Testing predictions of changes in benthic invertebrate abundance and community structure after flow restoration in a large river (French Rhône)

International audience1. Principal threats to running waters are linked to human-made discharge modifications, but tools to predict the quantitative consequences of flow restoration for benthic invertebrates in large rivers remain untested. 2. Quantitative benthos samples from two bypassed reaches (Pierre-Benite – PBE – and Chautagne – CHAU) of the French Rho^ne River were collected during four years each before and after minimum flow increases (from 10 to 100 m3 s␣1 at PBE and from 10 to 50 m3 s␣1 at CHAU). These samples provided observed ln-density changes for the 50 and 62 (PBE and CHAU, respectively) most abun- dant taxa (typically species or genera). For about half of the ‘model’ taxa among them, distinct pref- erence models for bottom shear stress categories were available from four reaches of the Upper Rho^ne River and from various German rivers. Linking these preference models with a statistical hydraulic model predicting frequencies of shear stress categories for any given discharge, we pre- dicted ln-density changes of the model taxa after restoration.3. Community structure of the abundant taxa changed clearly and rapidly after restoration at PBE but less clearly at CHAU. Our predictions explained a considerable amount of mean ln-density changes of our model taxa observed after restoration (75 and 30% at PBE and CHAU, respectively). They also explained (67 and 40% at PBE and CHAU, respectively) the model taxa scores along the principal components analysis axis that summarised the community variations of all abundant taxa before and after restoration. For taxa not identified to species (assessed as genus, tribe or family), the predicted ln-density changes were sometimes inaccurate at PBE (a Lower Rho^ne site), suggesting that the transferability of preference models for taxonomic levels above that of species can be problem- atic.4. If further developed, statistical habitat models focussing on ecologically relevant physical variables (in our case shear stress) should enable reliable quantitative assessments of associations between flow restoration efforts and achievable ecological improvement