A method for the objective selection of landscape‐scale study regions and sites at the national level

Abstract

Summary Ecological processes operating on large spatio‐temporal scales are difficult to disentangle with traditional empirical approaches. Alternatively, researchers can take advantage of ‘natural’ experiments, where experimental control is exercised by careful site selection. Recent advances in developing protocols for designing these ‘pseudo‐experiments’ commonly do not consider the selection of the focal region and predictor variables are usually restricted to two. Here, we advance this type of site selection protocol to study the impact of multiple landscape scale factors on pollinator abundance and diversity across multiple regions. Using datasets of geographic and ecological variables with national coverage, we applied a novel hierarchical computation approach to select study sites that contrast as much as possible in four key variables, while attempting to maintain regional comparability and national representativeness. There were three main steps to the protocol: (i) selection of six 100 × 100 km2 regions that collectively provided land cover representative of the national land average, (ii) mapping of potential sites into a multivariate space with axes representing four key factors potentially influencing insect pollinator abundance, and (iii) applying a selection algorithm which maximized differences between the four key variables, while controlling for a set of external constraints. Validation data for the site selection metrics were recorded alongside the collection of data on pollinator populations during two field campaigns. While the accuracy of the metric estimates varied, the site selection succeeded in objectively identifying field sites that differed significantly in values for each of the four key variables. Between‐variable correlations were also reduced or eliminated, thus facilitating analysis of their separate effects. This study has shown that national datasets can be used to select randomized and replicated field sites objectively within multiple regions and along multiple interacting gradients. Similar protocols could be used for studying a range of alternative research questions related to land use or other spatially explicit environmental variables, and to identify networks of field sites for other countries, regions, drivers and response taxa in a wide range of scenarios.

Risk of potential pesticide use to honeybee and bumblebee survival and distribution: A country‐wide analysis for The Netherlands

“…The risk assessment for active ingredients was done using a hazard quotient (HQ). The method was based on an insecticide loading assessment method presented in Gillespie et al (). The hazard quotient was calculated as follows:…”

Indirect effects of agricultural pesticide use on parasite prevalence in wild pollinators

“…This set of sites were selected from a larger database of field sites ( Fig. S1) originally identified by the IPI AgriLand project (Linking agriculture and land use change to pollinator populations, BB/I000364/1; Supplementary Material section S4; Gillespie et al 2017). The farms in this database are a randomised selection of farms that were chosen to encompass variation in four specific variables thought to be important in driving pollinator declines, yet were otherwise comparable (Gillespie et al 2017).…”

“…S1) originally identified by the IPI AgriLand project (Linking agriculture and land use change to pollinator populations, BB/I000364/1; Supplementary Material section S4; Gillespie et al 2017). The farms in this database are a randomised selection of farms that were chosen to encompass variation in four specific variables thought to be important in driving pollinator declines, yet were otherwise comparable (Gillespie et al 2017). These variables were pesticide use, habitat diversity, floral resource availability, and managed honey bee colony density (see Gillespie et al 2017 and Supplementary Materials, section S3 for specific details on how these were calculated).…”

An analytical framework for spatially targeted management of natural capital

“…At high coverage, croplands coalesce into larger patches, reducing edge effects and nutrient retention by other land covers such as forest. It is critical to determine when configuration is important 13 for natural capital, and is likely to be possible in the future, given that monitoring programmes are increasingly incorporating regional and landscapelevel considerations in their sampling strategies 71,72 .…”