Flood avalanches in a semiarid basin with a dense reservoir network

- Flood avalanches in a semiarid basin (5000 dams) are investigated (1991-2010).
- We use field data, imagery and a model to simulate each reservoir explicitly.
- Small reservoirs prevent the formation of large clusters at low storage level.
- The exponent of the avalanche-size distribution is stable, in the range 2.05 ± 0.15.
- Existing network performs better than numerous stochastically-generated networks.

SummaryThis study investigates flood avalanches in a dense reservoir network in the semiarid north-eastern Brazil. The population living in this area strongly depends on the availability of the water from this network. Water is stored during intense wet-season rainfall events and evaporates from the reservoir surface during the dry season. These seasonal changes are the driving forces behind the water dynamics in the network. The reservoir network and its connectivity properties during flood avalanches are investigated with a model called ResNetM, which simulates each reservoir explicitly. It runs on the basis of daily calculated water balances for each reservoir. A spilling reservoir contributes with water to the reservoir downstream, which can trigger avalanches affecting, in some cases, large fractions of the network. The main focus is on the study of the relation between the total amount of water stored and the largest observable cluster of connected reservoirs that overspill in the same day. It is shown that the thousands of small and middle-sized reservoirs are eminent for the retention of water upstream the large ones. Therefore, they prevent large clusters at a low level of water. Concerning connectivity measures, the actual reservoir network, which evolved without an integrated plan, performed better (i.e., generated smaller avalanches for similar amount of stored water) than numerous stochastically generated artificial reservoir networks on the same river network.