Go to Paper
Return to GeoComputation 99 Index
An Object-Oriented Framework for Distributed Hydrologic and Geomorphic Modeling using Triangulated Irregular Networks
TUCKER, Greg (firstname.lastname@example.org), GASPARINI, Nicole, BRAS, Rafael, RYBARCZYK, Scott, Massachusetts Institute of Technology, Department of Civil and Environmental Engineering, Building 48, Room 429, Cambridge, MA 02139; LANCASTER, Stephen, Oregon State University, Department of Geosciences, Forest Sciences Laboratory, 3200 SW Jefferson Way, Corvallis, OR 97331
Key Words: simulation, model, geomorphology, runoff, erosion, sedimentation, software, Delaunay triangulation, TIN
In recent years, spatially distributed models of land surface processes, such as runoff and erosion, have come into widespread use in the Earth and environmental sciences. As these models grow in sophistication, the software engineering effort required to implement them also expands. Therein lies the need for portable, modular codes that can implement many of the basic requirements of a distributed model in a flexible, efficient, and application-independent manner. Here, we describe a simple prototype of such a system and its use in modeling long-term landscape evolution and short-term flood forecasting.
Distributed models of surface processes such as runoff, vegetation growth, soil erosion, forest fires, landscape evolution, and other processes typically share a number of important features in common: all involve (1) spatial division of terrain into discrete elements, (2) storage of mass and/or energy within landscape elements, (3) routing of flows of mass (e.g., water) and/or energy among landscape elements, (4) dynamic updating of boundary conditions (e.g., rainfall input), and (5) dynamic updating of state variables (e.g., soil moisture and surface elevation) through time. Often, the programming effort required to implement these features is non-trivial and quite labor intensive. This is particularly true when the underlying spatial representation is irregular; for example, the case of models based on triangulated irregular networks. Although current GIS systems provide sophisticated capabilities for spatial representation of data, performance and other limitations make them unsuitable for computationally intensive dynamic (i.e., time evolving) simulations; thus, to reduce software development times and minimize duplication of effort, it would be advantageous to develop application-independent modeling routines that would provide the underlying space and time structure for distributed models without dictating the processes or state variables.
We present a simple prototype of one such system. The system is an outgrowth of parallel ongoing efforts in (1) modeling rainfall, runoff, and streamflow for real-time flood forecasting, and (2) modeling long-term drainage basin evolution via uplift, erosion, and sedimentation. The system is based on a triangulated irregular network (TIN) representation of terrain using the Delaunay triangulation. The TIN representation is implemented via a set of C++ classes, which are in effect decoupled from the process routines. Efficient storage of mesh elements is accomplished by using a "dual edge" data structure, an adaptation of the well known Quad Edge structure for Delaunay triangulations. Capabilities of the system include the ability to construct triangulations from a given set of points, calculation of polygon areas and edge lengths, and the ability to move, add, or subtract points dynamically during a simulation. A class inheritance hierarchy is developed in order to enable new applications (such as models of watershed slope stability) to take advantage of existing code for more general processes (such as routing of overland flow across a topographic surface).
The system is illustrated through several examples, including simulation of long-term river meandering within a floodplain, and simulation of event-driven rainfall and runoff. We discuss some of the advantages and disadvantages of the modular, object-oriented approach in terms of system performance, ease of development, and related issues.