1  Introduction

This project is part of a nationwide effort to monitor the status of birds in the underrepresented boreal region. In this report we describe the Quebec adaptation of the Boreal Optimal Sampling Strategy (BOSS). The BOSS design is a hierachical sampling approach stratified by ecoregions, habitat types, and cost constraits (Van Wilgenburg 2020). This structured design provides a spatially balanced coverage while accounting for rare habitats and sample cost. Here, we focus on the adaptation of the design for the Quebec province; for a thorough explanation and discussion of the national strategy, see Van Wilgenburg (2020).

In addition to stratifying the sampling based on habitat distribution and cost constraits, the BOSS design includes a function to take legacy sites and iconic sites into account. Legacy sites are existing or historical surveys with data extracted from randomly selected sites, whereas iconic sites are from non-randomly selected sites. The key reason for integrating legacy or iconic sites in the sampling design is to keep a representative sample of the community while reducing the sample cost. This is especially important in Quebec, as there are many historical data in the southern part of the province. Considering legacy sites in well-covered regions allows us to allocate ressources to remote areas with less data and higher sampling costs. In Chapter 5, we detail a novel approach accounting for the number and distribution of legacy and iconic sites to reduce sample size and maintain a representative sample of habitat types.

Once habitat types, cost constraints, and legacy sites are defined, the BOSS design uses the Generalized Random Tessellation Stratified Sampling (GRTS; Stevens Jr and Olsen (2004)) method to perform the random sampling. This is a widely used approach to ensure spatially balanced samples in a region. The GRTS uses a mapping function to transform two-dimensional space into one-dimensional space with an ordered spatial address. This one-dimensional ordered space is then randomly reordered before the sampling. This random reordering of the linear, one-dimensional space ensures a spatially well-balanced sample, whatever the sample size. After being sampled, this one-dimensional space is then mapped back to the original two-dimensional space.

This report is divided into two main sections. The first section details the spatial layers to feed the GRTS algorithm. We begin by describing the study area in the Quebec province, the selected ecoregions, and the Primary Sample Unit (PSU). We then detail de habitat and cost layers to weight the inclusion probabilities. Finally, we dedicate a complete section to describe the simulations used to create the new method to account for legacy and iconic sites. The first section contains most of the steps that have been regionalized for Quebec. The second section details the sample steps using the GRTS algorithm. In this second part, we will begin by describing the method used to calculate the stratified sample size for each of the ecoregions. We then detail the use of the GRTS to sample the PSUs and the Secondary Sample Unit (SSU).

Stevens Jr, Don L, and Anthony R Olsen. 2004. “Spatially Balanced Sampling of Natural Resources.” Journal of the American Statistical Association 99 (465): 262–78.

Van Wilgenburg, C. Lisa AND Campbell, Steven L. AND Mahon. 2020. “A Cost Efficient Spatially Balanced Hierarchical Sampling Design for Monitoring Boreal Birds Incorporating Access Costs and Habitat Stratification.” PLOS ONE 15 (6): 1–28. https://doi.org/10.1371/journal.pone.0234494.