| Water stress caused by soil water deficit in the field can greatly affect the crop growth and yield.Therefore,it’s not only improved crop yield by irrigating the water-deficient lands timely and accurately,but also improve water resource utilization efficiency,which is of great significance to set a suitable irrigation system in the arid regions of Northwest China.This is an important means to develop precision irrigation and smart agriculture by using continuous monitoring of crop soil moisture content and predicting soil moisture changes accurately.With the advantage of convenience,speed and a wide monitoring area,low-altitude unmanned aerial vehicle(UAV)remote sensing is widely used in various fields,especially in precision irrigation technology.In order to solve the problem of lacking measured spectral data of the UAV and monitored soil moisture discontinuously under bad weather conditions,this study conducted with summer maize with four irrigation treatments,three replicates,and a total of12 experimental fields.The four different water gradients were:0.95 gradient(95%-100%of the field capacity),0.80 gradient(80%of the field capacity),0.65 gradient(65%of the field capacity)and 0.5 gradient(50%of the field capacity).The high-definition images of the corn canopy were obtained by using the UAV platform equipped with multi-spectral sensor including 6 bands(490nm,550nm,680nm,720nm,800nm and 900nm),Then,the canopy spectral data were extracted by threshold segmentation,and calculated the solar zenith angle data at the same time.The measured canopy spectral data together with the solar zenith angle data were input into the Ambrals Kernel-Drive Model to obtain simulated spectral data of the corn canopy to analyze the simulation accuracy of each band.Meanwhile,nice vegetation indices as SIPI,SAVI,SRPI,ARVI,PSRI,m SR,RDVI,EVI and MCARI were established by using the measured and simulated reflectance,respectively.This study analyzed the simulation accuracy of each vegetation index,and compared the correlation that between measured vegetation indices and the simulated vegetation indices and soil moisture content at different depths.The feasibility of monitoring or predicting the soil moisture content at different depths were explored by establishing vegetation index with canopy simulation spectral sequences obtained through the Ambrals Kernel-Drive Model.The following results were obtained:(1)The measured canopy spectral reflectance extracted by threshold segmentation method and solar zenith Angle calculated by geographic information were substituted into the fitting parameters of Ambrals kernel driving model.The simulation accuracy of the four kernel combinations was similar,and RMSE values were also very small,but the RMSE values of Ross-thin and Li-Sparse were the smallest.Therefore,this paper chooses this nuclear combination as the best choice for Ambrals Kernel-Drive Model.It is feasible to simulate the spectral reflectance of summer maize canopy by Ambrals kernel driven model,and the simulation accuracy of all bands reaches a significant level.Among them,the simulation accuracy of 800nm near infrared band and 490nm blue band with 0.95 water gradient treatment is the highest,with R~2 reaching 0.594 and 0.572.It is feasible to obtain continuous canopy reflectance data from solar zenith Angle sequence by Ambrals Kernel-Drive Model.(2)The canopy reflectance of July 26,July 28,July 31 and August 2 at 13:00 was simulated by Ambrals Kernel-Drive Model with Rose-thin and Li-sparse cores,and the simulation accuracy was analyzed with corresponding measured reflectance,which is the same as conclusion(1).The reflectance of 490nm blue and 800nm near infrared band simulated by Ambrals nuclear drive model is more accurate than that of other bands,and the determination coefficients are R~2=0.610 and R~2=0.578,respectively.The results show that it is feasible to construct vegetation index by simulating canopy spectral reflectance.Among the four vegetation indices SIPI,SAVI,SRPI and ARVI,the accuracy of vegetation index SIPI is the highest,reaching 0.729.The vegetation index established by simulated reflectance and its extrapolation value can be used to monitor the surface soil moisture content,and the monitoring effect is the same as that established by measured data.The vegetation index with the best monitoring effect is SIPI(R~2=0.514).(3)The summer maize canopy reflectance sequence was obtained by Ambrals Kernel-Drive Model,and the vegetation index sequence was constructed based on it.The simulation accuracy of simulated vegetation index was analyzed,the monitoring accuracy of simulated vegetation index on soil moisture content at different depths was explored,and the change trend of soil moisture content at different depths was inverted and predicted,and irrigation warning was carried out.Nine vegetation indices were established based on simulated reflectance,and the highest simulation accuracy was vegetation index SIPI,R2=0.729.The simulation accuracy of vegetation index was related to the simulation accuracy of its band.The monitoring accuracy of simulated vegetation index for soil moisture content at different depths is similar to that of measured vegetation index.For soil moisture content at five different depths,the highest monitoring accuracy is vegetation index SIPI,whose monitoring accuracy is 0.602,0.512,0.344 and 0.271,respectively.The vegetation index SIPI sequence with high monitoring accuracy can be used to predict soil moisture content at different depths,and the lower limit of irrigation is set as the irrigation warning line,which has an early warning effect on the plots without water stress and the plots with mild water stress.There was no warning effect on moderate water stress and severe water stress. |
PDF Full Text Request