A simulation framework for evaluating multi-stage sampling designs in populations with spatially structured traits

Area of Science:

• Ecology
• Population Biology
• Statistical Ecology

Background:

• Biological surveys require efficient sampling strategies, especially for spatially structured populations.
• Multi-stage sampling involves selecting sites and then sub-sampling individuals for trait data.
• Sub-sampling strategies significantly impact population estimates used in management and conservation.

Purpose of the Study:

• To develop a simulation framework for evaluating sub-sampling strategies in multi-stage biological surveys.
• To quantitatively compare the precision and bias of random versus stratified sub-sampling designs.
• To assess the impact of sub-sampling on population estimates for a virtual fish population.

Main Methods:

• Developed a simulation framework to model multi-stage sampling with sub-sampling.
• Compared random and stratified sub-sampling strategies for age data collection.
• Applied the framework to a virtual Pacific cod population from the Eastern Bering Sea.
• Incorporated various error sources and analyzed sensitivity of population estimates.

Main Results:

• Stratified sub-sampling struggled to reproduce spatial patterns of individual traits.
• Differences between strategies were minimal with large sub-sample sizes.
• Random sub-sampling showed better precision across tested scenarios.
• Spatial autocorrelation significantly contributed to trait estimate errors, irrespective of the sub-sampling design.

Conclusions:

• The choice of sub-sampling strategy critically affects the accuracy of ecological survey estimates.
• Random sub-sampling can be more precise, particularly for spatially structured traits.
• The developed simulation framework is valuable for assessing populations with spatially structured traits in multi-stage designs.