Rainfall and river flow are one of the most difficult elements of hydrological cycle to predict. This is due to tremendous range of variability it displays over a wide range of scale both in terms of space and time. The situation is further aggravated by the fact that rainfall-runoff is also very difficult to measure at scales of interest to hydrology and climatologic. Computational intelligence techniques provide efficient and fast results for modelling non-linear and complex data. Computational intelligence methods which inspired by the capability of learning that derive meaning from unknown relationship provide guidance for a sensible decision making. This advantage creates them adaptable and talented methods for modelling real world problems. This paper is an attempt to present the introduction to computational intelligence methods; applications to river flow modelling and its performance with regards to the parameter and method used. The methods include artificial neural networks, fuzzy logic, evolutionary computation, support vector machine; swarm intelligence and hybrid method are critically compared mainly on computational results and prediction accuracy.