Using flood process information to support flood trend studies

Floods are a common natural hazard with costly and often fatal consequences. Under changing climatic conditions, extreme precipitation events have increased in severity. This is expected to affect flood magnitude. However, several previous studies were not able to reach comparable conclusions in regard to direction of flood trends and their drivers. We investigate some potential explanations for this inconsistency. These include (1) the importance of climate-dependent differences in flood generating processes and (2) the effect of alternative event sampling methods. We investigate these two aspects using a quasi-global dataset comprising catchment data from Australia, Brazil, Chile, Great Britain, and the United States. The compiled dataset represents a wide range of climates.

We first evaluate how streamflow reacts to precipitation under different initial soil moisture conditions. It is well known that extreme precipitation is more likely to cause flood events under high soil moisture conditions. Our results show that this interaction of precipitation, soil moisture, and streamflow changes with aridity. This indicates that the influence of soil moisture on flood trends might change with climate zone and should be considered in future flood projections.

Secondly, we compare two sampling strategies that are currently in use: sampling events by extreme precipitation and sampling events by extreme flow. Sampling events by extreme precipitation is likely to miss some significant flood events and include lower flow events due to the strong interaction between precipitation and soil moisture resulting in misclassification of flood trends. To summarise, we recommend that future studies of flooding (i) should stratify large samples according to flood process, e.g. by using climate as an indicator of process and (ii) should include extreme flow events which are caused by moderate precipitation combined with high initial soil moisture.