Modeling of intermittent streamflow and precipitation in arid and semi-arid regions using hidden Markov models

  • Catherine Wilcox

Student thesis: Master's Thesis

Abstract

A two-level model based on Markov and hidden Markov chains is developed for the prediction of rainfall and intermittent streamflow in arid regions. The model is characterized by an outer standard Markov chain switching between states of non-zero and zero valuesl, and an inner hidden Markov chain based on the distribution of non-zero values. The model was tested on data from semi-arid and arid regions of the southwestern United States, and also on data from the hyperarid United Arab Emirates. Comparison with a standard hidden Markov model illustrated the improvements the two-level model has over the base case. Preprocessing via clustering techniques greatly reduced the predictive error of the model. Results showed that although the model performed especially well on streamflow in semiarid regions, the sparsity of data in time series from hyper-arid regions prevented the model from properly calibrating.
Date of AwardJun 2013
Original languageAmerican English
SupervisorTaha B. M. J. Ouarda (Supervisor)

Keywords

  • Hidden Markov Chains
  • Semi-Arid Regions
  • Arid Regions
  • Hyperarid Regions
  • Southwestern United States
  • United Arab Emirates
  • Streamflow.

Cite this

'