Monitoring Leaf Senescence Under Water Stress After The Heading Stage With Hyperspectral Reflectance In Wheat

Senescence is one of the key factors which influence wheat production.Water,drought,temperature stresses and growth regulators could lead to early aging in wheat.Hyperspectral remote sensing is a rapid and nondestructive monitoring technology,which could be applied to monitor early aging.In this study,wheat pond culture and pot culture in two continuous years were carried out to study the aging process of capable leaves after heading.Biological parameters(such as morphological parameter(the ratio of green part length to the whole leaf,GLR),biophysical parameter(the content of chlorophyll,Chi)and biochemical parameter(malondialdehyde,MDA))sensitive to senescence were obtained and selected systematically.Hyperspectral reflectance were collected synchronously and the relationship between those biological parameters and hyperspectra was analyzed quantitatively.Vegetation index(Vegetation index,VI)and continuous wavelet analysis(Continuous Wavelet Analysis,CWA)were employed to analyze those parameters,such as GLR,Chi and MDA).Spectra sensitive to senescence were determined,which could be used to monitor wheat leaf senescence indirectly.Take GLR as an example,the criterion of wheat leaf senescence and the thresholds of biological and spectral parameters were determined through probabilistic method,which provides technical support and theoretical basis for rapid,accurate and nondestructive monitoring senescence of wheat.First,three parameters(GLR,Chi and MDA)sensitive to senescence were selected,which was based on having known biological parameters(such as morphological,biophysical and biochemical parameters)were sensitive to senescence.The trend of GLR and single leaf hyperspectra and the quantitative relation in the process of aging were analyzed under different water content and leaf positions.Vegetation index and continuous wavelet analysis were employed to establish the monitoring model for GLR.Results demonstrated that visible,near infrared high reflectance platform and red edge region were sensitive to GLR.The most sensitive was the red edge region.The new-proposed vegetation indexes NDSI(R734,R713),RSI(R713,R736),and the previous vegetation indices NDVI705,mNDVI705,could well estimate the GLR.Wavelet functions mexh(W726,S7),cmor2-0.1(W451,S5)and cmorl-0.1(W451,S5)performed best in monitoring GLR.Among all the methods,mNDVI705(R2 = 0.720)and mexh(W726,S7)(R2 = 0.747)performed best.Probabilistic relations between GLR and wavelet power mexh(W726,S7)and vegetation index mNDVI705 in the senescence judgement were further analyzed synthetically.Probability model and thresholds(mexh(W726,S7)= 0.17 and mNDVI705 = 0.6)for wheat leaf senescence were determined.Namely,when mexh(W726,S7)<0.17and mNDVI705>0.6,it indicated that the leaves did not grow old,otherwise,leaves did.The probability model was tested by independent data.Results indicated that the classification accuracy of the probability model was 85.30%,and the recall were 85.61%(GLR=1)and 85.06%(GLR?1).Next,biophysical parameter(Chl related to photosynthesis)and biochemical parameter(MDA reflecting the degree of cellular damage)were selected to improve simulation accuracy,because morphological parameter based on GLR has some hysteretic nature in monitoring leaf senescence.Results showed under different water treatment,Chlorophyll content was low at the beginning,increased as leaf grows,and decreased gradually in the aging process.Red edge region was sensitive to Chl and senescence.Vegetation index NDSI(R796 R709),NDSI(R581,R513),RSI(R710,R796),NDVI705,mNDVI705(R2= 0.586)and wavelet functions mexh(W710,S3),rbio3.7(W592,S6)performed well in monitoring the content of chlorophyll.Among all of vegetation indices and wavelet powers,mNDVI705(R2 = 0.586)and mexh(W726,S7)(R2 = 0.635)performed best in monitoring Chl content and indicating leaf senescence.Finally,MDA inverted by hyperspectra was applied to monitor wheat leaf senescence.Results indicated that under different water treatment,MDA content was low in the non-senescence conditions and high in the aging process.The higher degree of senescence and the heavier the stress,the higher the content.Results based on spectral index and wavelet analysis showed that the most sensitive spectral area of MDA was the visible region mainly located in all the blue and green.Vegetation index RSI(R442 R659)(R2 = 0.845)and wavelet power rbio3.1(W476,S8)(R2 = 0.882)had the best performance in monitoring MDA content and indicating the aging process.