Remote Sensing Mechanisms And Methods Of Freezing Injury In Winter Oilseed Rape Using Multi-source Data

Amidst global climate change,agricultural meteorological disasters are on the increase,and this has affected agricultural production in China.Remote sensing technology is capable of providing near real time monitoring of meteorological disasters related to crops at a variety of spatial scales.Oilseed rape is one of the major sources of cooking oil.Cultivated in winter,oilseed crop growth and productivity has grossly been affected by freezing injury,one of the major agro-meteorological disasters in China.It is therefore of great significance to use remote sensing technology for the monitoring of oilseed rape freezing injury and estimation of damage extent.Data derived from such investigations would provide information on areas most prone to this disaster and can be used in guiding the design and implementation of crop protection measures.In view of the above,this study aims at characterizing oilseed rape freezing injury using hyperspectral data obtained from pot experiments,field experiments and multi-source remote sensing images.The winter oilseed rape freezing injury mechanisms were studied at a regional level with inversion models for crop biophysical parameters.The current study lays a theoretical foundation for monitoring crop freezing injury using hyperspectral remote sensing technology.The main research contents and results include the following:(1)First experiment to measure oilseed rape leaf reflectance during freezing injury.Hyperspectral reflectance of oilseed rape leaves in supercooling state,different states of freezing and post-thawing process were measured and chlorophyll content,water content,leaf temperature,and photosynthetic parameters were also synchronously measured through artificial simulated freezing injury process.Some small strips(approximately 1mm×7mm)were cut from the leaf samples to examine the cellular structure of normal and frozen leaves using light microscopy and transmission electron microscopy during freezing injury.? and D indices were calculated to quantitatively study the changes of hyperspectral reflectance during freezing injury.Increasing D values represents a decrease in reflectance magnitude,while decreasing ?represents a flattening of the spectral shape.The results showed that compared with the spectral reflectance of oilseed rape leaves under normal condition,the spectral reflectance of leaves in supercooling state had no significant changes,and the spectral reflectance of leaves in frozen for 1 hour state decreased greatly.The most significant change was the positions of the water absorption features in leaf reflectance spectra shift to longer wavelengths when leaf water was transformed into ice upon freezing.During freezing process,the reflectance in visible and near infrared spectrum area gradually decreased,and water absorption features band was gradually moving towards longer wavelengths.During thawing process,the leaf spectral reflectance of the water vapor absorption band gradually returned to the same as the normal leaf water vapor absorption wavelength position.Leaf spectral reflectance in the near infrared region as the thawing time increased gradually,and at that stage of post-thawing,the water absorption bands at 970 nm,1200 nm,1450 nm and 1940 nm virtually disappeared and the dry matter absorption features became more prominent,such as those for lignin,cellulose,starch and protein at 1690 nm,1900 nm,2130 nm and 2300 nm.(2)Investigating the potential of hyperspectral techniques for detecting leaves at the stages of freezing and post-thawing injury,and for quantifying the impacts of freezing injury on leaf water and pigment contents.The spectral reflectance obtained by indoor experiment were firstly smoothed using a Savitzky-Golay filter with 15 sample points and a second order polynomial to reduce the effect of noise on the reflectance.Then,the mixed effect model was used to find sensitive wavelengths which can reflect the changes of pigments content,water content and cell structure.Based on these selected wavelengths,principal component analysis and support vector machines were applied to raw reflectance,first and second derivatives,and inverse logarithmic reflectance to differentiate freezing and the different stages of post-thawing from the normal leaf state.The results showed that for the detection of freezing and post-thawed leaves,support vector machine model based on the second derivative spectra obtained the highest classification accuracies and the overall accuracy was more than 95.6%,the Kappa coefficient was greater than 0.91.The optimal narrowband ratio vegetation index(NBRVI)generated the highest predictive accuracy for changes in leaf water content,with a cross validated coefficient of determination(R2cv)of 0.85 and a cross validated root mean square error(RMSEcv)of 2.4161mg/cm2.Derivative spectral indices outperformed multivariate statistical methods for the estimation of changes in pigment contents.The highest accuracy was found between the optimal RVI and the change in carotenoids content(R2cv=0.70 and RMSEcv=0.0015 mg/cm2).The spectral domain 400-900 nm outperformed the full spectrum in the estimation of individual pigment contents,and hence this domain can be used to reduce redundancy and increase computational efficiency in future operational scenarios.Our findings indicate that hyperspectral remote sensing has considerable potential for characterizing freezing injury in oilseed rape,and this could form a basis for developing satellite remote sensing products for crop monitoring.