Forest leaf water content estimation using LiDAR and hyperspectral data

Abstract

Forest leaf water content (LWC) plays a key role in many physiological processes such as plant growth, photosynthesis, transpiration, and thermal regulation. The estimation of LWC and knowledge about its spatial variation can provide important information for assessing forest drought conditions and predicting future forest change associated with climate change. Recently it was proposed as one of the essential biodiversity variables (EBVs) to assess progress towards the Aichi Biodiversity Targets. Remote sensing data have been proved effective to provide a non-destructive, rapid and economical way for the estimation of LWC at various scales. At the leaf scale, terrestrial laser scanning (TLS) data enable vegetation information to be explored at small spatial scales (mm). The backscatter intensity value of the reflected backscatter signal recorded by TLS is, to some extent, insensitive to ambient light and atmospheric conditions and provides good spectral separability for detecting and classifying objects. At the individual canopy scale, TLS data are capable of providing vertical information of LWC within canopies. However, at the regional scale, unlike TLS, due to the large footprint size of airborne laser scanning (ALS), most laser beams hit multiple targets within the footprint. With insufficient prior knowledge about full hits from the same target, the true reflectance of partially hit individual targets cannot be unmixed because the target reflectance and the collision area between the laser beam and partial hits both contribute to the returned intensity. The thesis aimed to estimate LWC at the leaf, canopy and regional scales using LiDAR and/or hyperspectral data. The research in the thesis firstly presented a new method to eliminate the incidence angle effects on TLS intensity data. The results showed that LWC and its 3D spatial distribution can be estimated at a moderate accuracy using corrected TLS intensity data. Secondly, the thesis evaluated the partial hits and incidence angle effects on LWC estimation at the individual canopy scale. LWC vertical distribution within plant canopy can be retrieved from TLS. To quantify the impact of woody material on the estimation of LWC, the thesis subsequently investigated a new algorithm to classify foliar and woody materials using TLS data. An adaptive radius near-neighbor search algorithm was employed to extract both geometric and radiometric features. The findings suggested that an accurate classification between foliar and woody materials can be achieved in a mixed natural forest by using TLS. Then the research moved on to the regional scale with the emphasis on removing canopy structural and background effects on LWC estimation from airborne LiDAR and hyperspectral data using radiative transfer models. Prior knowledge about canopy cover and background reflectance using the integration of LiDAR and hyperspectral data was used in the inversion of radiative transfer models. The results showed that the structural and background effects could be effectively minimized and the accuracy of LWC estimation was significantly improved. This thesis explored the potential of LiDAR remote sensing alone for LWC estimation at the leaf and individual canopy scales, and the combination of LiDAR and hyperspectral data for LWC estimation at the regional scale. It demonstrates that the LWC distribution within individual leaves and the vertical distribution of LWC within vegetation canopy can be estimated using terrestrial LiDAR, and the prior information derived from LiDAR and hyperspectral data can significantly improve the estimation of LWC through the inversion of radiative transfer models at the regional scale. Using the calibration methods developed in this thesis, the approaches have the potential to be transferred to other sites. Global maps of LWC can be generated when satellites data such as the Global Ecosystem Dynamics Investigation (GEDI) LiDAR become operational.