Spectral models for estimating water content in Eucalyptus leaves

Abstract

The relative water content (RWC) and equivalent water thickness (EWT) are parameters that provide a diversity of information about the plant’s hydric condition. Hyper-spectral remote sensing is a fast and non-destructive technique capable of generating data that allow the quantification of vegetation’s hydric condition. This study seeked to establish the existing relations among the water parameters (RWC and EWT) with the leaf spectral response of different species and hybrids of Eucalyptus. For the determination of the relative water content (CRA) it was necessary to determine the fresh weight (FW), the saturated weight (SW) and the dry weight (DW) and later carry out the spectral readings of each Eucalyptus leaf with the Fieldspec hyper-spectral sensor. Eight spectral indices were tested for the prediction of water parameters: Water Index (WI), Moisture Stress Index (MSI), Normalized Difference Water Index (NDWI), Normalized Difference Infrared Index (NDII), Simple Ratio 701 and 820 (SR701,820), Simple Ratio 1300 and 1450 (SR1300,1450), (R850-R2218)/(R850-R1928) and (R850-R1788)/(R850-R1928) in which the SR1300,1450 index was found R2= 0.72 when correlated with CRA and an R2=0.81 when correlated with EEA. The spectral data were correlated to the water parameters and it was found that the RWC at 1881 nm presented a maximum negative coefficient of correlation of r=-0,89 whereas the EWT presented a maximum negative coefficient of correlation of r=-0,79 at 2165 nm. Four methods of selecting hyperspectral variables were tested to generate a mathematical model through linear regression. For the RWC parameter the stepwise variable selection method generated the higher R2=0,86 with a RMSE = 13,85%, considering that just six predicting variables were left in this method. While the variable selection method by spectral regions was the most precise to predict the EWT parameter with an R2=0,87 and an RMSE=0,0012g/cm2. It is possible to predict CARA and ELA, for the generation of Eucalyptus by means of mathematical models derived from hyperspectral data.