Soil High Spectral Characteristics And Soil Fertility Indexs Quantitative Inversion In Hainan Rubber Plantation

In our study, the spectra data and characteristics of rubber plantation soil were analysed and described by the high spectrometer (Field Spec ASD). In addition, the spectral reflectivity data was converted to the three types data including R, R’ and Log(1/R), which and the corresponding soil fertility indexs were used to establish the prediction models. And investigate the establishment of robust model. Finally, we selected the optimal model to predict soil nutrients quickly and effectively. The conclusions were as follows:For four different types of soil developed by the granite, the iron absorption peaks at the 760-1080 nm spectrum ranges were obvisous in the obvisous curves of the mountain brown, laterite and mountain lateritic, respectively, while it was no obvisous in the absorption curve of the brown laterite. At the 1400-1900 nm spectrum ranges, except for the absorption trough ranges, the spectrum reflectivity of the brown laterite was increased, while the spectrum reflectivity of the mountain brown laterite, laterite and mountain laterite was significantly decreased. The spectra reflectivity of four different types of soil followed in the orders:brown laterite> laterite> mountain brown laterite> mountain laterite. Except for the mountain laterite, the spectra reflectivity in different types of soil was consistent with soil organic matter, clay content.The spectra reflectivity at the 350-600 nm spectrum ranges of the rubber plantation soil which were developed from different parent materials showed a rapid rising trend, corresponding to the scarp type. And the curve trendcy varied from parent material to material. The spectra reflectivity of the laterite developed by the basalts was the lowest despite soil organic matter content was highest. Additionally, The spectra absorption troughs of different soil were obvious at near 1400 nm,1900 nm and 2200 nm points of the spectral curve.By the contrast of quantitative inversion models of soil fertility indexs, the MLR model showed a better inversion result to SOM, available-P content, and the PLSR were a better model to soil pH. As for the convertion of data, the R and R1 showed a deal result to establish the model, while the inverse-log spectrum (Log[1/R]) showed no a desired result. In the terms of the quantitative inversion of soil fertility indexs, total nitrogen content inversion, the optimal model(r=0.9253) was the one established by the R’ and MLR through the inversion to the total nitrogen content, which has high stability. The optimal model (r=0.9312) was the one established by the R’ and MLR through the inversion to soil organic matter content; As for available phosphorus and pH, the optimal model was the one established by the R and MLR (r=0.9042) and by R and PLSR (r=0.8801). The inversion model of available potassium were to be analyse further. Thus, except for available-K, the quantitative inversion of rubber plantation soil to fertility indexs was feasible on the basis of hyperspectral characteristics of soil and spectrum analysis technology, and the model possessed good accuracy and precision. All the above laid a solid foundation to detect quikly soil fertility index of rubber plantation soil.