Incorporating Complexity and Uncertainty in Planning Models for Engineering: Economic Analysis within the U.S. Army Corps of Engineers

摘要

Established by Congress in 1802, the U.S. Army Corps of Engineers (USACE) has both military and civil responsibilities. Among its responsibilities for civil works, the Corps investigates, develops, and maintains the nation's water and related environmental resources, including navigation, shore protection, and flood control projects. In so doing, the Corps must deal with natural and man-made systems on a scale ranging from small watersheds and localized problems to the Mississippi River Basin.The Corps is charged with analyzing and understanding the systems for which it has stewardship responsibility, and has developed a planning process that attempts to display the benefits and costs of implementation of various project alternatives, and compare those benefits and costs with the “do-nothing” or “without project” alternative. A fundamental question that is to be answered when the Corps undertakes a study of a proposed project is “Is it worth it?” That is, do the benefits to the nation of implementing a project exceed the costs? The calculation of benefits and costs is done over a period of time, typically 50 years.When planning is viewed in this framework, it is clear that, in order to estimate benefits and costs, assumptions must be made regarding natural phenomena and human behaviors into the future. How many hurricanes will take place in the project area? How severe will they be? What will people do to their houses in the face of repeated flooding? Will they rebuild to a safer construction standard? If a port channel is deepened to allow larger vessels to service a port, how will this affect traffic at other ports? If a lock is persistently out of service, at what point will shippers choose to use other modes of transportation (rail, truck)?What we have, then, is the need for analysis of a natural-engineering-economic-behavioral system that is clearly beyond complete predictive capability. Unlike systems that are largely man-made, where failure rates can be estimated and reliability established in an analytical sense, the long periods of time of a life cycle, the limited knowledge and data available to predict the natural system behavior, and the difficulty of predicting human behavioral responses, all combine to create an analysis problem that we know we cannot solve; we can only estimate. That is, we understand that the true complexity of the problem is beyond our capabilities to analyze, and that simplifications must be made if we are to approach answering the “Is it worth it?” question.In the face of this evolving understanding of the problem, the Corps has elected, as noted above, to adopt a life cycle approach, coupled with the incorporation of “risk and uncertainty” as a fundamental component of the analysis process (Moser 1996). What does this mean? It means that the Corps recognizes that any historical record of natural events can be considered as a single sample from a large possible population of such records. It means that there is an understanding that data are inherently uncertain, and are often best characterized as distributions rather than point estimates. It means that the future will not be like the past, and that the selection of any single “future” is not a good basis for analyzing the benefits of a particular project. Because investments are made in the near term, at a specific (and largely fixed) scale, they will produce different outputs (in terms of actual damages prevented) depending upon the future conditions that will occur. A low-cost shore protection project may work perfectly acceptably under one possible future sequence of storms under which few hurricanes hit the area, but lead to large economic losses (and possible loss of human life) under another sequence of storms of more severe weather impact. Thus, most importantly, it means that decision-making must take into account the variability and uncertainty in project outputs, and that analysts must be able to provide estimates of that variabi