Characterizing and understanding grassland ecosystem functioning. From functional to optical plant traits

Worldwide, grasslands form important ecosystems, providing essential habitats to a wide variety of species. However, these ecosystems experience various pressures, such as climate change and plant invasion, potentially affecting their functioning and thus jeopardizing the services and benefits they provide to humanity. Grassland conservation and restoration initiatives are thus important, and various policy frameworks have been set up. In support of such programs, it is essential to understand how these ecosystems function so that, based on scientific insights, effective management practices can be implemented, and progress towards policy goals can be monitored. The concept of plant functional traits highly contributes to such understanding. In fact, plant functional traits, being the morphological, physiological, biochemical and phenological characteristics that determine a plant’s fitness and function in general, indicate how plant communities respond to pressures and management actions on the one hand, and determine how such modifications result in changes in the functioning of, and services provided by, the ecosystem. However, the use of functional traits is to a great extent constrained by its limited potential for generalization across time and space. Therefore, promising alternative, or at least complementary, approaches deserve further study. In this dissertation, we investigated the potential of hyperspectral remote sensing technology to measure functional traits, also referred to as “optical traits”, and advance our understanding of the dynamics in grassland ecosystems. The research consisted of two parts: in a first, methodological, part (Chapters 2 and 3) we aimed to provide and recommend technical tools that enable grassland optical trait measurements; in a second, applied, part (Chapters 4 and 5) our intention was to demonstrate how these optical traits can in turn be adopted to assess plant community functioning and address more conceptual questions at the forefront of functional ecological research.

Reflectance can be recorded from various platforms, with different spatial and spectral resolutions, and subsequent quantification of optical traits can be accomplished using various signal processing techniques, with different technical strengths and weaknesses. Driven by a lack of knowledge on the reliability of these approaches, we performed a global meta-analysis, summarizing trait estimation accuracies reported in 77 studies (Chapter 2). We found that most studies have focused on a few traits only (chlorophyll, carotenoid, phosphorus, nitrogen, LAI, water and lignin), and estimation accuracy was generally high (R² ranged between 0.64 and 0.80, nRMSE ranged between 0.09 and 0.26). Our findings supported the increasing use of multivariate signal processing because they generally performed better than univariate approaches. Moreover, we found that the upscaling of existing methods to airborne and satellite data is promising, and may allow for functional mapping at broader spatial scales. Despite these technical recommendations and encouraging outlook, in practice, spectral measurements of individual herbaceous species in the field turned out challenging, because these species generally have tiny leaves and grow in ecosystems with small scale heterogeneity. Such information is highly valuable for many ecological applications, policy targets and species mapping exercises. Therefore, we developed a novel in situ measurement procedure (Chapter 3). The procedure consists of measuring monospecific arrangements of plant individuals on a black, light absorbing table, as such preserving structural plant properties, while avoiding confounding effects of other species, soil or non-photosynthetically active vegetation. In a case study, we demonstrated that the procedure enables an accurate representation of spectral shape and amplitude, as well as functional trait differences between species.

Having clarified and advanced the technological and methodological capabilities of hyperspectral remote sensing for the quantification of grassland traits, this dissertation aimed at taking this know-how one step further to address two leading issues in functional and community ecology. The central idea was to deploy the combined strengths of functional traits and spectral reflectance, by integrating optical trait measurements in ecological analysis frameworks. First, we showed that emergent plant optical types (POTs), obtained through agglomerative hierarchical clustering of optical traits, are well suited to represent trait variation among locally co-occurring species (Chapter 4). Indeed, the resulting POTs better captured multidimensional trait variation among species than four commonly used pre-defined conventional plant functional types. Second, we demonstrated that optical traits can contribute to an enhanced understanding of the causal pathways of environmental and anthropogenic pressures on ecosystem functioning, more specifically by studying the case of plant invasion (Chapter 5). We focused on two functionally distinct species that are non-native and invasive in Belgium: the annual forb Impatiens glandulifera Royle, and the rhizomatous perennial forb Solidago gigantea Ait. We revealed that both invasive alien species (IAS) altered aboveground biomass (decrease and increase under I. glandulifera and S. gigantea respectively), litter stabilization (decrease under both IAS) and soil available phosphorus (increase under both IAS) through selection effects, rather than through decreasing the functional diversity of the community.

Together, our results indicate that hyperspectral remote sensing may lead to important insights into vegetation diversity and ecosystem dynamics. We propose that an interdisciplinary framework, coupling ecosystem functioning and remote sensing through optical traits, allows for a mechanistic understanding of ecological processes. The  presented concepts can be easily extended to study various ecological cutting-edge issues, e.g. by including explicit links to ecosystem services or studying other drivers of change such climate change and fertilization. Moreover, the developed concepts entail promising perspectives for upscaling to larger spatial scales.