Stated preference and revealed preference are two commonly conducted non-market value evaluation methods which can also be applied to make evaluation of forest ecosystem. In the application of these evaluation methodologies, there always exists limitation from the data collection and empirical analysis. In the dissertation here, I extend the traditional evaluation methods with novel design or statistical analysis approaches to solve the practical problem we met in evaluation of forest ecosystem. The first and second chapters are based on stated preference methods. The first chapter employ both the mail survey and on-site survey to investigate the preference for attributes of low-impact timber harvesting programs. In the second chapter, we recruit three interest groups for on-site survey and compare their preference for the low-impact timber harvesting programs. In these first two chapters, choice modeling method is employed to elicit the respondents' preferences, and I also use bootstrap method to get robust estimation results for small sample size data. The last chapter employed revealed preference method to evaluate the economic losses from hemlock damages caused by forest pest. Three different interpolation methods are employed to scale-up the analysis from sites to states. Based on the findings of all three chapters, we can see that these survey design and statistical methods help to overcome the limitations in empirical analysis of forest ecosystem and make more robust inferences for design forest protection policies. / Ph. D.