Designing cost-effective assemblage-based ecological models by optimizing taxonomic resolution and sampling effort

Background/Question/Methods

Statistical models relating ecological assemblages to environmental conditions are vital components of many biomonitoring programs and environmental assessments. Programs incorporating these models often involve comprehensive sampling, and incur substantial costs. Limited resources mean that expensive sampling and analytical procedures should be planned to maximize information gain while minimizing unnecessary expense. However, there has been little consideration of cost effectiveness in parameterizing ecological assemblage models, and no explicit consideration of cost-effectiveness in balancing investment in the crucial aspects of sample size and taxonomic resolution.

Using lacustrine diatom (Bacillariophyceae) assemblages from regional datasets comprising >1200 lakes, we address the following questions: 1) how does taxonomic resolution affect model information content; 2) how does sample size affect information content for different taxonomic resolutions; and 3) what are the most cost-effective combinations of taxonomic resolution and sample size for models across a range of budgets? We use weighted averaging regression models for pH, phosphorus, salinity and lake depth to determine model information content, measured as R² and root mean squared error for predicted versus measured environmental variables in independent datasets. We then use realistic data collection costs to examine relationships between cost and model information content across budget scenarios.

Results/Conclusions

Investment in taxonomic resolution is more cost-effective than investment in additional samples, across all tested variables and nearly all conceivable budget scenarios. Cost savings at lower taxonomic resolutions do not allow enough additional samples to be collected to overcome sacrifices in information content. Information content exhibits greatest gain during initial sample sizes increases, reaching an asymptote at or before ~100 samples for all variables. Smaller sample sizes can achieve surprising predictive power in some cases (e.g. mean r² 0.68 between predicted and measured pH, using randomly-selected 10-sample model datasets), suggesting low-cost ecological models for rapid bioassessments may be achievable. However, caution is necessary when using such an approach, because spatial dependencies in predictions may be missed and ecological interpretation may be hampered by poor analogues with predicted assemblages.

We demonstrate the utility of explicitly considering cost estimates to determine optimal sampling effort and taxonomic resolution for ecological assemblage models. For regional biomonitoring programs using diatoms, cost-effective sampling could save millions of dollars that could subsequently be used for managing impacted ecosystems. Our general framework for determining optimal trade-offs in ecological assemblage models is easily adaptable to a variety of ecological assemblages used in biomonitoring and environmental assessment.