| Soil water content (θ) was an important factor for the plant growth and crop production. Hence, it was very important to effectively and non-destructive monitor surface soil water content for the irrigation and crop growth.In this study, the θ data and concurrent spectral parameters were acquired in the laboratory experiment in for sand, loam, clay loam and sandy loam soils by two methods (integrating sphere and handheld spectrometer), the image information (Value, Saturation and Hue) were extracted and analyzed to establish the prediction model of soil water content (θ) under different soil bulk density (ρ). The major conclusions are as follows:First, the reflectance gradually decreased with increasing θ and the regression equationbased on the spectral feature parameters could be used to estimate θ. The spectral feature parameters sum reflectance with green edge (coefficient of determination, R2=0.24; root mean square error, RMSE=0.09 m3 m-3) and 780-970 nm absorption depth (R2=0.31 and RMSE=0.11m3 m-3) were best correlated with θ in the sand. The 0 model based on maximum reflectance with 900-970 nm and sum reflectance with 900-970 nm had a high correlation (R2=0.92 and RMSE=0.05 m3 m-3) for the loam. The maximum reflectance with 900-970 nm (R2=0.86 and RMSE=0.03 m3 m-3) and sum reflectance with 900-970 nm (R2=0.85 and RMSE=0.03 m3 m-3) had a high correlation to estimate θ for clay loam soil. And the maximum reflectance with 900-970 nm (R2=0.87 and RMSE=0.02 m3 m-3) and sum reflectance with 900-970 nm (R2=0.87 and RMSE=0.02 m3 m-3) were the best correlation for sandy loam soil. The maximum reflectance with 900-970 nm (R2=0.48 and RMSE=0.05 m3m-3) and sum reflectance with 900-970 nm (R2=0.47 and RMSE=0.05 m3 m-3) were the best correlation for all soils. And the ANN was presented better estimation for θ (R2=0.95 and RMSE=0.03 m3 m-3) in four soils. Thus, the ANN model had great potential for θ estimation.Second, soil reflectance was obtained by the integrating sphere method in different soil texture under different soil bulk density and soil water content. The results showed that for soil with different p and texture (sand, loam, clay loam soil and sandy loam soil), the soil reflectance was decreased with increasing water content on The estimation model was the best correlation for sand (R2=0.79 and RMSE=0.05 m3 m-3), loam (R2=0.91 and RMSE=0.04 m-3), clay loam soil (R?2=0.88 and RMSE=0.04 m3 m-3) and sandy loam soil (R2=0.86 and RMSE=0.04 m3 m-3), were highly significant (p<0.01); the soil water content of four kinds had estimation accuracy higher (R2=0.68 and RMSE=0.07 m3 m-3).Third, the image information (Value, Saturation and Hue) were extracted and analyzed to establish the prediction model (Value and Saturation) of soil water content (0) under different soil bulk density.The results showed that:1) the estimation model was the best correlation (R2=0.82 and RMSE was 0.05 m3 m-3), when the sand bulk density was 1.50 g cm-3; the correlation of estimation model was the best (R2=0.82 and RMSE=0.05 m3 m-3), when the loam bulk density was 1.50 g cm-3; the correlation of estimation model was the best (R2=0.93 and RMSE=0.04 m3 m-3), when the loam bulk density was 1.40 g cm-3; the best estimation model for clay loam soil was under soil bulk density was 1.60 g cm-3, corresponding to R2 and RMSE were 0.93 and 0.03 m3 m-3, respectively; the correlation of estimation model was the best for sandy loam soil (R2=0.87 and RMSE=0.07 m3 m-3) in the soil bulk density was 1.60 g cm-3. With considering the change of soil bulk density, the correlation of estimation model was the best for clay loam soil (θ=1.89-0.72×V-1.50×S, R2=0.72 and RMSE=0.06 m3 m-3), the R2 and RMSE of loam were 0.70 and 0.07 m3 m-3, respectively; the R2 and RMSE of sand were 0.70 and 0.07 m3 m-3, respectively; the R2 and RMSE of sandy loam soil were 0.60 and 0.07 m3 m-3, respectively. Artificial neural network (ANN) could improve the measurement accuracy of soil water content four types soil, the R2 and RMSE of sand were 0.67 and 0.07 m3 m-3, respectively; the R2 and RMSE of loam were 0.79 and 0.07 m3 m-3, respectively; the R2 and RMSE of clay loam soil were 0.82 and 0.07 m3 m-3, respectively; the R2 and RMSE of sandy loam soil were 0.88 and 0.05 m3 m-3, respectively.In summary, in this study, soil reflectance and image information were obtained using the integrating sphere method, spectrometer method and image method on four kinds of different soil bulk density of and surface soil water content, two θ estimation models were established based on soil reflectance and color; then analysis and comparison estimate 9 model accuracy of these three methods. The results showed that the accuracy of estimation of θ in the order of p and different soil texture (sand, loam, clay loam soil and sandy loam soil) as follows:integrating sphere> spectrometer method>image processing. Artificial neural network model should improve the measurement accuracy of soil water content. |
PDF Full Text Request