Using Hyperspectral Remote Sensing Of Foliar Chemicals To Predict The Quality Of Tea (Camellia Sinensis)

Objective, precise and real-time information of vegetation can help both land users and land resource managers to acquire deeper understanding of the land productivity at present and in the future, which is conductive to making rational and precise land resource management policies. Therefore, improved quantification and monitoring of biophysical and biochemical attributes of vegetation in regional scale and eventually global scale have a great importance in land resource management. In contrast to the traditional methods available for detecting vegetation properties, remote sensing is widely viewed as a time and cost-efficient way to monitor various biochemical and biophysical vegetation variables, because of its wide coverage, repetitiveness and non-destructive characterization.Developments in hyperspectral remote sensing have provided new possibilities for estimation of vegetation attributes and particularly foliar biochemicals, for the narrow spectral channels in hyperspectral techniques make it possible to detect subtle changes in narrow absorption features caused by the biochemical characteristics of the object, while in traditional broadband remote sensing, some critical information in specific narrowband may be lost. Today, much research has been undertaken to estimate the chemical composition of plants using reflectance spectroscopy, in the fields of grass, savanna trees, agricultural crops and forests. However, often these studies have focused on the detection of chlorophyll and nutrients such as nitrogen. Although some studies have been carried out on the estimation of other biochemical constituents of vegetation, the use of high spectral resolution data for estimating the biochemical parameters of vegetation needs to be explored.This study aimed to detect the potential of hyperspectral remote sensing for predicting the biochemical compounds (total tea polyphenols, free amino acids and soluble sugars) as quality indicators of tea (Camellia sinensis). From dried and ground leaves, via fresh whole leaves to canopies, the complexity in spectroscopy typically increases, which complicates the detection of biochemical absorption features at the canopy level. Thus, the experiments were carried out firstly for tea powders and fresh leaves in controlled laboratory conditions, and then scaled up to the canopy level for living tea plants, as a way to explore the possibility of using airborne or spaceborne imaging spectrometer data to detect the quality-related biochemicals of tea in the future. In order to obtain generalized result, the spectral-chemical relationships were analyzed for different tea varieties. A number of methods to measure the biochemical concentrations of tea using hyperspectral techniques have been demonstrated in this study.The results show that the concentration of biochemical components important for tea quality (total tea polyphenols, free amino acids and soluble sugars) can be successfully predicted from tea powder spectra, fresh leaf spectra and also canopy spectra of living plants (for one tea variety or across different tea varieties). With the regard that canopy spectral data was collected in the field under natural environmental conditions, the success of retrieving foliar biochemical concentrations of tea in this study has shown that there is a potential to use imaging spectrometer to predict in situ tea garden quality at a landscape scale. By comparing retrieving ability at different levels, an important finding in this study is that canopy spectra may have more potential than leaf level spectra to predict the foliar biochemical concentration of tea. However, more investigations are needed to confirm this result. By comparing the results using different methodologies, both partial least square regression and artificial neural networks based predictive models worked well for tea foliar biochemical predictions at canopy level. This study also shows that soil nutrients conditions affect the spectral signatures of one green tea varieties, grown in greenhouse. And the foliar chemical variation as a response to different soil nutrient levels can be detected from canopy spectra using the integrated approach that involves genetic algorithms and artificial neural networks.This study makes a contribution in the domain of foliar biochemical information extraction from hyperspectral data. This non-destructive and objective method may be used at canopy level with airborne/spaceborne remote sensing, which is a step towards regional tea quality monitoring. Despite the success at leaf and canopy level reported in this study, the use of imaging spectroscopy for predicting foliar chemistry of tea still faces a great deal of challenges. More efforts and investigations are encouraged in the future work, such as developing algorithms to minimize the effects of canopy structure, sensor geometry, soil background and atmosphere on canopy spectra, or developing hybrid approaches by linking empirical and physically based models for predicting foliar biochemistry.