Testing habitat-based predictions of community responses to river flow restoration: generic lessons from a data-rich case study.

Testing our ability to predict the ecological effects of flow and habitat restoration is often complicated by the weaknesses of monitoring strategies. In this context, data-rich flow experiments are unique occasions to draw generic lessons concerning the predictive power of habitat-based models and the optimization of monitoring strategies. We provide an overview of such a data-rich restoration case study (hydrological changes in the Rhone River in France) and discuss some general lessons learnt from the project. The restoration started in 1999 and has combined so far minimum flow increases in four reaches (total length 47 km) and the dredging and/or reconnection of 24 floodplain channels. The project concerned many sites and involved quantitative ecological data (hydraulics, temperature, sediment, fish, macroinvertebrates) collected over long periods. Tests of habitat-based predictions indicated that general habitat models, calibrated in independent sites and focused on the main physical drivers, could predict population and community-level changes after restoration. An analysis of statistical power revealed that, even in the data-rich situation of the Rhone, the probability to detect a moderate change (50 - 200%) in a given taxon abundance was typically 15%. These results indicate how to improve feedback from future flow experiments.