The simulation of terrain and ecosystems based on environmental and climate parameters is a scientific and technical field fraught with numerous challenges. The challenge stems not only from the complexity of the underlying natural phenomena that shape the terrain, but also from the design of generation tools that must strike a balance between physical accuracy and precise user control to enable interactive editing. Existing techniques—whether procedural, data-driven, deep learning-based, or based on physical simulations—often focus separately on mountainous terrain such as large-scale dendritic mountain ranges, lowland ecosystems, or high-altitude glaciers. In contrast, the generation of landscapes at the interface of different environments—such as plains with complex and diverse forms of meandering river networks, or coastal regions with varied landscapes ranging from flooded deltas to steep, rocky shores—has received less attention in computer graphics. 
The scientific objectives of this dissertation are to realistically simulate and control landscape dynamics. More specifically, the goal is to simulate phenomena at interfaces and the interactions between natural phenomena as environmental climate parameters evolve; to enable the introduction of control mechanisms for generating realistic landscapes according to the artist’s or designer’s intent; and to overcome existing resolution limitations by extending generation models to large-scale scenes. From a strategic perspective, this thesis aims to strengthen a new research focus within the ARCHES team at LIRIS, which is firmly centered on the modeling and simulation of climatic effects in computer graphics.
Scientific Positioning and Challenges
The originality of this topic lies in seeking to push the boundaries of computer graphics simulation by pursuing synergy at the interface between computer graphics and other scientific fields such as geomorphology and botany.Our approach consists of combining appropriate physical models, introducing corrective terms into the equations, utilizing the parallel processing capabilities offered by graphics cards, and employing vector models capable of representing objects with a higher level of detail while minimizing memory usage.In terms of the phenomena studied, we seek to model realistic landscapes that can be characterized as lying at the interface between regions with different physical, biological, or environmental properties. The goal is to simulate coastal erosion—that is, at the interface between the sea and the land—to simulate vegetation at the boundaries of mountainous or desert terrain, or to generate glaciers at the snowline.