Mapping spatial variation of Foliar Nitrogen using hyperspectral remote sensing

Foliar nitrogen is a critical parameter in leaf photosynthesis, respiration and transpiration, which relates to canopy and stand-level traits such as light use efficiency, wood growth and net primary production. Nitrogen plays an important role in terrestrial ecosystem carbon dynamics, and is an important input parameter of ecosystem process models. Recently, foliar nitrogen content has been proposed as one of the key essential biodiversity variables (EBVs) for satellite monitoring of progress towards the Aichi Biodiversity Targets. Hyperspectral remote sensing provides a time-efficient and cost-effective solution to estimate foliar nitrogen. Empirical approaches have been the dominant when estimating nitrogen, and good accuracies achieved, though any empirical relationship so established is hard to be transferred across species, sites and sensors. Physically-based approach, or radiative transfer models (RTMs) simulate the process of radiation transfer based on physical laws, which is more robust and transferable. However, nitrogen has been considered to be impossible to be retrieved from fresh spectra by inversion of a leaf RTM model. Furthermore, at the canopy level, hyperspectral remote sensing of foliar nitrogen may be confounded by canopy structure which causes more variations in canopy bidirectional reflectance factor (BRF). The thesis aimed to retrieve foliar nitrogen at the leaf and canopy level using radiative transfer models and with hyperspectral remote sensing. The research in this thesis firstly evaluated the feasibility of retrieving nitrogen by inversion of a leaf radiative transfer model PROSPECT. The indirect estimation of nitrogen by inversion of leaf PROSPECT through links between nitrogen and leaf traits was investigated, and it was proved that nitrogen can be estimated at a moderate accuracy by inversion of the leaf PROSPECT model and the empirical relationship between nitrogen and leaf traits (e.g. leaf water, leaf mass per area and chlorophyll). Secondly, the thesis recalibrated the specific absorption coefficients of protein and cellulose + lignin, and proved that the recalibrated model can accurately simulate leaf reflectance and transmittance. Nitrogen can be estimated at a moderate accuracy by inversion of the recalibrated model. Then research moved on to the canopy level with emphasis on addressing canopy structural effects on nitrogen estimation. The thesis evaluated the feasibility of nitrogen retrieval from the canopy BRF in the full wavelengths between 425 - 2500 nm corrected for canopy structure effects. The results showed that nitrogen can still be detected from canopy BRF after correcting for canopy structure, and continuum wavelet analysis was shown to be an efficient method of revealing the nitrogen absorption features. The selected wavelet feature most sensitive to nitrogen variation were found to be correlated with water, dry matter and protein. Lastly, the thesis tested the capability of coupled leaf-canopy models in foliar nitrogen estimation. Global sensitivity analysis showed that the influence of nitrogen content on canopy reflectance is small but noticeable. Canopy nitrogen content was more accurately predicted than leaf nitrogen content. Maps of leaf and canopy nitrogen were generated for the study area, and spatial variations of nitrogen appeared consistent with ecological knowledge. This thesis demonstrates an approach of estimating foliar nitrogen at the leaf and canopy level using leaf and canopy radiative transfer models. The robustness and transferability of radiative transfer models suggests that the approaches might be transferred to other sites. The combination of spectral invariants theory and continuous wavelet analysis have been shown to be an efficient and robust way of hyperspectral remote sensing of canopy foliar nitrogen regardless of the confounding effects of canopy structure. Accurate estimation of foliar nitrogen is beneficial in a variety of ecological applications. Global maps of foliar nitrogen content could be generated when hyperspectral satellites such as EnMAP and HyspIRI become operational. In turn, such monitoring would support the development and demonstration of Essential Biodiversity Variables (EBVs), which are used for assessing biodiversity. Regional and global mapping of foliar nitrogen will further our understanding of the photosynthesis process, net primary productivity, and carbon dynamics. In a broader context, foliar nitrogen as an input to ecosystem process models is essential in understanding the nutrient cycle and climatic feedbacks