDI
%K WRDVI
%K Multiple Linear Regression
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=59539