This article examines whether the temporal clustering of flood events can be explained in terms of climate variability or time-varying land-surface state variables. The point process modelling framework for flood occurrence is based on Cox processes, which can be represented as Poisson processes with randomly varying rate of occurrence. In the special case that the rate of occurrence is deterministic, the Cox process simplifies to a Poisson process. Poisson processes represent flood occurrences which are not clustered. The Cox regression model is used to examine the dependence of the rate of occurrence on covariate processes. We focus on 41 stream gauge stations in Iowa, with discharge records covering the period 1950-2009. The climate covariates used in this study are the North Atlantic Oscillation (NAO) and the Pacific/North American Teleconnection (PNA). To examine the influence of land-surface forcing on flood occurrence, the antecedent 30 d rainfall accumulation is considered. In 27 out of 41 stations, either PNA or NAO, or both are selected as significant predictors, suggesting that flood occurrence in Iowa is influenced by large-scale climate indices. Antecedent rainfall, used as a proxy for soil moisture, plays an important role in driving the occurrence of flooding in Iowa. These results point to clustering as an important element of the flood occurrence process.
All Science Journal Classification (ASJC) codes
- Atmospheric Science
- Cox regression
- Temporal clustering