Deliverable 2.7 (D14) v2.1 Report on the improvement of ecological effectiveness of collaborative schemes through multiscale spatiotemporal modeling of biodiversity indicators and their relations with environmental and anthropic drivers

The report presents the results of project Task 2.7 tackling the spatiotemporal modelling and the upscaling from the field to the landscape scale of five SHOWCASE core biodiversity indicators: wild bee abundance and species richness, spider abundance and species richness, and vascular plant species richness. The indicators have been modelled in five case study areas (CSAs) encompassing the SHOWCASE EBAs in Hungary (Kiskunság), Spain (Guadalquivida), Portugal (Alentejo), the Netherlands (Zuid Limburg) and Switzerland (Solothurn). After a brief introduction in Chapter 1 of the state of the art and the goals of Task 2.7, Chapter 2 of the Deliverable describes the methodology adopted to model the impacts of the implementation of biodiversity friendly farm management, as implemented by project partners in the SHOWCASE EBA fields. For all the EBAs with available data for the two years of surveys, a set of common predictors were associated with field data, providing a harmonized set of covariates to be used for the spatiotemporal modelling. The set of covariates encompasses terrain attributes, landscape structure descriptors, indices from remote sensing describing vegetation and soil status, and variables describing the implementation of the biodiversity friendly management practices. All the predictor data used are extracted from free data sources resorting as much as possible to EU official data sets. To fulfill the goal of upscaling field results, two different data driven approaches were tested: machine learning algorithms, in particular Random Forest (RF), and multiple linear regressions (MLR). Based on a set of error indices, the most accurate model was selected for each indicator in each CSA. The use of the measured and modelled biodiversity indicators as proxies of ecosystem services provision in the five EBAs is also discussed. Data driven model results for pollinator abundance were compared with the outcomes of the mechanistic model of Lonsdorf et al (2009), and selected ecosystem services were identified based on biodiversity indicator data as foreseen in the description of Task 2.7. Chapter 3 of the Deliverable reports the results of the spatiotemporal modelling in each CSA for the five selected indicators, expressed as 0-1 interval normalized scores. In all cases MLRs outperformed RFs in terms of accuracy and agreement with observed data. A sixth composite indicator was calculated by summing and normalizing the five considered indicators. Raster maps at 10m resolution are presented for each indicator, for the two rounds of sampling of the two years under two different management scenarios: control and intervention. A total of 240 indicator maps were produced and analyzed. These maps provide a spatially explicit and time dynamic assessment of the potential impact of the implementation of biodiversity friendly management practices at landscape scale. Chapter 4 of the Deliverable focuses on the comparison of the mechanistic and data driven model results for pollinator occurrence in the five selected EBAs, focusing on the impacts of the different drivers in the two approaches and on the resulting spatial patterns. Chapter 5 closes the Deliverable by first comparing the biodiversity indicators modelling results across the five CSAs, as well as the related ecosystem services estimates and maps, and then summarizing the key findings of Task 2.7, providing an outlook on their further use for scientific analyses within SHOWCASE.