A Comprehensive Hidden Markov Model for Hourly Rainfall Time Series
For hydrological applications, such as urban flood modelling, it is often important to be able to simulate sub-daily rainfall time series from stochastic models. However, the literature is currently lacking owing to several challenges with modelling rainfall at this resolution, namely a complex temporal structure including long dry periods, seasonal variation in both the occurrence, and intensity of rainfall, and extreme values. To address this, we propose a flexible model, an advanced hidden Markov model set in the Bayesian hierarchical framework, which is capable of capturing and reproducing these characteristics well. To demonstrate this, we apply the model to an 8-year time series of hourly observations. We also present thorough model checking and a critical discussion of our approach.
READ FULL TEXT