Assessing Canopy Water Status Of Winter Wheat With Multiple Spectrometers Mounted On Unmanned Aerial Vehicle

It is of great significance that monitoring quickly and evaluating accurately crop water content(CWC),canopy temperature(CT),crop water stress index(CWSI),water deficit index(WDI),which are related closed wih crop water status,for gruiding scientific field water management.Through a variety of crop water statuses setted by arranging different irrigation modes,and based on the monitoring data of a multiple spectrometer mounted on unmanned aerial vehicle(UAV),this study analyzes and researches the spatio-temporal variation of crop spectral reflectance,canopy temperature,canopy equivalent water thickness(EWT),crop water stress index,and water deficit index,and their relationship with crop water status,in order to determine the suitable techniques and indicators for rapid and non-destructive assessment of canopy water status.For all experiments,UAV and destructive sampling was performed from March to May in 2019.Exp.1 was conducted in rain-out shelter facility using split-plot design with four irrigation amount(0,120,240,360 mm)and two nitrogen(N)levels(75 and 255 kg N ha^-1)treatments.In addition to the Exp.1,two field experiments(Exps.2-3)were conducted under field condition to compare different precision wheat water status.For Exp.2 and Exp.3,different rates of nitrogen fertilizer have been applied(0-300 kg ha^-1)in single plots without replication.Field measurements including destructive sampling and remote sensing data collection were conducted for specific growth stages.The main research results are as follows:1.Wheat water content are affected by water stress.Fields sampling,often used to evaluate the canopy water content(CWC)did not represent the assessment of crop water status spatial variability.The use of unmanned aerial remote sensing platform is capable to capture the variability of crop water stress in a whole field condition.The first objective of this study to develop machine learning models derived from multispectral and thermal images to predict the crop canopy water spatial variability of wheat fields.Statistical parameters such as the coefficient of determination(R²),root mean square error(RMSE),mean absolute error(MAE),nash sutcliffe efficiency(NSE)have been used to assess the relationship between inputs data,outputs models and ground-truth measurements of crop water measurement.The best performance of the model has been based on different inputs.Validation analysis have been performed on another field dataset using mean absolute error(MAE),root mean square error(RMSE)and relative error(RE).The finding results showed high accuracies performance in PWC modeling with ANN algorithms in the range of 0.786-0.955(R²),0.786-0.954(ENS),1.882-4.066%(RMSE),and 3.234-1.456%(MAE).While for DNN,they were in the range of 0.927-0.976(R²),0.927-0.9758(ENS),1.3689-2.386%(RMSE)and 1.828-0.977%(MAE).Also,these results indicated that DNN and ANN are very effective for predicting PWC when compared to linear regression models.However,the DNN model(R²=0.976,ENS=0.976,RMSE=1.369%,MAE=0.977)has a great capability for PWC modeling and outperforming ANN(R²=0.954,ENS=0.954,RMSE=1.882,MAE=1.456)and the stepwise regression model(SRM)(R²=0.665,ENS=-0.505,RMSE=10.793,MAE=9.034).These statistics values at the calibration stage confirmed the trained ANN and DNN networks are a powerful tool to predict PWC accurately.It is concluded that remote sensing type UAV measurements coupled with machine learning techniques are a promising and adequate tool to rapidly determine PWC.2.To investigate the relationship between vegetation indices derived from UAV multispectral for assessment,equivalent water thickness using linear and machine learning models.A feature selection(FS)algorithm named the decision tree(DT)was used as the automatic relevance determination method to obtain the relative relevance of vegetation indices(Vis),which were used for estimating EWT.The selected VIs was used to estimate EWT using multiple linear regression(MLR),deep neural network multilayer perceptron(DNN-MLP),artificial neural networks multilayer perceptron(ANN-MLP),boosted tree regression(BRT),and support vector machines(SVMs).The results show that the DNN-MLP with R²=0.934,NSE=0.933,RMSE=0.028 g cm^-2,and MAE of 0.017 g cm^-2 outperformed other ML algorithms(ANN-MPL,BRT,and SVM-Polynomial)owing to its high capacity for estimating EWT as compared to other ML methods.Our findings support the conclusion that ML can potentially be applied in combination with VIs for retrieving EWT.Despite the complexity of the ML models,the EWT map should help farmers by improving the real-time irrigation efficiency of wheat by quantifying field water content and addressing variability.3.Wheat water content was evaluated from rainfed and irrigated fields by using canopy temperature with different nitrogen and water levels.We compared the diferences between the CWSI and WDI values of winter wheat across diferent N and irrigation treatments,to examine the efect of crop N status on CWSI and WDI,and to recalibrate the CWSI and WDI values for better assessment of water status in winter wheat using nitrogen nutrition index.The result demonstrated that CWSI and WDI values decreased with the increasing irrigation rates within the same N treatments.However,both indices could not be used for assessing plant water status under diferent N treatments at the same I treatments.The increase in CWSI and WDI values with decreasing N rates indicated that there is a signifcant efect of plant N status on CWSI and WDI values.The pure canopy foliage temperature and mixed soil canopy temperature showed a signifcantly negative relationship with NNI across different N treatments.The recalibrated Tc and Ts values were lower than the original values under the N-limiting treatments.Using the recalibrated Tc and Ts value,the CWSI and WDI values were recalculated,which can better signify plant water status under diferent N conditions.