Simulation is defined as a technique for imitating the evolution of a real system by studying a model of the system. The model is an abstraction, a simplified and convenient mathematical representation of the actual system, typically coded and run as a computer program. If the model has a stochastic element, then we have
stochastic simulation
. The term stochastic simulation sometimes is used synonymously with the
Monte‐Carlo method
.
Stochastic simulation is regarded as mathematical experimentation and is appropriate for complex systems, whose study based on analytical methods is laborious or even impossible. For such systems, stochastic simulation provides an easy means to explore their behavior by answering specific ‘what if’ questions. Moreover, stochastic simulation can be viewed as a numerical method for solving mathematical problems in several fields such as statistical inference, optimization, integration, and even equation solving. Under certain conditions, stochastic simulation is more powerful than other more common numerical methods (e.g., numerical integration of high‐dimensional differential equations).
Due to their complexity, hydrosystems, including water resource systems, flood management systems, and hydropower systems, are frequently studied using stochastic simulation. A generalized solution procedure for hydrosystems problems, including systems identification, modeling and forecasting, hydrologic design, water resources management, and flood management, is dicussed. Emphasis is given on the stochastic representation of hydrologic processes, which have a dominant role in hydrosystems. Peculiarities of hydrologic and other geophysical processes (seasonality, long‐term persistence, intermittency, skewness, spatial variability) gave rise to substantial research that resulted in numerous stochastic tools appropriate for applications in hydrosystems. Four examples of such tools are discussed: (1) the multivariate periodic autoregressive model of order 1 [PAR(1)], which reproduces seasonality and skewness but not long‐term persistence; (2) a generalized multivariate stationary model that reproduces all kinds of persistence and simultaneously skewness but not seasonality; (3) a combination of the previous two cases in a multivariate disaggregation framework that can respect almost all peculiarities except intermittency; and (4) the Bartlett‐Lewis process that is appropriate for modeling rainfall and emphasizes its intermittent character on a fine time scale.