|
Key words:
Continuing advances in computing technology have made distributed numerical modeling of terrain-surface processes an attractive approach for many different applications in the Earth and environmental sciences. Dynamic landscape models are attractive not only for their potential predictive ability, but also for their role as "idea animators" that can simulate the consequences of simple physical rules when iterated over time and space. This paper presents a new hydro-geomorphic modeling system and discusses its application to diverse research problems in landscape evolution, flood forecasting, and archaeological site simulation. We also review some of the prospects and challenges in the future development of terrain modeling systems.
The CHILD model, which has roots in theoretical geomorphology, is a software system designed to model the flow of water and sediment across a topographic surface, and to simulate the resulting changes in the shape of the terrain itself. The model discretizes terrain using a triangulated irregular network (TIN), which has two important advantages: first, the variable spatial resolution capability means that certain landscape features (such as stream channels) can be represented at a locally high resolution; second, computational points can be added or deleted in real time in response to changing spatial variables as the system evolves. A deformable TIN mesh also allows for simulation of tectonic deformation. Several examples are presented which highlight potential applications in geomorphology, including gully erosion and alluvial fan development. The close connection between landscape evolution and the spatial distribution of prehistoric archaeological sites has also motivated the extension of the model to simulate archaeological site distributions in an active floodplain environment; results from this application suggest ways in which archaeological surveys can be "tuned" according to landscape position. Finally, application of the model to real-time flood forecasting illustrates both the potential for "generic" landscape modeling systems as well as some of the challenges that lie ahead.
Landscape models vary widely in the processes they represent, from forest fires and floods to mountain range evolution over millions of years. Yet most landscape models also share a number of common computational ingredients, including the discretization of a landscape surface into a set of connected elements, and the computation of fluxes of mass or energy between neighboring elements. As landscape models grow in sophistication, there is an increasing need for generalized modeling systems that, by providing basic spatial and temporal data structures and other capabilities, provide a canvas on which particular process algorithms or hypotheses can be applied. Developing such systems will involve challenges in both software engineering and in issues related to spatial scale, system complexity, and model validation.