Hyperspectral Monitoring Of Photosynthetic Parameters And Chlorophyll Fluorescence In Cotton Leaves Under Different Water Treatments

Photosynthesis is the physiological basis of crop growth and can reflect the response of crop growth and water stress.Monitoring and diagnosing water stress and water demand accurately and timely by hyperspectral technique is of great significance for improving water management of cotton and guiding water-saving agricultural production.In this study,Aksu National Field Scientific Observation and Research Station of Farmland Ecosystem,Chinese Academy of Sciences was selected as the research area.Five irrigation gradients were set in the research area:900m³·hm^-2,1 800m³·hm^-2,2 700m³·hm^-2,3600m³·hm^-2,4 500m³·hm^-2.At the same time,hyperspectral data,photosynthetic parameters and chlorophyll fluorescence indexes of cotton leaves under different water conditions were obtained,and the change characteristics of physiological indexes and spectral characteristics of leaves of cotton under different water conditions were analyzed.Spectral index was constructed by reflectance data,and spectral index was constructed by selecting a band combination with a good correlation with 9 physiological indexes.Principal component regression（PCR）,partial least squares（PLSR）and support vector machine（SVM）were used to construct the physiological index estimation model of cotton leaves and verify the accuracy.The machine learning algorithm was used to build a physiological index estimation model,which provided a theoretical basis for real-time and rapid monitoring of cotton growth and water demand in large areas of Xinjiang by hyperspectral remote sensing.The main findings are as follows:（1）The original spectral reflectance of cotton leaves at different growth stages showed an obvious peak near 451nm and 553nm,and an obvious"valley"near 505 and 670nm.The order of the red valley reflectance and green peak amplitude was different at different growth stages.The reflectance of the first derivative spectrum shows an obvious peak near 425nm and 525nm in the blue-green band,and an obvious absorption valley appears near 570nm and 930nm.In the red band,the highest peak of the whole first order spectral curve appears near 725nm,and the position of the red edge moves towards the long-wave direction with the increase of irrigation amount.The phenomenon of"red shift"occurs.After the envelope was removed,the absorption valley appeared near 500nm and 675nm.W500（absorption width near 500nm）at flowering stage showed significant difference under different treatments.W675（absorption width near675nm）reached its maximum value at water T4 at flowering stage,and A675（absorption area near 675nm）at T4 at early flowering stage reached its maximum value.With the increase of irrigation amount,BD500（absorption depth around 500nm）and BD675（absorption depth around 675nm）showed a decreasing trend during different growth periods.（2）The photosynthetic parameters Pn,Tr,Gs and Ci firstly increased and then decreased with the increase of irrigation amount.The Ci values of T1 and T2 treatments were significantly different from those of other treatments,and the values of Pn,Tr,Gs and Ci were the highest at flowering stage and initial boll stage.The chlorophyll fluorescence parameters Yeild and qP first decreased and then increased with the change of growth stage,and the values of the five treatments were the lowest at flowering stage.The size of qN in the five growth stages under different water treatments was as follows:T5>T4>T3>T1>T2.Fv/Fm in different growth stages under different water treatments had a large difference,and both Fv/Fm and Fv/Fo reached the maximum value under T4 treatment at flowering stage.（3）Based on the estimation model constructed by PCR,PLSR and SVM,it is found that the SVM model has better prediction effect on physiological indicators,while the PCR model has the worst effect.Based on the SVM model,the model of Gs and spectral index DVI measured at full flowering stage in photosynthetic parameters was the best（R²=0.8041,RPD=2.5016）,the model of Fv/Fo measured at full flowering stage and spectral index RVI measured at chlorophyll fluorescence parameters was the best（R²=0.6374,RPD=2.0118）.However,the three models constructed by qP and qN do not have the ability to predict it.