Optimizing Large-Scale Biodiversity Sampling Effort: Toward an Unbalanced Survey Design

Enrique Montes,Francis Choi,Nicolas Moity,Brian Helmuth,Frank Muller-Karger,Gregorio Bigatti,Edgardo Londoño-Cruz,Cesar Cordeiro,Edlin Guerra Castro,Eulogio Soto,Jonathan Lefcheck,Patricia Miloslavich,Ana Carolina De Azevedo Mazzuco,Eduardo Klein,Nuno Simoes,Erasmo Macaya

doi:10.5670/oceanog.2021.216

Abstract

Acquiring marine biodiversity data is difficult, costly, and time-consuming, making it challenging to understand the distribution and abundance of life in the ocean. Historically, approaches to biodiversity sampling over large geographic scales have advocated for equivalent effort across multiple sites to minimize comparative bias. When effort cannot be equalized, techniques such as rarefaction have been applied to minimize biases by reverting diversity estimates to equivalent numbers of samples or individuals. This often results in oversampling and wasted resources or inaccurately characterized communities due to undersampling. How, then, can we better determine an optimal survey design for characterizing species richness and community composition across a range of conditions and capacities without compromising taxonomic resolution and statistical power? Researchers in the Marine Biodiversity Observation Network Pole to Pole of the Americas (MBON Pole to Pole) are surveying rocky shore macroinvertebrates and algal communities spanning ~107° of latitude and 10 biogeographic ecoregions to address this question. Here, we apply existing techniques in the form of fixed-coverage subsampling and a complementary multivariate analysis to determine the optimal effort necessary for characterizing species richness and community composition across the network sampling sites. We show that oversampling for species richness varied between ~20% and 400% at over half of studied areas, while some locations were undersampled by up to 50%. Multivariate error analysis also revealed that most of the localities were oversampled by several-fold for benthic community composition. From this analysis, we advocate for an unbalanced sampling approach to support field programs in the collection of high-quality data, where preliminary information is used to set the minimum required effort to generate robust values of diversity and composition on a site-to-site basis. As part of this recommendation, we provide statistical tools in the open-source R statistical software to aid researchers in implementing optimization strategies and expanding the geographic footprint or sampling frequency of regional biodiversity survey programs.

Highlights

Human society has long been interested in understanding the patterns, drivers, and consequences of biodiversity change both in specific locations and across the globe (Mittelbach et al, 2007; Barnosky et al, 2011; Dornelas et al, 2014)
Because the objective here is to compare changes in species composition along the rocky shores of the Americas, we propose that community composition estimates should be made using a common bounded measure of dissimilarity (i.e., Jaccard) and a common level of precision (i.e., 0.1)
Sampling effort was generally comparable among localities, with an average of ~25 quadrats collected per survey over the three strata: high, mid, and low tide

Summary

Introduction

Human society has long been interested in understanding the patterns, drivers, and consequences of biodiversity change both in specific locations and across the globe (Mittelbach et al, 2007; Barnosky et al, 2011; Dornelas et al, 2014). Where effort cannot be equalized, rarefaction techniques have been applied to minimize biases by reverting diversity estimates to equivalent numbers of samples or individuals (Hessler and Sanders, 1967). Such approaches often result in considerable loss of hard-earned information and wasted effort. They fail to detect and report undersampling, and do not provide information about where more samples are needed to adequately characterize biodiversity

Methods

Results

Discussion

Conclusion