Nitrogen Nutrition Index For Winter Wheat Under Different Irrigation Rates

High-efficient production of winter wheat has a direct impact on food security in China.Nitrogen nutrition index(NNI),as an important characterization of nitrogen deficiency,is of great significance in evaluating winter wheat growth,estimating yield and directing field management.Irrigation and fertilizer are two main management measures affecting the growth and development of winter wheat.However,there are two main methods to determination of critical nitrogen dilution curve(Nc)based on the coupling effect of water and fertilizer: One is to characterize all irrigation conditions directly with an Nc curve.Although this method is simple and convenient in field management,it cannot diagnose accurately.The other is multiple curves under different irrigation conditions.Although the Nc curves under different irrigation conditions have high accuracy,there are many problems in field management.In order to diagnose nitrogen in winter wheat under the coupling effect of water and fertilizer,this study discussed the correction of Nc under different irrigation conditions,remote sensing monitoring of agricultural parameters involved in NNI,and how to invert NNI using remote sensing information based on three consecutive years of water and fertilizer experiments.The main conclusions are as follows:(1)Under the same water treatment,with the development of growth period,aboveground dry biomass(AGB)showed a gradual upward trend.AGB ranged from 1.20 to 4.53 t hm-2,1.95 to 8.27 t hm-2 and 3.42 to 12.74 t hm-2 at jointing stage,booting stage and flowering stage,respectively.Under the same water treatment,plant nitrogen concentration(PNC)decreased with the development of growth period.The range of PNC at jointing stage,booting stage and flowering stage was 2.07-3.94%,1.36-2.85% and 1.01-2.10%,respectively.Under the same water treatment,the plant water content decreased gradually with the development of the growth period,and the difference of plant water content(PWC)was not obvious at the same growth period.The range of jointing stage,flag-picking stage and flowering stage were 0.75-0.85,0.73-0.84 and 0.59-0.78,respectively.(2)The critical diluted nitrogen concentration curves of winter wheat were all negative power type.The traditional Nc model without considering irrigation effects was Nc=4.91*AGB-0.53.The R2,n RMSE and RPD of the calibration model were 0.64,0.20 and 1.80,respectively.Traditional construction methods cannot accurately predict Nc under different irrigation conditions.(3)The accuracy of the corrected model of winter wheat with 1/(1-PWC)as the second layer information was the highest.The model expression was Nc = a AGB-b,where a = 1.33/(1-PWC)-1.88,b = 0.13/(1-PWC)-0.22,R2,n RMSE and RPD were 0.85,0.12 and 2.53,respectively.The n RMSE of the validated model was 0.15.The agricultural NNI model based on hierarchical linear model(HLM)also has high accuracy.The calibration models R2 and n RMSE were 0.74 and 0.10,respectively.The validation models n RMSE was 0.14.(4)Vegetation index has good correlation with AGB,PNC and PWC in single growth period,but poor correlation in multiple growth period.Compared with the traditional vegetation index,the vegetation index combined with growing degree days(GDD)had significantly improved the prediction accuracy of AGB,PNC and PWC in the whole growth period.In this paper,SRWI/GDD,TVI/GDD and NDVI/GDD were selected to construct the estimation models of AGB,PWC and PCN respectively.The predictive model expression of AGB was y=28.64*e-345.90 x,R2 and RMSE were 0.64 and 1.75 t hm-2,respectively,and the verification set RMSE was 1.32 t hm-2;the predictive model expression of PNC was y=0.17*e277.65 x,R2 and RMSE were 0.77 and 0.36%,and the verification set RMSE was 0.24%;the predictive model expression of PWC was y=41.27x+0.43,R2 and RMSE were 0.75 and 0.03,respectively,and the verification set RMSE was 0.03.(5)Compared with the four types of remote sensing monitoring models,the accuracy of the remote sensing monitoring model based on multi-layer linear model was the highest.The model expression is Nc=β1*TCARI/GDD+β2,where β1=210.94-299.79*NDVI/GDD,β2=-0.96+3397.91*NDVI/GDD.The calibration model R2 and n RMSE were 0.62 and 0.19 respectively,and the validation model n RMSE was 0.18.Although the other three types of Nc models also had high accuracy,the verification results were poor.However,the NNI monitoring model constructed by MTVI2 ha D the highest accuracy.The calibration model,NNI=2.02*MTVI2+0.60,R2 and n RMSE were 0.21 and 0.17 respectively,and the validation model n RMSE was 0.20.The accuracy of the other two indirect remote sensing monitoring models has not reached the ideal level.The NNI monitoring combined with remote sensing data needs further study.