The methodology proposed here is aimed at providing both the deterministic water demand forecast and an estimation of the uncertainty affecting this forecast. Two models based on non-homogeneous and homogeneous Markov Chains (named Markov Chain A - MCA and Markov Chain B – MCB, respectively) are developed. These models are applied to three real case-studies with the aim of forecasting hourly water consumptions up to 6 hours ahead. The results obtained show that model based on the homogeneous Markov Chain (MCB) provides more accurate short term forecasts than the non-homogeneous Markov Chain model (MCA) and concurrently is capable of providing useful information about the forecasting uncertainty.