Uncertainty and confidence intervals in optical design using the Monte Carlo ray-trace method

摘要

The increasing use of probablistic methods, such as the Monte Carlo ray-trace (MCRT) method, in thermal radiation and optical modeling, has created a general awareness in the community of the need for a protocol to predict, to a specified level of confidence, the uncertainty of the results obtained using these methods. This paper presents such a protocol applied to models of radiometric channels used in spaced-based earth observations. It is anticipated that the same protocol, with suitable modification, may be extended to data from actual instruments. The authors and their colleagues have developed a powerful generic MCRT-based computational environment that, among other features, is capable of simulatign radiative exchange among surfaces and within enclosures. As in any MCRT thermal-radiative/optical model, the spatial resolution and accuracy of the results obtained depend on the fineness of the mesh, the number of rays traced, and the accuracy of directional and spectral surface property models. The protocol presented in this paper identifies and quantifies the contribution of these factors to the ultimate uncertainty in predicted results and to their relate4d confidence intervals.