Increasing nonstationarity increases the uncertainty of hydrological processes. Consequently, intrinsic randomness increases in
magnitude and occurrence. In order to generate a reliable and robust predictive model, time series need to be better understood. This study addresses this challenge through the internal causality of annual runoff series. According to new methodological tendencies for hydrological research, complex temporal dependences within time series have been adequately captured through causal reasoning implemented by Bayes’ theorem. Those dependences permit quantifying the relative percentage of annual runoff change attributable to causality. This was later useful for calculating the temporally conditioned/nonconditioned runoff (TCR/TNCR) fractions. Results satisfactorily show the high and low temporally conditioned behavior of Porma-Esla and Adaja subbasin runoff, respectively. This study also provides a new stochastic approach for the return period (RP) assessment. Using TCR and TNCR fractions, the RP for each fraction was calculated and called the temporally conditioned RP (TCRP) and temporally nonconditioned RP (TNCRP), respectively. Results show coherent behavior, and, consequently, the highest RP corresponds to the largest runoff and vice versa.
