Traditionally, the field of reliability has been concerned with failure time data. As a result, degradation-based reliability methods have not been very well developed. This is especially true of analysis of degradation data resulting from highly variable environments. This dissertation, comprised of three papers, proposes two simulation-based methods to estimate reliability metrics for materials or products that degrade from exposure to the outdoor weather. In the first paper, time series modeling is used to estimate probability distribution of cumulative degradation in x years and probability distribution of failure time. A procedure to construct approximate confidence intervals for metrics of interest is also given. The second paper is an extension of the work presented in the first paper to include the case where there is an additional uncertainty due to unit-to-unit variability. The paper discusses reliability quantities of interest induced by the presence of two sources of variability and techniques to estimate them. Bayesian methods are used to estimate the distribution of the population of units, and an approximation technique to overcome computational difficulties is described. The third paper uses a model-free block bootstrap scheme to estimate reliability quantities in the context of periodic data. The degradation data has periodic structure due to the seasonality of the outdoor environment. The paper also proposes two methods to choose block size. The choice of block size is an important issue in the implementation of a block bootstrap scheme. A comparison is also made between the results from time series modeling and from block bootstrap.