Balancing sampling intensity against spatial coverage for a community science monitoring programme

Community science is an increasingly integral part of biodiversity research and monitoring, often achieving broad spatial and temporal coverage but lower sampling intensity than studies conducted by professional scientists. When designing a community-science monitoring programme, careful assessment of sampling designs that could be both feasible and successful at meeting programme goals is essential. Monarch butterflies (Danaus plexippus) are the focus of several successful community-science projects in the U.S., but broader coverage is needed to monitor breeding areas and explain population declines observed in overwintering areas. The U.S. Monarch Conservation Science Partnership's Integrated Monarch Monitoring Program (IMMP) will representatively monitor monarchs and milkweed across North America. We performed a simulation-based power analysis to predict trade-offs between sampling breadth (number of sites and years) and sampling intensity (number of visits or subplots per site and year) for the IMMP. We evaluated whether each sampling design would produce sufficient statistical power to detect population trends and differences among land-use sectors in densities of milkweed, monarch eggs, and adult monarchs. Sampling breadth had much stronger effects than sampling intensity on statistical power for all three monitoring targets. Depending on land-use sector, monitoring 400–800 sites over 10–15 years would detect trends in densities of milkweed and adult monarchs, but no scenarios were successful for monarch eggs. Sampling 400–800 sites would also detect small (for adult monarchs) to large (for milkweed) differences among land-use sectors in density of all three monitoring targets within the first 2–5 years. As more data become available from the IMMP, the sampling goals can be updated. Synthesis and applications. Careful sample design is an essential step in developing a successful monitoring programme. For monarchs and milkweed, we found that sampling breadth (number of sites and years) had a much stronger effect on statistical power than sampling intensity (number of visits or subsamples per site), suggesting field protocols could be tailored to maximize recruitment and retention of volunteers by minimizing the effort required to monitor each site. Many long-term monitoring programmes might similarly benefit from evaluating the statistical trade-offs between sampling breadth and intensity in their sampling designs.