Exploring the Behavior of the Spring Ecosystem Model Using An Object-Oriented Database System

Landscape and ecosystem models have typically been viewed as “black boxes”, that given a set of inputs yield a set of outputs. This view does not easily lend itself to investigations as to why a model behaves as it does, especially when multiple models are coupled together to create a larger model. We hypothesize that, for this type of investigation, a visual multimedia tool is necessary to gain insight into the temporal behavior of the model. In order to facilitate the exploration of the behavior of one particular well-known landscape model, we have coupled the model with an object-oriented database system and a data plotting and visualization package. Using this system, we have created animations of the model’s output and used them to discover interesting properties of the model. In our demonstration, we will display these animations and discuss how such a system can be used as an aid to explore hypotheses about the model’s behavior. 1. System Description The EcoProbe system is composed of four parts: the Spring terrestrial ecosystem model, the ADAMS objectoriented database language system, the EcoProbe application interface code, and the data visualization tools. The system was built primarily using C/C++, and runs under Solaris. 1.1. The Spring Terrestrial Productivity Model The Spring terrestrial productivity model is described in detail by Woodward, et al. [15], and was derived from the DOLY model highlighted by Baskin [I]. Briefly, the model represents the biochemical processes of photosynthesis and the dependence of gas exchange on stomatal conductance, which in turn is dependent on temperature and moisture. Separate modules within the Spring model represent these biochemical processes on cellular, leaf-level, plant-level, and canopy-level scales. The long-term average annual carbon and hydrologic budgets dictated canopy leaf area. Finally, the model allocates carbon and estimates respiration consistent with this leaf area to derive net primary productivity. Woodward, et al. demonstrated that this model, based on fundamental processes of photosynthesis and leaf physiology, can be generalized sufficiently to allow realistic global simulations, and numerous models now follow this approach. The Spring model assumes that responses to environmental change are local and independent of spatial effects. Furthermore, Spring assumes that responses to environmental change are isolated on an annual time scale. Changes in the landscape as indicated by the model over the course of a year are not considered by the model for subsequent years. This has some interesting implications, as the model may indicate that a region has a high leaf area index value one year, and a greatly reduced LAI the next. While we might expect a year of decreased rainfall or unusual temperatures to have some effect on LAI, we would not expect to see dramatic year-by-year changes, but rather we would probably expect a smoother change. This is one aspect of the model that we wished to explore further. Originally, Spring was designed to accept input parameters as ASCII via s t d i n , and produce text and numeric output to s t d o u t . This is typical of the “black box”mode1 approach; one which does not lend itself to querying the data or examining intermediate values of variables within the model. By coupling Spring to a database system, we created a tool that would allow us to gain insight into the workings and interactions of the model components that we would not otherwise get by searching through source code or looking at columns of numeric output. 267 0-7695-1218-6/01 $10.00