Realised and predicted changes in the invertebrate benthos after restoration of connectivity to the floodplain of a large river

River-floodplain interactions (i.e. lateral connectivity) are a major driver of the structure and function of the biota in floodplain waterbodies. Therefore, the enhancement of lateral connectivity is a key focus of large river restoration. The Rhone River restoration project in France includes an array of floodplain habitats and restoration measures, offering opportunities to test the predictability of changes in taxon abundance after connectivity restoration. Quantitative samples of benthic macroinvertebrates taken before restoration were used to derive generalised additive models (GAMs) relating the density of taxa to an index of lateral connectivity. The index combined the diversity of sediment grain size, its organic content, vegetation cover and the water's electrical conductivity. GAMs were used to predict changes in taxon density and between-site (i.e. beta) diversity 2 and 4years after restoration at 34 sites (14 unrestored, eight reconnected to the river, six dredged, six restored by increasing river flow). Observed post-restoration changes included an increase in the density of rheophilous taxa and an increase of beta-diversity among restored sites. Fifty-four GAMs (27% of the taxa) were selected for post-restoration predictions. Predictions from these models matched observations for a set of Trichoptera and other taxa with large density changes after restoration. The quality of predictions of density changes was poor on average among sites, but higher for sites where large changes in lateral connectivity occurred. Taxon-level predictions partly explained increased beta-diversity in restored sites, suggesting that predictions based on a subset of individual taxa can be used to predict community-level changes. Our models can guide restoration measures in the Rhone and other rivers. Our results also show the benefit of diversifying restoration actions at the scale of entire floodplains. Post-restoration trajectories should be monitored over long periods to improve the identification of the key environmental predictors of biotic changes, especially at the less connected floodplain sites.