Research On Key Parameters Inversion And Growth Monitoring Of Corn At Different Fertility Stages Based On Multi-Source And Multi-Scale Data

The Leaf Area Index(LAI)and the Leaf Chlorophyll Content(LCC)are key parameters and core indicators for crops,which directly reflect their nutrition,health and growth status.Therefore,an accurate quantification and rapid estimation of LAI and LCC is of great significance for crop growth monitoring,yield prediction,ecological environment change analysis,and ecosystem assessment.The traditional methods of LAI and LCC calculation can only be conducted at a "point scale" accompanied by post event and destructive effects,make it difficult to meet the needs of modern high-tech agriculture with challenging demands such as real-time information,large-scale farming and efficiency.In recent years,remote sensing has developed into a viable technology to obtain LAI and LCC by its advantages of non-destructive and fast measurements.With the increasing number of operational remote sensing systems and associated inversion methods,particular attention can now be focused on the selection of the most suitable data source and the optimization of processing and evaluation methods.This reaserch focuses on the above mentioned issues,based on a quasi-synchronous experiment of "star-air-ground" joint observation,to obtain the LAI and LCC of corn canopies by multispectral Sentinel-2 data as well as UAV hyperspectral UAV data.Data acquisition was conducted at four days in 2022 on July 2,July 24,August 12 and September 4,during the respective stages of maize,namely trefoil,jointing,heading,and maturity.For the processing of the data suitable spectral bands were selected and the PROSAIL model and lookup table method were utilized.The key parameters of maize at different growth stages at various scales have been derived through quantitative inversion modeling of the recorded data.Thereby,the"star-air-ground" integrated joint monitoring of the growth information of the complete growth period of maize could be realized.The main results and conclusions are as follows:(1)The change law of growth indicators and spectral characteristics at different growth stages was obtained by interpreting the characteristics of maize growth indicators,analyzing the spectral characteristics of the canopy,and judging the nutrition,health and growth status of maize at different growth stages.The law shows that during the trefoil stage,the LAI and LCC are in a low-level state,with significant reflectance characteristics such as the "green peak" and "red shoulder".During the trefoil to jointing stage,the LAI and LCC rapidly increased,while the characteristic curve gradually decreased.During the jointing to heading stage,the LAI and LCC were at a high level,and the characteristic curve continued to decline,with obvious absorption characteristics such as "blue valley" and "red valley".During the heading to maturity stage,the LAI and LCC showed a gentle decrease and remained at a high level,while the characteristic curve begins to rise and remains unchanged.(2)By exploring the correlation relationship between LAI,LCC and the canopy spectrum,it was found that the wavelength bands λ=544nm,λ=658nm,λ=676nm,λ=762nm and λ=798nm in the sensitive range are significantly correlated.Thus,to construct a growth index estimation model,a characteristic response band X is used,to estimate and discuss the prediction results of the LAI and LCC in each growth period of the maize.The LAI prediction accuracy R2 ranges from 0.649 to 0.738,the RMSE and NRMSE range from 0.106 m2/m2 to 0.283 m2/m2 and 5.89%to 8.72%respectively.The LCC prediction accuracy R2 lies between 0.658～0.752,the and ranges from 2.603～3.405 μg/cm2 and 5.58%～9.93%respectively.The results confirm that the characteristic band λ can quickly and effectively estimate and predict the growth parameters of maize.(3)Analyzing the inversion results derived from different growth stages by using the normalized difference vegetation index(NDVI)for the establishment of a LAI inversion model for maize,we found a saturation effect when the LAI reached a higher level of>3.25 m2/m2.For this reason,we introduced adaptive coefficients and optimal selection factors to improve the NDVI to obtain an NDVI-k.That was then used to reconstructed the inversion model.Thereby,it was found that the inversion accuracy was improved in all four fertility stages,and most obviously at the heading and ripening stages.The R2 increased by 0.188 at the heading stage,the RMSE and NRMSE decreased by 0.164 m2/m2 and 3.88%respectively.At the maturity stage the R2 increased by 0.176,and the RMSE and NRMSE decreased by 0.171 m2/m2 and 3.72%respectively.The results further reveal that the optimized and reconstructed NDVI-k can effectively overcome the saturation defect of the NDVI and improve the inversion accuracy and prediction ability of the inversion model.(4)In the final step,we established a quantitative inversion model of the complete fertility time phase of maize LAI,compared the accuracy of the inversion,and discussed the advantages and disadvantages of the inversion effect.This is conducted by optimizing the band combinations of the spectral parameters,simplifying the parameters of the PROSAIL model and refining the cost matching function of the look-up table method.Finally,we got an optimum inversion of the recorded data and an advanced estimation method for the time phase scale.At the trefoil stage,it has the optimum LAI inversion when using the UAV hyperspectral data and the optimal recombined spectral parameter method(R2=0.852,RMSE=0.072 m2/m2,NRMSE=4.48%).At the jointing stage,based on Sentinel-2 multispectral data,the PROSAIL model provides the optimum inversion(R2=0.839,RMSE=0.137 m2/m2,NRMSE=4.88%),while at the heading stage,also based on Sentinel-2 multispectral data,the optimal recombined spectral parameter method turns out best(R2=0.894,RMSE=0.155 m2/m2,NRMSE=3.56%).At the mature stage,based on UAV hyperspectral data,the PROSAIL model method offers the optimum inversion(R2=0.847,RMSE=0.219 m2/m2,NRMSE=4.74%).Finally,the overall analysis shows that the inversion result of the heading-mature growth period performs superior as compared to that of trefoil-jointing growth period.Furthermore,the inversion result based on UAV hyperspectral data is improved towards Sentinel-2 multispectral data,and the inversion result using the optimized spectral parameter method outperforms the inversion results based on the PROSAIL model method.Therewith,the presented research provides a deeper theoretical basics accompanied with practical experience for the high-precision inversion of LAI and LCC.It offers reliable and precise crop information and thus,ultimately serves the development of precision agriculture and automated agricultural management.