Abstract: Weather extremes, including precipitation, are often analyzed using extreme value analysis (EVA). But many aspects of precipitation extremes over short to medium time scales relate to patterns of precipitation that are not captured by EVA of daily values. Furthermore, analysis is complicated by the presence of missing values and potential interest in trends in multiple metrics (e.g., storm intensity and storm frequency). I'll describe a flexible spline-based Bayesian hidden Markov model stochastic weather generator that I have developed. The goal of the generator is not as a generator per se but rather to statistically model daily precipitation over time by season at individual locations so as to allow for subsequent trend analysis. The model naturally accounts for missing data (considered missing at random), avoiding potential sensitivity from systematic missingness patterns or from arbitrary data inclusion cutoffs. The fitted model can then be used for inference about trends in arbitrary measures of precipitation behavior, either by multiple imputation of the missing data followed by frequentist analysis or by simulation from the Bayesian posterior predictive distribution. Using three stations from the western United States, I'll show that the model fits precipitation data well, including a variety of multi-day characteristics, indicating fidelity to the autocorrelation structure of the data. I then will assess trends in various aspects of precipitation (such as dry spell length and precipitation intensity). Limitations include that the low signal to noise ratio from analyzing single stations impedes trend detection, extension to modeling multiple stations involves much greater complexity, and MCMC mixing should be improved.