Time series analysis

Abstract

A characteristic feature of hydrologic events is that they are not independently distributed in time. The degree of linear dependence between events k time units apart is measured by the serial correlation coefficient. This correlation, which tends to decrease with an increase in k, is attributed to storage processes in the atmosphere or within the drainage basin. The correlation measures the degree of redundancy of information yielded by each hydrologic event. This redundancy implies that statistical parameters computed from a sequence of events are less reliable than is indicated by the sequence length. Hydrologic sequences are formulated in terms of generating models to account for serial dependence between events, to allow prediction of events at future time points, and to assess the effect of serial dependence on hydrologic studies.