Towards More Accurate Species Distribution Modeling Through Enhanced Methods And Data Usage
Loading...
Files
Date
Authors
Publisher
Philipps-Universität Marburg
Abstract
The biodiversity crisis, which is recognized as one of the major challenges of our time, highlights the urgent need for effective nature conservation planning. With increasing rates of species extinction and habitat loss, the development of effective mitigation strategies has become a priority for researchers, policymakers, and conservationists alike. To combat this crisis, Species Dsitribution Modeling (SDM) is widely employed in research and conservation practice. SDM involves upscaling species occurrence points with environmental data to create area-wide maps of species distribution, which are used to support the identification of areas in need of protection, the monitoring of invasive species, the assessment of environmental niche dynamics under different climate change scenarios, and the identification of occurrence sites of rare species, among other applications.
However, despite its widespread use, SDM is not without challenges. The performance of species distribution models is influenced by the chosen model algorithms, parameter settings, and validation and testing strategy. Although these topics are often discussed in the literature, important recommendations for SDM, such as implementing spatial cross-validation strategies or fine-tuning model parameters for each species, are frequently overlooked. Another great challenge in SDM is the availability and quality of input data, including species occurrence records and environmental variables. Obtaining comprehensive and reliable data can be difficult, as field surveys are often limited by time and financial constraints. While globally available environmental variables, such as climate data, are commonly preferred for SDM, it remains uncertain whether they are also the most suitable variables.
Therefore, the overall aim of this thesis is to enhance SDM by developing advancements in methodology, and optimizing data usage and acquisition. These efforts can be divided into two main components: firstly, the refinement of methods, with a key focus on the development of a software extension for the popular and widely used SDM software Maxent. For this purpose, the software extension spatialMaxent is developed, which integrates spatial validation and tuning procedures. spatialMaxent improves model performance compared to traditionally trained Maxent models and is accessible for users across various expertise levels, as the software extension can be operated via a Graphical User Interface (GUI). Additionally, the usability of the software beyond nature conservation purposes is demonstrated by mapping tick attachment to humans in Switzerland to assist healthcare management.
Secondly, the thesis introduces innovative data usage and acquisition methods for SDM. To achieve this, an automated sensor network for biodiversity monitoring is developed, capable of automatically capturing species occurrences (such as bird species through audio recordings). Furthermore, a workflow for processing high-resolution optical satellite data is presented, and the value of heterogeneous LiDAR data for environmental modeling demonstrated. These detailed analysis-ready variables demonstrate greater value for modeling than widely available generic datasets, as demonstrated by their application in modeling endangered forest dwelling bat species using spatialMaxent.
In conclusion, this thesis improves SDM by introducing enhanced methodologies, as well as more reliable and detailed approaches to data usage and acquisition. These combined advancements improve the overall performance of SDMs, providing solutions that are beneficial for effective conservation planning and therefore, mitigating the biodiversity crisis.