The Hyperspectral Monitoring Of Agronomic Parameters In Winter Wheat Under Different Nitrogen Levels

Using hyperspectral technology to realize accurate and rapid monitoring of plant nitrogen content(PNC and aboveground dry biomass(AGDB)of winter wheat is of great significance for scientific and reasonable fertilization and accurate management according to crop growth condition.In order to explore the influence of different pretreatment methods on the precision of PNC and AGDB spectral models,and choice a preprocessing method which can improve the precision of the model.the study based on the experiment of nitrogen treatment,and analyzed the dynamic changes of canopy spectral characteristics and the correlation between AGDB and PNC in winter wheat from jointing stage to maturity stage under different nitrogen application conditions.Then established the spectral monitoring model based on SPA-MLR,and evaluated and compared the model.The results are as follows:(1)With the increase of nitrogen application,the spectral reflectance of the canopy decreased in the visible wavelength(400?720nm),but increased significantly in the near infrared(720?2500nm).With the development of the growth period,the reflectivity increased at 400?720nm,but increased at 740?110nm,then decreased,and the reflectivity at booting stage was the highest.The first order differential spectrum shows obvious peak at 730nm,and the variation of the first order differential spectrum is consistent with the original spectrum with the growth period.In addition,the red edge amplitude(Dr)and the red edge area(SDr)of the red edge position increased at first and then decreased with the growth period,but with the increase of nitrogen application,the difference of REP between the treatments was not obvious,the difference between REP and SDr increased first and then decreased(2)From jointing stage to mature stage,with the development of growth stage,PNC decreased with the same nitrogen application,while AGDB increased first and then decreased.With the increasing of nitrogen application,the PNC of AGDB increased,and the change of PNC was not obvious after nitrogen application exceeded a certain range.(3)The correlation between PNC and AGDB and spectral reflectivity can be increased in different degree by all pretreatment methods.The correlation between PNC and AGDB is improved obviously by first order differential(1-Der),second order differential(2-Der)and multivariate scatter correction(MSC)(4)Among the PNC and AGDB spectral monitoring models established by SPA-MLR,the 2-Der pretreatment model is the best one,and its modeling set(R2=0.7121,RMSE=0.113,RPD=1.856)and the verification set(R2=0.679,RMSE=0.121,RPD=1.773)are also the best models established by 2-Der pretreatment,and the modeling set(R2=0.862,RMSE=6.923,RPD=2.419)is the best.The accuracy of model verification is also the best(R2=0.776,RMSE=8.382,RPD=1.833).