Automated construction of rulebases for forest resource planning

Abstract

Abstract In recent years the application of geographic information systems (GIS) has moved from highly specialized areas to general use. This trend stimulates the need for easy-to-use GIS based tools. In forest resource management, expert systems to support resource planning is one such tool. The STARR LAB in a cooperative project with the USDA Forest Service has developed the Rulebase Toolkit to expedite the process of developing rulebases to support Forest Service district-specific resource planning tasks. The design concept of the Rulebase Toolkit is that the complexity of integrating GIS, database management systems and expert systems should be hidden from an ordinary resource manager. This can be done by providing templates through the graphical user interface of the Toolkit. The templates provide transparent links to the GIS and underlying databases. A set of pre-defined spatial functions and database query functions are also built into those templates. With such utilities, the resource manager, as a rulebase developer, can simply point and click to define parameters, conditions and weights of the `IF THEN' rules based on the underlying spatial data layers and the database tables. Once all the heuristic rules are defined, the construction of the rulebase is then automated by the system. There are three types of rules supported by the Rulebase Toolkit. Score rules assign a quantitative weight to the relevant decision based on the truth of the facts involved within the condition(s) of a heuristic rule. Query rules retrieve data from the database related to a pertinent spatial data layer. Spatial rules process all needed spatial manipulation to generate spatially oriented facts for use by the score rules. The current implementation of the Rulebase Toolkit supports only a limited set of spatial functions such as calculating slopes and aspects or generating buffers and calculating distances. To make this toolkit more useful, a more comprehensive and generalized framework should be re-engineered.