Coniferous forests are important in the regulation of the Earth’s climate and thus continuous monitoring of these ecosystems is crucial to better understand potential responses to climate change. Optical remote sensing (RS) provides powerful methods for the estimation of essential climate variables and for global forest monitoring. However, coniferous forests represent challenging targets for RS methods, mainly due to structural features specific for coniferous trees (e.g. narrow needle leaves, shoot clumping) whose effects on the RS signal are not yet known or not yet fully understood.
Recognizing the need for a better adaptation of RS methods to such spatially heterogeneous and structurally complex canopies, this thesis contributes to improving the interpretation of the remotely sensed optical signal reflected from coniferous stands by focusing on specific knowledge gaps identified in the RS methods at different scales of the coniferous canopies. In addition, it explores the application of approaches that simplify the way the structural complexity of such an environment is tackled when using canopy-level radiative transfer approaches. Three main levels based on the identified gaps were defined for the analysis: (needle) leaf level (chapter 2 and 3); shoot level (chapter 4) and canopy level (chapter 5).
At leaf levelthis thesis contributes to minimizing the uncertainties and errors related to leaf optical measuring methods adapted for needle leaves. Although optical properties of coniferous leaves are extensively used in RS approaches (i.e. as input or as validation data), there is only a limited number of techniques available for measuring coniferous leaves. The first focus of this thesis was to review the shortcomings and uncertainties of such methods in order to identify application limits and potential improvements (chapter 2). A review showed that a more standardized measuring protocol was needed, for which measurement uncertainties and errors had to be identified, quantified and preferably removed or minimized. Thus, an experimental set-up improving the original method of Mesarch et al. (1999) was presented (chapter 3), which focused on analyzing uncertainties caused by the presence of the sample holder and by the multiple scattering triggered by both the shape of the specific needle cross-section, and the distance between the needles composing a sample. Results showed that both the sample holder and the multiple scattering, triggered specially by the shape of the non-flat cross section of the coniferous needle-leaves, had a non-negligible effect on the optical signal when measured using a standard spectroradiometer coupled to a single-beam integrating sphere and following the method suggested by Mesarch. Thus, approaches designed to measure optical properties of non-flat coniferous needle samples more comprehensively should take into account these effects in their current signal correction algorithms.
Needle clumping into shoots quickly transforms the optical signal making the description of the canopy radiative transfer a complex task and encouraging the search for simplified yet robust approaches. Thus, subsequent steps in this thesis focus on one such simplified approach, known as the recollision probability theory (“p-theory”), applied at two hierarchical levels, i.e., shoots (Chapter 4) and the whole canopy (Chapter 5).At shoot level, an empirical verification of the relationship between the photon recollision probability and a structural parameter called STAR was investigated. The approach allows upscaling needle albedo to shoot albedo and was previously theoretically tested only (chapter 4). For this analysis empirical optical measurements of Scots pine needles and shoots were used. Results showed that the approach works well for the VIS and SWIR spectral regions. However, it was less accurate for the NIR and also for sparse shoots (STAR <0.15) with an uneven distribution of photon–needle interactions and a larger influence of the twig bark.
Finally, accurate modelling of the reflectance signal at canopy levelfor coniferous canopies requires realistic representations of the forest stands, which in general implies a large number of input parameters and computationally demanding algorithms. Radiative transfer modelling based on the photon recollision probability offers an alternative for a simplified definition of the forest canopy structure. The performance of such approach for estimation of the leaf chlorophyll content from satellite imaging spectroscopy data acquired by the CHRIS-PROBA sensor was investigated. The approach was compared to a computationally more demanding one based on a detailed 3D structural description of a forest (chapter 5). For this purposes two canopy models, PARAS and DART, representing the first and second approach respectively, were used. Top-of-canopy bidirectional reflectance factors (BRF) were simulated for both models and used to calculate two optical indices, ANCB670–720 and ANMB670–720.Subsequently, the empirical relationships established between the optical indices and the needle-leaf chlorophyll content (Cab) were applied to the CHRIS-PROBA image of a Norway spruce forest stand to retrieve a map of Cab estimates. Results showed that for the spatial resolution of CHRIS-PROBA (17 m), the simpler model PARAS can be applied to retrieve plausible needle-leaf Cab estimates from satellite imaging spectroscopy data with less intensive model parameterization and reduced computational powerthan when using a model like DART. The ANMB670–720 optical indexwas more robust andresulted in a linear relationship between the Cab estimated by both models. This relationship showed, however, a systematic offset that is potentially caused by differences in the implementation of woody elements in each model or by a different parameterization of leaf optical properties. Thus, further investigation on the impact of parameterization differences related to the needle optical properties and the implementation of woody elements in such a model is recommended.