Note on the December 2013 Floods in Brazil

Editor's Note: Late last year, three southeastern states in Brazil suffered their worst flooding in decades. This article describes many of perspectives that can be used to understand the severity of a flood event and future tools to manage flood risk in Brazil.

Rainfall accumulation, however, is not the sole indicator of flood intensity. While heavy precipitation is a necessary condition for river flooding, a multitude of other factors—including pre-existing soil moisture levels and river heights, and the presence of flood defenses—affect the severity of a flood. Thus, record-setting precipitation does not necessarily translate to record-setting flooding, and calling a flood a "1 in 100 year" event can mean several different things. Hydrologists and flood engineers examine several other measures, including but not limited to: water height (height measured from the bottom of the body of water), river flow, flood footprint, and inundation depth (height of water in flooded area).

For last December's event, river gauge data suggest that the severity of the flood event is not as extraordinary as the precipitation amounts would suggest. For example, the observed peak flow of the Rio Doce in Colatina for the December 2013 event was 962 m³/s. This level has been achieved or exceeded several times in recent decades, suggesting an annual probability of occurrence significantly higher than 1%. Risk managers can also look at the severity of a flood event from an insured loss or economic loss perspective (keeping in mind that exposure growth and inflation need to be considered when comparing a recent loss to those of past historical events).

Because of Brazil's continued economic development and expanding middle class, the insurance sector continues to grow; as a result, flood losses are an increasing concern for risk managers. Like any other natural catastrophe, flood risk is best managed with a probabilistic model that can consider the complete range of potential adverse events. Building fully probabilistic inland flood models, however, is computationally resourceful and can take many years. They also require extremely high-resolution and high-quality data on elevation and hydrology—data that may simply not be available for an individual country.