江西农业大学学报2024，Vol.46Issue(2)：367-378,12.DOI:10.3724/aauj.2024033
银杏雄花序光谱特征时序变化分析及氮含量估测
Temporal evolution analysis of spectral characteristics in male ginkgo inflorescences and estimation of nitrogen content
摘要
Abstract
[Objective]Nitrogen is a key element that significantly impacts the growth and development of ginkgo trees.As a dynamically active physiological organ,the nitrogen content in male inflorescences of ginkgo trees has a direct impact on its fertilization and senescence processes.However,previous research has primarily focused on the spectral estimation of nitrogen content in leaves of horticultural crops like fruit trees.Few studies have utilized spectral techniques to monitor the nitrogen content of ginkgo male inflorescences.Therefore,the objective of this study is to establish an optimal estimation model for nitrogen content in ginkgo male inflorescences,thus enabling rapid and non-invasive monitoring of nitrogen content fluctuations.[Method]This study focused on the male inflorescences of ginkgo trees,quantitatively analyzing the dynamic changes of nitrogen content in the inflorescences over time.A portable spectrometer was used to simultaneously acquire high-resolution spectral reflectance of the inflorescences,and nitrogen-related spectral indices were extracted.Through correlation analysis and the All-subsets Regression(ASR)algorithm,target spectral indices were selected.Subsequently,models were developed based on Multiple Linear Regression(MLR)and Random Forest Regression(RFR),and the Akaike Information Criterion(AIC)was employed to assess the fit of the MLR model.Finally,the accuracy of the models was tested using leave-one-out cross-validation.[Result]The results indicate that the nitrogen content in ginkgo male inflorescences exhibited a gradual decreasing trend over time.Among numerous spectral indices,sensitivity analysis and the All-subsets Regression algorithm identified the spectral indices NI_Ferwerda,MCARI/OSAVI,and TCARI as being particularly sensitive to nitrogen content in ginkgo male inflorescences(adjusted R2=0.73).The performance of the Multiple Linear Regression model based on these selected spectral parameters(CV-R2=0.70)generally outperformed the Random Forest Regression model(CV-R2=0.63).[Conclusion]The Multiple Linear Regression model constructed based on the selected spectral indices demonstrated a favorable estimation performance,accurately predicting the nitrogen content in ginkgo male inflorescences.Furthermore,this study offers theoretical and technical supports for the effective monitoring of nitrogen nutrition in the male inflorescences of ginkgo trees.关键词
银杏/氮含量/光谱特征/时序光谱特征Key words
Ginkgo/nitrogen content/spectral characteristics/time-series spectral characteristics分类
农业科技引用本文复制引用
裘赛铤,朱玉婷,周凯..银杏雄花序光谱特征时序变化分析及氮含量估测[J].江西农业大学学报,2024,46(2):367-378,12.基金项目
国家自然科学基金项目(32101521) Project supported by the National Natural Science Foundation of China(32101521) （32101521）
