| Land available for crop cultivation in Yangtze River Reaches(YR)is gradually decreasing due to growing population,urbanization and land degradation.Hence,increasing crop yield per unit area is the target for crop scientists.Excessive nitrogen(N)fertilizer has been used to achieve higher yield.N influences all aspects of crop growth and development,shaping yield throughout the growing season.N availability is an important factor in crop production,while its deficiency can cause a severe reduction in yield and economic returns to growers.Even being considered as an important factor in increasing rice yield and quality,excessive use of N fertilizer reduces N use efficiency and causes severe environmental threats.China's national average rate of N used in rice cultivation is about 70%higher than the world average.To curtail the damage to the environment caused by excessive N fertilizer application and to increase N use efficiency,precise N fertilizer management has become one of the burning issues in the modern agriculture,especially for rice production in YR of central China.An effective plant N management strategy at critical growth stages optimizes N management to minimize the environmental impact of the N while ensuring optimum N status of the crop for good product quality and maximum growth.Above ground plant biomass and N are important for plant growth monitoring,which has a decisive influence on N management.Destructive measurements of these parameters are time-consuming and labour-intensive.Non-destructive estimation by hyperspectral remote sensing may contribute to improving precise and timely N management.This dissertation mainly focuses on non-destructive monitoring of the key components of biomass such as leaf area index(LAI),canopy nitrogen(CN)and plant nitrogen(PN)level indicators estimation at critical growth stages using hyperspectral data from non-imaging field spectrometers.CN level indicators include canopy dry weight(CDW),canopy nitrogen concentration(CNC),and canopy nitrogen accumulation(CNA),and PN level indicators include plant dry weight(PDW),plant nitrogen concentration(PNC),and plant nitrogen accumulation(PNA).The main objectives of study were(?)to investigate the sensitive spectral regions of canopy spectral reflectance(CSR)for LAI,CN and PN level indicators over phenological stages of rice(?)to evaluate the potential of various hyperspectral vegetation indices(HVIs)for LAI estimation at phenological stages and(?)to establish the quantitative relationships for estimation of CN(CDW,CNC and CNA)and PN(PDW,PNC and PNA)level indicators with new HVIs based on sensitive spectral regions at phenological stages of rice.Multi-N rate field experiments were conducted using rice hybrid in Hubei province of central China.The N rates were varied from 0-293 kg N ha-1.During the experimental periods,phenological stage specific observations were made on the sensing and agronomic parameters like LAI,CN and PN level indicators.The main highlights of the study are as follows:(1)Our study demonstrated that CSR of rice characteristics sensitive reflectance varied in visible(530-700 nm),red edge(700-780 nm)and NIR(750-860 nm)regions according to the three phenological(elongation,booting and heading)stages.We found that sensitive spectral regions such as 524,534,583,687,707 nm and infrared>760 nm were prominent at three phenological stages for rice LAI estimation.The biomass,CN and PN level indicators at three phenological stages were more precisely identified in 400-720 nm and 560-710 nm,720-900 nm of CSR.Therefore,phenological stages have a significant effect on the relationship between the optical and biophysical and biochemical characteristics(LAI,CN,and PN level indicators)of the rice crop.(2)We found that some HVIs have the potential for crop LAI estimation at early-mid vegetative phase(elongation-booting)and others have potential to estimate the sensitivity of LAI over the reproductive phase(heading-maturity).The rapid growth period at early stages and loss of leaf area at later growth stages could be easily identified due to the performance of HVIs.Therefore,HVIs are capable of detecting changes in LAI sensitivity at three phenological stages under varied N fertilization rates.(3)Our study formulates the two new HVIs based on sensitive CSR regions for prediction of CN and PN level indicators at three phenological stages.We used newly constructed HVIs with other reported HVIs to develop dynamic models for CN and PN level indicators estimation.These dynamic models demonstrated the excellent performance for estimation of CDW and PDW at booting,CNC,CNA,PNC and PNA at heading phenological stage of rice.This study suggested that dynamic models have potential to estimate the dynamic CN and PN level indicators at different phenological stages with specific HVIs compared with the univariate non-linear regression with a number of HVIs.This study improves the understanding of hyperspectral remote estimation of LAI,CN and PN level indicators by considering the dynamical co-variations between plant biomass and N across different growth stages.Our study suggests that canopy hyperspectral reflectance of remote sensing have potential to improve the ability of precise N management in rice crop.Changes in the LAI and spectral sensitivity at three phenological stages evaluate the potential of different HVIs for LAI estimation.Additionally,this study indicates that identified sensitive spectral regions facilitate the construction of new HVIs for dynamic models.Dynamic models based on newly developed HVIs may better explain the variation in PN compared to CN level indicators due to a number of HVIs and issue of overfitting and co-linearity in bands.The methods developed in this thesis can be used to improve systems of field information extraction by canopy hyperspectral reflectance of remote sensing.Moreover,this study enhances our understanding of biophysical and biochemical characteristics estimation of rice crop for precision N management. |
PDF Full Text Request