Evaluation of the transferability of hydrological model parameters for simulations under changed climatic conditions

Conceptual hydrological models are widely used for climate change impact assessment. The implicit assumption in most such work is that the parameters estimated from observations remain valid for future climatic conditions. This paper evaluates a simple threshold based approach for testing this assumption, where a set of behavioural simulators are identified for different climatic conditions for the future simulation i.e. wet, average and dry conditions. These simulators were derived using three different data sets that are generated by sampling a block of one year of data without replacement from the observations such that they define the different climatic conditions. The simulators estimated from the wet climatic data set showed the tendency to underestimate flow when applied to dry data set and vice versa. However, the performances of the three sets of basin simulators on chronologically coherent data are identical to the simulators identified from a sufficiently long data series that contains both wet and dry climatic conditions. The results presented suggest that the issue of time invariance in the value of parameters has a minimal effect on the simulation if the change in precipitation is less than 10% of the data used for calibration