(3)Monitoring of winter oilseed rape freezing injury was carried out in Jiangling County,Jingzhou City,Hubei province based on RS and GIS techniques.Firstly,winter oilseed rape planting area before winter was estimated using multi-temporal HJ-1A/B CCD images based on decision tree algorithm through the analysis of NDVI time series of winter oilseed rape and other ground objects(especially winter wheat),and therefore the spatial distibution of winter oilseed rape were obtained for multiple growthing seasons.The results indicated that the range of user's accuracies and producer's accuracies were 80.4%?95.56%and 82.56%?91.43%,respectively.The results were ideal,but completely distinguishing winter oilseed rape from winter wheat still need further exploration.Then,daily minimum temperature from regional automatic meteorological stations for winter oilseed rape wintering period(December to next early march)in 2013,2014 and 2015 growthing seasons(the number of meteorological stations were 114,122,and 121,respectively)in jingzhou city,hubei province were interpolated by the inverse distance weighting and ordinary kriging interpolation methods.The results showed that for the 2013,2014 and 2015 growing seasons,the accuracy using ordinary kriging interpolation were superior to the accuracy of the inverse distance weighted interpolation method and the interpolation accuracies were MAE = 0.4391? and RMSE = 0.6952?;MAE = 0.4689? and RMSE = 0.6663 ?;MAE = 0.3379? and RMSE = 0.4394?,respectively.Finally,according to the results of oilseed rape planting area and daily lowest temperature interpolation,monitoring of freezing injury of winter oilseed rape was performed based on the national standard for oilseed rape freezing injury in the study area.The results showed that winter oilseed rape suffered from light freezing injury for 2013 growing season,and freezing injury did not happen for 2014 and 2015 growing season on the basis of the national standards for oilseed rape freezing injury.(4)Estimation and Mapping leaf area index(LAI)and aboveground biomass(AGB)of winter oilseed rape using very high spatial resolution satellite data at Parcel Scale.In this study,a hybrid inversion method is developed to estimate LAI values of winter oilseed rape during growth using high spatial resolution optical satellite data covering a test site located in southeast China.Based on PROSAIL(coupling of PROSPECT and SAIL)simulation datasets,nine vegetation indices(VIs)were analyzed to identify the optimal independent variables for estimating LAI values.The optimal VIs were selected using curve fitting methods and the random forest algorithm.Hybrid inversion models were then built to determine the relationships between optimal simulated VIs and LAI values(generated by the PROSAIL model)using modeling methods,including curve fitting,k-nearest neighbor(kNN),and random forest regression(RFR).Finally,the mapping and estimation of winter oilseed rape LAI using reflectance obtained from Pleiades-IA,WorldView-3,SPOT-6,and WorldView-2 were implemented using the inversion method and the LAI estimation accuracy was validated using ground-measured datasets acquired during the 2014-2015 growing season.Our study indicates that based on the estimation results derived from different datasets,RFR is the optimal modeling algorithm amidst curve fitting and kNN with R2>0.954 and RMSE<0.218.Using the optimal VIs,the remote sensing-based mapping of winter oilseed rape LAI yielded an accuracy of R2 =0.520 and RMSE = 0.923.These results have demonstrated the potential operational applicability of the hybrid method proposed in this study for the mapping and retrieval of winter oilseed rape LAI values at field scales using multi-source and high spatial resolution optical remote sensing datasets.Our study therefore has significant implications for field crop monitoring at local scales,providing relevant data for agronomic practices and precision agriculture.Similarly,image features such as VIs and principal components(PCs)were extracted from satellite data over the whole season of winter oilseed rape growth.Curving fitting methods and machine learning methods including kNN and RFR were compared and the optimal features and models were used to map the temporal and spatial variability of winter oilseed rape AGB.The result of ten-fold cross-validation demonstrated that the RFR model based on the three optimal features(EVI,MNLI and PVI)produced the best estimation of winter oilseed rape AGB(R2cv = 0.698,RMSEcv = 0.287 kg/m2)when compared with models established by curve fitting and kNN.This study therefore proposes that selecting a few ideal image features reduces input data redundancy and computational time while being sufficient for AGB estimation and mapping at a field scale,which would support future operational scenarios in precission agricultural applications.