PhD defence Tawanda Gara

Quantitative remote sensing of essential biodiversity variables: understanding the role of canopy vertical heterogeneity

Tawanda Gara is a PhD student in the department of Natural Resources. His supervisor is prof.dr. A.K. Skidmore from the faculty of Geo-information Science and Earth Observation.

Understanding spatial and temporal dimension of leaf traits is key in monitoring ecosystem function, processes and services. Plant traits provides an insight in improved understanding of ecosystem services across biomes. Tracking changes in foliar nutrient content within the earth system is vital in assessing the effects and adaptation capacity of vegetation communities to climate change. Remote sensing provides a cost effective and practical means of charactering plants traits from spectra over large spatial extents.

We sought out to understand the role of vertical heterogeneity in leaf traits across canopy in estimating foliar traits using in-situ hyperspectral measurements and Sentinel-2. Results presented in this thesis demonstrated that leaf spectral reflectance mirror variation in trait content across canopy. Leaf spectral reflectance shifted to longer wavelengths in the 'red edge' spectrum (685 - 701 nm) in the order of lower > middle > upper canopy positions. Key wavebands that enhance leaf samples discrimination have been reported to be sensitive to variation in chlorophyll, EWT, N, carbon and SLA. These leaf traits exhibited significant variation across the canopy vertical profile.

Our results at field level showed that reflectance spectra of leaf samples collected from the lower canopy matched PROSPECT simulated reflectance spectra better compared to reflectance spectra measured from upper canopy across the growing season. Leaf chlorophyll and Equivalent Water Thickness for leaf samples collected from the lower canopy were retrieved with higher accuracy compared to leaf samples collected from the upper canopy. This observation imply that variation in leaf biochemistry and morphology through the canopy vertical profile potentially affects the performance of the PROSPECT model.

Results obtained using in-situ canopy hyperspectral measurements and simulated Sentinel-2 data showed that leaf-to-canopy upscaling approaches that consider the contribution of leaf traits from the exposed upper canopy layer together with the shaded middle canopy layer yield significantly (p < 0.05) lower error as well as high explained variance (R2 > 0.71) in the estimation of canopy leaf mass per area, nitrogen and carbon. At landscape level, canopy leaf mass per area, nitrogen and carbon estimated based on the weighted canopy expression yielded stronger correlations and higher prediction accuracy from Sentinel-2 MSI data compared to the top-of-canopy traits expression across all seasons. This observation imply that remote sensing instruments sense leaf traits beyond the sunlit upper canopy. These results have a strong implication in modelling leaf traits using remote sensing. We also demonstrated the capability of the newly launched Sentinel-2 to map seasonal changes in leaf traits at landscape level.

This thesis demonstrated the importance of vertical heterogeneity of leaf traits in estimating plants traits at leaf, canopy and landscape level. We showed that incorporating the leaf traits content of foliage material from the shaded canopy improves the estimation accuracy of plants traits at canopy and landscape level using in-situ hyperspectral measurements in the laboratory and Sentinel-2 multispectral data at field level. We also demonstrate that the performance of the PROSPECT model and retrieval of chlorophyll, equivalent water thickness and leaf mass per area is likely to be affected by the leaf biochemistry and morphological changes through the vertical canopy profile over the growing season. These results are important in canopy reflectance modelling and retrieval of canopy traits for various application ranging from forestry to agriculture.