Methods For Estimating Water Content Of Bifacial Leaf Plants

Water content is very important to the health status of vegetation,so it is very meaningful to monitor the water content of vegetation.The traditional water content measurement is time-consuming and costly,and the development of remote sensing technology can solve these problems.The vegetation index based on reflectance spectrum of vegetation leaves can be used to retrieve the water content of vegetation leaves well.However,most vegetation indices published,which estimated the water content,were developed by investigating the statistical relationship between spectral reflectance of the adaxial surface and water content of plants.In fact,some reflection information is also derived from the abaxial surface because of the effects of canopy structure or leaf inclination.It is difficult to guarantee the accuracy of the estimation of the water content by using the vegetation index that only uses the spectral reflectance information of the adaxial leaf surface.This study will create models to assess water content of plants based on adaxial and abaxial surface of plants.The vegetation water content parameter this study used is equivalent water thickness.In this research,the MDATT index was created and then compared with the published vegetation indices.The results showed that MDATT was better in the accuracy of the calibrating and validating models,compared with other indices in dataset of adaxial,abaxial surface and both surfaces,which indicated that MDATT was robust.In addition,the MDATT index derived from the bands near the water absorption could better retrieve the EWT.In addition to the vegetation indices,the spectrum of samples was processed by wavelet analysis and partial least squares in this study.It showed that in the estimation of EWT by wavelet analysis,the single surface(adaxial or abaxial)associated with the EWT was relatively strong,and the correlation in both surfaces was relatively poor.In case of predicting ability,except for the abaxial surface and both surfaces of Virginia creeper,it performed well.In the estimation of EWT by partial least squares regression,the modeling accuracy was very high(R~2>0.9)and in predicting performance.Besides the abaxial surface of Virginia creeper(R~2=0.83),abaxial surface(R~2=0.71)and both surfaces of White polor(R~2=0.82),the rest of the predictions were good(R~2>0.9).Overall,in most of data sets,the prediction ability of partial least squares was the best(highest R~2,minimum RRMSE).Besides the abaxial of White poplar,prediction of others were better(R~2>0.8),the prediction ability of partial least squares was the best in dataset of all samples.Therefore,the partial least squares can be considered to estimate the EWT of leaves accurately and without the influence of the adaxial and abaxial surfaces.