Objective   A hyperspectral estimation model for the chlorophyll content in the leaves of honey pomelo was developed using spectral measurements on the characteristic parameters for the establishment of a rapid, noninvasive, accurate determination method.

  Method   Characteristic bands of spectrum of the leaves were obtained. The first-order differential spectrum and spectral characteristic variables were used to analyze the correlation between the hyperspectral bands and the relative chlorophyll content (SPAD) of pomelo leaves. Univariate estimation and multiple regression models were constructed and compared to arrive at the best prediction model.

  Result   At the wavelengths between 350 nm and 1 050 nm, the reflectance spectra on the pomelo leaves of varied SPADs decreased significantly with increasing chlorophyll content. SPAD significantly correlated with either the original or the first-order differential spectrum at several wavelengths in the visible light range. The sensitive wavelengths for the original spectral curve were 576 nm and 701 nm, and 691 nm and 748 nm for the first order differential spectrum. Based on the precision of fitting, the one that based on the hyperspectral Y_SPAD=54.67−15.75 NDVI_′691,748−10.60 GRVI_550,770+6 565.6 R′₆₉₁−6 784.58 DVI′_691,748 of 3 multiple regression models displayed the R² of 0.894, the verification R² of 0.835 6, and the RMSE of 7.07 and was selected. The univariate regression models applied the first-order differential normalized vegetation index NDVI′_691,748 and the differential vegetation index DVI′_691,748 had R² of 0.824 and 0.798, respectively, the validation R² of 0.797 and 0.7918, respectively, and RMSEs of 13.21 and 12.56, respectively.

  Conclusion   The multiple regression model based on hyperspectral indices NDVI′_691,748, GRVI_550,770, R′₆₉₁ and DVI′_691,748 could be adequately applied for estimating chlorophyll content of honey pomelo leaves.