Estimation of the nitrogen concentration of rubber tree using fractional calculus augmented NIR spectra

Natural rubber (Hevea brasikensis) is a high-valued industrial crop in tropical areas. Reasonable fertilization can improve the quality and production of the natural rubber. Hence, there is a great interest in developing an accurate and robust model to detecting the nitrogen content of rubber tree to improve the production. In this study, we propose a new method for detecting the nitrogen content using fractional calculus augmented NIR spectra. Fractional calculus is applied as a tool to extract additional information from the original spectra. Three different mathematical transforms (R, 1/R, log(R)) are utilized before the fractional order derivative analysis. After derivative analysis with different orders (0-2), the PLS regression method is utilized to develop the estimation models with the wavelengths selected by significant test of correlation coefficient at 0.01 level. The optimal prediction result R-p = 0.9245 is achieved by reciprocal spectrum with the order 0.6. Results show that fractional calculus with NIR can be utilized to build a more accurate nitrogen estimation model for rubber tree. The result presented here provides us a fast and non-destructive strategy to evaluate the nitrogen condition of rubber tree and provides a direct way to support the field management.