Hyperspectral Models for Estimating SPAD Value of Cotton Leaves under Waterlogging Stress

Object] To setup the hyperspectral sensing models for estimating SPAD value of cotton leaves under waterlogging stress. Method] Irrigation and drainage controllable plots were introduced to simulate the waterlogging stress treatment in the flowering and boll forming stage, during which the change characteristics of the cotton leaf spectral reflectance and SPAD value were observed after 1 d, 3 d, 6 d, 9 d waterlogging, respectively. To find out the hyperspectral sensing models for estimating SPAD value of cotton leaves under waterlogging stress, the correlation and regression relationships between SPAD value and spectrum parameters were analyzed. Result] (1) The SPAD value of the fourth cotton leaf from the top was significantly lower than control when suffers from waterlogging for 3 d, when waterlogged 9 d the SPAD value decreased by around 15% compared with the control. (2) The cotton suffering from waterlogged damage in the flowering and boll forming stage caused the reflection peak in green light wave band became steep, while the near infrared spectral reflectance increased, and caused the reduction of red absorption and red edge position "blue shifts", the red edge position drifts towards short wave with 4～5 nm when suffers from waterlogging for 9 d. With increase of the waterlogged days, the red edge slope and red edge area increased with a maximum value at 6 d of waterlogging, meanwhile, the skewness and kurtosis of red edge increased. (3) After waterlogging, the SPAD value of the fourth cotton leaf from the top (chlorophyll content) had a remarkable correlation with red edge slope(Dr), red edge position(λr), green peak reflection(Rg), green peak position(λg), red well position(λo), blue edge area(SDb), yellow edge skewness(Sy), yellow edge kurtosis(Ky), red edge skewness(Sr), red edge kurtosis(Kr), etc. An experience linear, polynomial and exponential models for estimating SPAD value had been built through using the Sy, Sr, Kr as independent variables, respectively, their determination coefficient (R²) were greater than 0.9, and the root mean square error (RMSE) were less than 1; and an experience binary linear regression equation for estimating SPAD value had been built through multivariate regression using the λg, SDr/SDb(VI3), Sb, Sy, Ky as independent variables, the R² was as high as 0.973, and the RMSE was 0.393. Conclusion] The model can be remote sensing model used as estimating leaf SPAD of cotton value under waterlogging stress.