Spectral characteristic parameter-based models for foliar nitrogen concentration estimation of Juglans regia

Abstract: With the establishment of the models based on spectral characteristic parameter for estimating foliar nitrogen concentrations of Juglans regia 'Xinwen 185', this paper aimed to explore spectral technical approaches to rapidly monitoring the foliar N nutritional status of Juglans regia 'Xinwen 185'. Based on the field experiment of fertilizer efficiency, the foliar spectral reflectance and N concentrations of the foliar samples under different amounts of N fertilizer, which were applied in the fruit setting period, the fruit rapid-growth period, the fruit fat-change period and the fruit near-mature period, were measured separately with Unispec-SC and chemical analysis in the laboratory. The Pearson correlation analysis of the foliar spectral reflectance and foliar N element concentrations was done in the 4 phenological periods of fruit development, and the spectral characteristic parameters highly significantly correlated with the foliar spectral reflectance and foliar N element concentration were selected. Finally, the spectral characteristic parameter estimating models of the foliar N element concentration were established by employing regression analysis, with the adopted functions including power function, exponential function, semi-log function, linear function, quadratic function and cubic function, and then the independent sample was used to test the precision of the models, of which the one with the highest fitting degree was screened out. The results indicated that the spectral characteristic parameters, highly significantly correlated with the foliar N element concentration (P<0.01), were reflectance of the green peak position and ratio of red index in fruit setting period, yellow edge position, red valley position, ratio of green index, ratio of red index, normalized difference of green index and normalized difference of red index in the fruit fast-growing period, reflectance of the green peak position and ratio of red index in the fruit fat-change period, and reflectance of the green peak position, he spectra slop of yellow edge and red edge area in the fruit near-mature period. Moreover, the absolute values of the correlation coefficient were all greater than 0.96 in the 4 phenological periods of fruit development. Through cubic function analysis with the reflectance of the green peak position, yellow edge position, ratio of red index and the spectra slop of yellow edge as independent variables, the regression estimation models of the foliar N element concentration in the 4 phenological periods of fruit development were established respectively, and their fitting degreesR2 were all above 0.99. In addition, the predicted values for foliar N concentration of Juglans regia 'Xinwen185' were relatively close to measured values, and all the models were examined by the confidence ellipse test; and RMSE and the relative errors (RE) of the independent sample tests were only 0.4533 g/kg and 0.4403%, 0.7716 g/kg and -0.3390%, 0.8793 g/kg and -0.3201%, 0.8145 g/kg and -0.1229% separately in the 4 phenological periods of fruit development. The models showed their significant stability and high estimation accuracy. The findings in this paper manifested that it was feasible to monitor foliar N nutritional levels of Juglans regia 'Xinwen 185' in the 4 phenological periods of fruit development by establishing spectral characteristic parameter models of foliar N element concentration estimation. Spectrum technology possesses great application potential in monitoring N element nutritional status of Juglans regia 'Xinwen 185'.