| China is a big country of rice and rice planting area is in the forefront of the world.Rice is an important food for Chinese people.However,the acceleration of urbanization is gradually reducing the arable land area,and China is one of the countries suffering from salinization.The salinization of land further reduces the area of conventional available arable land,and more and more studies on rice cultivation in coastal salinization areas are conducted.Nitrogen nutrition index(NNl)is a relatively accurate index to determine the status of nitrogen in crops.It specific refers to crop aboveground plant actual nitrogen concentration and the ratio of critical nitrogen concentration.Its characteristic is scientific and accurate.However,field investigation,sampling and analysis in the laboratory are required in the practical application process.The cost is high and relatively complex,which has certain limitations and is not fast and real·time enough.Therefore,a more real-time and reliable monitoring technology is needed.In actual rice production,the final yield is one of the most important indices.Rice cultivation in saline soil area has not been developed for a long time and its yield still has great potential.To accurately estimate the spatial variation of yield and to search for low-yielding plots is helpful to explore the constraints of rice yield,so as to further improve the yield.Traditional rice production measurement relies on a large number of field samples.It is costly,time-consuming,and requires extensive laboratory measurements.Therefore,a rapid and reliable monitoring technique is also needed.With the rapid development and wide application of remote sensing technology,it provides a more convenient,effective and scientific method for the diagnosis of nitrogen nutrient level and yield estimation of rice in field.In this paper,the reflectance of red(R),green(G)s blue(B)and near-infrared(NIR)bands in the north rice reclamation area of Tiaozini in Dongtai county,Yancheng city,Jiangsu province was extracted by using the WFV image of GF-1 satellite.Through the combination of vegetation index,the nitrogen nutrition index estimation model and yield prediction model suitable for this region in different periods were established,which could guide the diagnosis of nitrogen nutrition level and yield estimation of rice in this region.The main conclusions are as follows(1)Comparison of common vegetation index and the relation between single-band reflectance and nitrogen nutrition index at tillering stage showed that the best correlation between nitrogen nutrition index and tillering stage was in near-infrared band,the regression model R2 reached 0.7038,RMSE=0.07439,RRMSE=7.653%.NIR/BNDVI had the best relationship with NNI after the combination of near-infrared bands with various common vegetation indices in this period.The R2 of NIR/BNDVI and nitrogen nutrition index model reached 0.748,RMSE=0.0707 and RRMSE=7.274%.Compared with the single-band model,the correlation between NIR and nitrogen nutrition index was significantly improved and the error was reduced.The model can be used to diagnose the nitrogen nutrition level of rice at tillering stage after the model test.By comparing common vegetation index and the relation between single-band reflectance and nitrogen nutrition index at booting stage,the best correlation between n nutrition index and the booting stage was in the near infrared band,and the R2 of its regression model reached 0.6728,RMSE=0.0410 and RRMSE=6.336%.NIR/BNDVI had the best relationship with NNI after the combination of near-infrared band and various common vegetation indices in this period,and its regression model R2 with nitrogen nutrition index reached 0.7062,RMSE=0.0388,and RRMSE=6.004%,which significantly improved the correlation between NIR/BNDVI and single band and reduced the error.After the model test,it was verified that the model could guide the diagnosis of nitrogen nutrition level in rice booting stage in this region.Compared with the previous two stages,the correlation between the mature stage and rice nitrogen was significantly weakened,which could not reflect the nitrogen level of rice.(2)By comparing common vegetation index and the relation between single-band reflectance and yield at booting stage,the best correlation between booting stage and yield was green light band,and the R2 of its regression model with yield reached 0.7958,RMSE=89.7554,RRMSE=24.273%.The relationship between G/GNDVI and yield was the best after the combination of green light band with various common vegetation indices in this period.The R2 of its regression model with yield reached 0.8289,RMSE=82.1485kg/mu,and RRMSE=22.215%.Compared with single band,the correlation is improved and the error is reduced.After testing the model,it is verified that the model can be used to estimate the yield.By comparing common vegetation index and the relation between single-band reflectance and yield at maturity stage,the vegetation index with the best correlation between the maturity stage and yield was RVI,and the R2 of its regression model with yield reached 0.7458,RMSE=100.1428,and RRMSE=27.082%.RVI-NDVI had the best correlation with yield after combining RVI with various common vegetation indices in this period,and the R2 of its regression model with yield reached 0.7509,RMSE=99.1247kg/mu,and RRMSE=26.806%.Compared with the single vegetation index,the correlation with yield was slightly improved and the error was slightly reduced.After testing the model,it is verified that the model can be used to estimate the yield.The relation between the yield and the common vegetation index and the reflectance of single band at tillering stage was relatively poor,which could not be used to estimate the final yield. |
PDF Full Text Request