Estimation Of Maize Growth Parameters And Yield Under Plastic Film Mulching From UAV-based Multispectral Images

In precision agriculture,the most important thing is to monitor crop growth conditions so that we can conduct effective field diagnosis and predict yield accurately.The biophysical parameters that indicate the status of field crops include the leaf area index(LAI),and leaf nitrogen content(LNC)and biomass.Unmanned Aerial System(UAS)can obtain the crop growth parameters quickly and non-destructively,therefore,it is helpful to field managers to make field management decisions and predict yield timely and accurately.In this study,we explored the response of multi-temporal and multi-spectral remote sensing images acquired by UAV(Unmanned Aerial Vehicle,UAV)to the growth stages of plastic film mulching maize and crop growth parameters such as LAI,LNC,and biomass in the Loess Plateau and we established the inversion model between vegetation indices(VIs)and maize yield from UAV-based multispectral images.The results show that combined with the crop phenology and the long-term vegetation indices we can realize crop monitoring and the estimation of maize yield using UAV multi-spectral remote sensing data.The main conclusions are as follows:1.A variety of vegetation indices obtained by UAV have a good regression relationship with LAI.Among them,the regression model based on the two-band enhanced vegetation index(EVI2)performed best.The coefficient of determination(R2)of linear regression model established in 2019 has a value as high as 0.85,and the relative root mean square error(RMSE)is 0.24;the R2 in 2020 is 0.84,and the RMSE= 0.23;the regression model established with all data based on 2019-2020 with a R20.85,and RMSE = 0.24.2.Compared the correlation between different vegetation indices and leaf nitrogen content(LNC),the DGCI has the best prediction performance,with R2 of 0.85 and RMSE = 1.45 g,the model can improve fertilizer management decisions.3.Constructing regression models between VIs and maize leaf biomass,aboveground biomass and total biomass respectively,it can be found that the model between VIs and leaf biomass achieved the best prediction.The regression model established by EVI2 and leaf biomass with R2 of 0.82 and 0.88,RMSE of 8.84 g and4.34 g in 2019 and 2020,respectively;the inversion model established by two-year VIs data of 2019 and 2020 with R2 of 0.78,RMSE = 8.83 g.4.In 2019,the regression model established by VIs acquired at a single growth period performed best when the maize of plastic film mulching entered the tasseling period.The inversion model established by EVI2 and yield with a R2 of 0.93,and the RMSE = 0.52 t /ha in 2019;When the maize without mulching entered the tasseling period the regression model effect established by VIs and yield is the best in 2020,and the inversion model R2 established by EVI2 and yield is 0.94,RMSE = 0.44 t/ha;when the maize under all treatments entered the reproductive growth period,the multiple statistical regression model established by multi-temporal VIs also showed good prediction results.The R2 value is 0.89 and 0.95,the RMSE is 0.49 t/ha and 0.34 t/ha in 2019 and 2020 respectively,and the multiple regression model established by the two-year data had a R2 of 0.87,and the RMSE = 0.60 t/ha.5.Establishing statistical regression models between different VIs obtained by UAV and yield at different growth stages separately,it can be found that the model of yield prediction had the best accuracy when maize in the reproductive stage.6.Based on the thermal infrared image obtained by UAV,the plastic film and bare soil could be identified,the average temperature difference was about 7℃.The thermal infrared image can also identify the canopy growth conditions caused by different plastic film mulching and nitrogen fertilization processing,and it is a potential method to study the differences of crop growth conditions under different management measures.