Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Continuous technological advancements allow for the rapid development of drones, sensors that can be mounted on them, and software to process the collected data. More and more, this technology is also used in ecology to gather high-quality data over large areas, where this used to be very time and labour intensive. One of the many types of data that can be gathered this way is three-dimensional (3D) data (e.g. the height of the terrain and vegetation or the circumference of trees). This can provide insights in various applications such as nature conservation, modelling of erosion or flood risks, or research in climate change. Currently, two methods are mainly used to collect this type of data: Light Detection and Ranging (LiDAR) and Structure from Motion photogrammetry (SfM). Both methods return a point cloud as an output after data processing. These consist of a very large number of points, which each have their own position in space. Together, they are a virtual representation of the objects or terrain of interest. LiDAR sensors send out a large number of laser pulses themselves in short intervals, which can be reflected on the object of interest (e.g. trees or the soil), and can then be redetected. Based on the direction of the pulses and the time required to return to the sensor, the position of all reflected points can be determined. For the SfM method, multiple overlapping images are collected, under various angles. Specialized computer programs can recognise specific points which occur on multiple overlapping images, and can create a point cloud based on geometrical principles. This thesis has compared both methods over 1 ha forest to determine how well results from SfM can equal those from LiDAR in this setting. In general, the LiDAR method is considered most accurate, especially because the laser pulses can better penetrate through the vegetation down to the soil. The images gathered for SfM can however not ‘see’ that well through the canopy, and thus less information about the lower vegetation layers and the terrain can be derived. Nevertheless, LiDAR sensors are nowadays much more expensive than the cameras for SfM. The flight altitude, camera angle and image overlap of SfM were varied in the comparison. Furthermore, it was investigated to what extent the LiDAR point cloud could be used to determine the diameters of trees. The results showed that also in this study, LiDAR could best determine the terrain and lower vegetation. On the other hand, SfM point clouds had more points and detail at the top of the canopy or over unvegetated terrain. Besides, they also represented the terrain well, there where the terrain was best detected. Which method performs best, thus strongly depends on the application (for example, LiDAR for terrain coverage, or SfM for high detail at the top of the canopy or in case of little vegetation). For SfM, a lower flight altitude, a higher image overlap and a camera angle that included more of the sides of the object caused a larger number of points. A low altitude and more overlap improved the point elevations, whereas the camera angle especially improved the tree height. A rather limited number of tree diameters was well determined from the LiDAR cloud. One of the most important causes here was the need for a division of the point cloud in individual trees, which often did not happen correctly, with missing or even multiple stems.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

As ecosystem engineers, earthworms significantly impact soil properties, nutrient cycles, other soil biota (e.g., mites and springtails), and plants. They provide ecosystem services such as enhanced food production, flood mitigation, carbon sequestration and environmental education. Furthermore, earthworms serve as bioindicators, offering insights on soil health and quality. Conservation efforts are needed and require knowledge on their distribution, but traditional sampling methods are labour-intensive and can fail to detect all species present at a site. This makes Species Distribution Models (SDMs) valuable tools for inferring their distribution over large spatial extents. Combining models (ensemble modelling) may improve the quality of predictions, but use presence-absence models that are less preferable for earthworms. Presence-only methods like MaxEnt are thus particularly interesting, but these require the optimisation of several parameters. Combining models with different MaxEnt settings (‘within-algorithm ensemble modelling’) may thus improve predictions. In this study, I modelled earthworm distribution across Europe by creating 200 MaxEnt models per species, varying in background point selection, data thinning of training observations, and model complexity. Models performing well in block-cross validation were included in ensembles to generate suitability maps, later converted to binary presence-absence maps for further analysis. The results of this study show that the within-algorithm ensembles did not significantly outperform single MaxEnt models with default or species-optimised settings, which could be due to insufficiently independent evaluation data and excessive complexity of models with default settings. Furthermore, key factors influencing earthworm distribution were land cover, soil type and soil temperature seasonality. The output further showed that species richness was highest in the Iberian peninsula, around the Balkan Mountains and Carpathians, the north coast of Turkey and the Caucasus and Ural Mountains, although most species showed greater model extrapolation to the east of Europe. More specifically, epigeics were predicted to reach significantly higher latitudes than endogeics, indicating their greater cold resistance and acid tolerance. These maps provide testable hypotheses regarding species ecology. In conclusion, within-algorithm ensemble modelling could improve predictions of species distribution. Nevertheless, more research is needed, which could target more data-rich species groups (e.g., birds) to further test this hypothesis regarding within-algorithm ensembles put forward in this thesis.

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Climate change has an important influence on the geographical space in which species can live. As locations are becoming unsuitable, species will have to shift their geographical range, in general towards higher elevations and latitudes, following the moving isotherms. However, many species (especially forest plants) will not be able to track climate change, because they have too low dispersal rates or are living in fragmented habitats. Hence, they are at the risk of extinction. However, most studies do not integrate the microclimate when estimating species’ responses to climate change. In temperate forests, microclimate is buffered, so species in the understorey of forests experience overall cooler maximum temperatures and higher minimum temperatures, hence they have the potential to provide microrefugia for species threatened by climate change. In this way, species do not have to travel long distances to find new suitable habitats. Efforts have been made to model the microclimate in the future to estimate if decoupling, an increasing difference between macroclimate and microclimate temperatures, is the case, but they do not take the changing forest structure into account (both in time and space). This thesis seeks to partly fill this research gap by integrating changes in forest structural variables, such as canopy cover and canopy height, for three future climate scenarios (moderate, hot and hot & dry) in BRT models. These models aim to predict the mean and maximum microclimate temperature of the summer season in the current situation and the mean and the maximum microclimate temperature for Meerdaal forest under future conditions (period 2096 2100). The findings indicate that Meerdaal forest currently exhibits a buffering capacity for both studied temperature variables. The mean temperature offsets, the difference between the microclimate and the macroclimate, continue to be negative in the future scenarios, but the maximum temperature offset is unexpectedly positive (i.e. the microclimate is warmer than the macroclimate) for the hot scenario, whereas temperature cooling occurs under the driest climate scenario (hot & dry). These deviances from the expectations are mostly related to extrapolation issues. By recognizing the importance of microclimate, forest management should protect microclimatic areas using adapted techniques and minimize canopy disturbances. Despite its limitations, this study demonstrates the potential for integrating microclimate variables into future ecological research, highlighting their role in detecting microrefugia and enhancing species distribution models (SDMs).