Using an Error Budget to Evaluate the Importance of Component Models Within a Large-Scale Simulation Model : Mathematical Modelling of Forest Ecosystems. Proceedings

Frankfurt am Main

1992

S. 62-74

15 Lit. Ang.

Bandaufführung

76784

Large scale forest growth simulation models are used extensively in developing long-term forestry management plans. With few exceptions, the existing forest growth models are complex models with many components. Traditionally, the prediction quality of these models are assessed by using either Monte Carlo or validation methods. Even with today's powerful computers, Monte Carlo methods still require a prolonged period of computation and cannot easily be used to assess the quality of model components. Good validation studies are rare, mainly due to the limited availability of high quality independent data. These shortcomings prohibit the use of the two approaches to assess the quality of predictions on a routine basis. In this paper, an error budget approach is described for assessing the quality of model predictions. An error budget approach is an alternative for assessing the quality of model predictions. As a catalog of errors, an error budget shows the effects of individual errors and groups of error on the accuracy of model predictions. The goal in developing an error budget is to account for all major sources of errors that can be expected in a forest growth model. Error propagation methods can be used to develop error budgets. An error propagation approach requires much less computation time than a Monte Carlo approach. With an error budget, the prediction quality of a model and its components can be assessed on a routine basis without the need for independent data. The overall objective of developing an error budget is to develop methodology that will systematically account for and approximate the sources of errors that can be expected to influence the performance of a growth model and its components.