Construction And Application Of A Precise Nitrogen Fertilizer Topdressing Model For Winter Rapeseed Based On Multispectral UAV

Winter oilseed rape is the most common oil crop in China,nitrogen is one of the most important nutrient elements for its growth.Proper application of nitrogen fertilizer is beneficial to improve yield and quality of oilseed,excessive use of nitrogen fertilizer will reduce fertilizer utilization efficiency and cause potential negative impact to the environment.In recent years,precision agriculture has become the main direction of modern agricultural development,precise nitrogen fertilizer application method with low cost and easy to use can improve the nitrogen fertilizer utilization efficiency of oilseed rape and reduce negative environmental effects.In recent years,UAV-based remote sensing platforms have emerged.Multispectral field images can be obtained through these platforms real-time and rapidly.Field experiments on nitrogen fertilizer and oilseed-specialized fertilizer application rates were conducted in Wuxue city and Shayang county,Hubei Province during 2014-2015 and 2020-2021.To solve the problem of excessive nitrogen fertilizer application,during the experiment,nitrogen concentration（NC）and above-ground biomass（AB）of the plant were determined,corresponding in-situ hyperspectral data and unmanned aerial vehicle（UAV）multispectral image data were also obtained.The NC and AB estimation model were constructed based on data above.Combined with the nitrogen dilution curve of direct-seeded winter oilseed rape,the nitrogen nutrition index（NNI）was calculated to evaluate the nitrogen status during the critical growth periods of oilseed rape,and a precise nitrogen fertilization model was constructed for direct-seeded winter oilseed rape.To facilitate the application of the fertilization model,a winter oilseed rape nitrogen fertilizer precise fertilization software based on multispectral UAV was developed using Qt and QGIS.To verify the precision fertilization model for direct-seeded winter oilseed rape,a variable fertilization experiment was conducted in Shayang County,Hubei Province in 2021-2022.The specific conclusions are as follows:（1）The multispectral reflectance of oilseed rape canopy with different nitrogen treatments.By calibrating the oilseed rape multispectral reflectance with near-infrared reflectance of vegetation（NIR_V）,the canopy scattering factor（CSC）can be obtained,which can reduce the difference in CSC in the near-infrared wavelength band under different nitrogen application levels and amplify the difference in the red and red-edge wavelength bands,which means that CSC can be used to estimate the nutrition status of oilseed rape more effectively.（2）The optimal method for estimating the nitrogen concentration（NC）of oilseed rape is XGBoost regression model.This model takes CSC as input.The model achieves an R² value of 0.8105 and an RMSE value as low as 0.4366%in the validation set.The optimal method for estimating the above-ground biomass（AB）is through a quadratic polynomial regression model constructed using the CI_{Red Edge} vegetation index.This model achieves an R² value of 0.9377 and an RMSE value of 440.2 kg/ha in the validation set.（3）A nitrogen fertilizer topdressing calculation model were constructed based on the nitrogen dilution curve of winter oilseed rape in the Yangtze River Basin and nitrogen nutrition diagnosis model constructed in this study.A winter oilseed rape nitrogen fertilizer precision topdressing software based on multispectral UAV imagery was preliminarily implemented using the above models,the Qt UI framework,and the QGIS open-source GIS library.In the field experiments,nitrogen fertilizer was applied during the wintering period and budding stage with this software.The result shows that the fertilization methods commonly used by farmers cannot meet the needs of rapeseed,the yield and fertilizer contribution rate of the spectral diagnosis topdressing treatment were close to those of the conventional topdressing treatment,but the total nitrogen application was reduced by 10.3%.This indicates that the winter oilseed rape nitrogen fertilizer precision topdressing technology based on multispectral UAV imagery proposed in this study can effectively reduce nitrogen fertilizer input without significantly affecting the yield of oilseed rape.This study is based on unmanned aerial vehicle（UAV）multispectral images,using NIRV to reduce the impact of canopy structural changes on multispectral reflectance characteristics.Based on vegetation indices and machine learning models,a nitrogen nutrition diagnosis model for rapeseed was developed.On this basis,a nitrogen fertilizer recommendation and topdressing calculation model was established according to the nitrogen dilution curve of winter oilseed rape in the Yangtze River Basin.To facilitate model application,software for making nitrogen fertilizer recommendation based on UAV multispectral images was developed,providing a new method for promoting more rational nitrogen fertilizer application in oilseed production.