Significant advances have been made that integrate landscape issues in forest-level models. These advanced models are designed to simulate and evaluate economic, ecological, and social goals that are included in the management of forests. The application of multiple-objective heuristics such as tabu search and simulated annealing, combined with remarkable advances in computing power, now allows us to explore highly complex management scenarios over long time horizons and over vast geographic scales. While the power of these decision support systems is highly appealing, and even intoxicating, we still face three sobering challenges on the path towards generating credible forecasts. First, advanced data acquisition and data management systems are needed to support these systems. Data management systems must have high storage capacity, be capable of rapid updates, and accommodate a seemingly endless demand for queries from customers, government agencies, and the public. Planning is an interdisciplinary, hierarchical process, and team members have different data demands, depending on where they fit in the hierarchy. Second, the models must be verified. Multiple-objective models have dozens of parameters, and when these are combined with random search techniques, they become difficult to understand and replicate. Thorough sensitivity analysis is needed to test model parameters, goal weights, and assumptions of uncertainty. Finally, our ability to formulate and run large-scale, long-term forecasting models often exceeds the scientific credibility of the data, especially for complex forest ecosystems. In the absence of critical thinking, such powerful models can become dangerous weapons.
Monitoring systems for parallel distributed data management systems
