Dependance of restoration outcomes on contemporary spatial context

Task 3.1: Extract and process the spatial context data (drivers, pressures and status) (INRAE, WEnR)

Based on European/national available data (Table 3) and enquiries amongst stakeholders (WP6), we will extract and process the spatial context information of the restoration sites, in terms of both biotic and abiotic conditions, to derive hypothesized explanatory factors of restoration outcomes at different scales. At catchment scale, we will consider context in terms of environmental descriptors; hydrology (e.g. HydroSHEDS databases / GloRiC), morphology (e.g. local data, aerial photos), water chemistry (e.g. Waterbase - Water Quality), riparian characteristics (EEA’s databases), connectivity (e.g. European Barrier Atlas AMBER), land-uses (Corine Land Cover), point sources (stakeholder enquiries), and socio-ecological descriptors such as population density (GHS-POP), river accessibility and recreational activities (e.g. OpenStreetMap, enquiries amongst stakeholders). At the sub-catchment scale, we will consider hydrology, land-uses and riparian characteristics too, but also management policies (Natura 2000, RAMSAR sites), the occurrence of nearby restoration projects/sites (based on the restoration datasets themselves, RAMSAR database, etc.), and navigation intensity (e.g. INE). At the reach and floodplain scales, we will consider land uses, hydrology, riparian characteristics, but also touristic information (e.g. touristic harbours) and management measures (e.g. Natura 2000, RAMSAR sites). At all scales, biotic factors such as fragmentation/connectivity (e.g. barrier cumulative height, connectivity index) and size of regional species pools (based on e.g. regional atlas data, IUCN Red List, GBIF; see Barbarossa et al. 2020) will also be considered. Data homogenisation and standardisation will allow building appropriate environmental descriptors needed for subsequent analyses (e.g. stress scores at different scales; see de Vries et al. 2019) and translating these variables into drivers, pressures and states (in line with the DPSIR framework).

Task 3.2: Quantify the links between spatial context variables (drivers, pressures and status) and ecological outcomes of restoration (WEnR, INRAE, UCB)

We will identify and quantify the significant links between environmental descriptors (generated in Task 3.1) and the ecological metrics of restoration outcomes (according to WP1 & WP2 outputs), using statistical analyses that account for dataset heterogeneity, complex relationships, variable interactions, and multi-scale responses (e.g. hierarchical, generalized mixed models and boosted regression tree techniques). We will analyse each individual project dataset separately with the same procedure and will integrate all the project-specific results with a “meta-reanalysis” approach (Jeliazkov et al. 2018; Pilotto et al. 2020) in which the effects of spatial context are tested, while accounting for both noise due to dataset heterogeneity, and effect sizes (Osenberg et al., 1994; Nakagawa & Schielzeth 2013). These analyses will allow us to rank the relative importance of the different context variables / drivers and pressures across scales in the ecological outcomes of restoration (Hyp. 1) and to test and quantify the causal relationships (e.g. via structural equation modelling; Hair Jr. et al. 2016) previously hypothesized in WP2 within the DPSIR framework. 

Task 3.3: Quantify the links between spatial context parameters (drivers, pressures and status) and societal outcomes of restoration (UCB, EAWAG, INRAE)

Similarly to Task 3.2, we will identify and quantify the links between environmental descriptors (generated in Task 3.1) and the identified societal metrics of restoration outcomes (according to WP1 & WP2 outputs) with the methods used in Task 3.2 (Hyp. 1). The main methodological difference compared to Task 3.2 is that here, we use a unified source of data (social media information) to calculate the societal outcomes of each restoration project. Consequently, we won’t need the “meta-reanalysis” approach and will be able to directly analyse the outcomes (instead of effect sizes) as responses of context drivers. We will estimate the potential predictive efficiency of these models in order to assess their transferability to other geographical areas and scales, e.g. via cross-validation procedures considering the hierarchical spatial structure of the data. These analyses will allow us to further refine the BBN (WP2) and to better understand what drives the societal success of restoration.