Ecological Survey Design Notes
Context and goals of the lecture
Focus on decisions researchers must make before, during, and after fieldwork related to ecological survey design and field assessments.
Grounded in practical principles rather than lots of journal articles; emphasis on key literature and foundational concepts.
Suggested core references:
Quinn and Kearns? (incorrect spelling in transcript) but cited: experimental design and data analysis for biologists – a foundational text (PDF available via library).
Elliot (papers related to benthic and coastal ecology).
Kingsford and Battershill – chapters on temperate marine environments, rocky shores, surveys, and sampling design.
Preview of upcoming topics: monitoring of disturbed/disturbed habitats, actual experimental design, and data analysis.
Core aim of surveys in ecology
Surveys aim to describe spatial and temporal patterns as accurately as possible (stock assessments, diversity estimates, community structure).
Basis for state-of-the-environment monitoring and fisheries management (regional councils, catch allowances).
Real-world constraint: budgets are finite; need to balance accuracy and cost.
Sampling unit should represent a usable portion of the ecosystem to tell a true story of what exists elsewhere in the system.
Linking questions to methods
Type of question dictates methods (e.g., census vs general surveys):
Census: often used to survey individuals, common in social sciences (psychology, pre-election polls).
In marine benthic and ventic ecology, focus on distribution and abundances of organisms.
Important to have a representative sample to avoid biased inferences.
Over-arching design decisions are driven by the question and prior knowledge about the area.
Key concepts: accuracy, precision, and bias
Accuracy: closeness of the estimate to the true value.
Precision: variability of repeated measurements (how tight the distribution around the mean is).
Bias: systematic error that shifts estimates away from the true value.
Trade-offs:
High accuracy with low precision can still mislead if variability is large.
High precision with low accuracy means estimates are consistently wrong.
How to improve accuracy and precision:
Increase replication (more samples) to narrow the spread and reduce SE.
Improve sampling design to remove biases.
Conceptual examples (stock/biomass):
Case 1: Accurate and high precision – estimates cluster tightly around the true value.
Case 2: Accurate mean but low precision – tight around a mean that is close to true value but with wide tails elsewhere.
Case 3: High precision but biased – estimates tightly clustered but far from the true value.
Case 4: Biased and imprecise – both mean is off and high variability.
Key takeaway: design to minimize bias, reduce error, and lower variability through robust sampling.
Quantitative links:
Standard error of the mean: where is the population or sample standard deviation and $n$ is the number of replicates.
Accuracy and precision are about how close estimates are to true values and how tightly repeated estimates cluster, respectively.
Fundamental design choices based on questions
Your sampling design depends on the area of interest, habitat heterogeneity, and the question:
Random sampling (unstratified) tends to produce less precision in heterogeneous systems but is unbiased overall.
Stratified random design: divide area into strata (habitat types) and sample within each stratum; allows better representation of heterogeneity and targeted inference.
Systematic sampling (gridded/spatially regular): easy to implement and can be representative, but risks aliasing if there is periodicity or cyclical patterns.
Systematic with potential bias: starting point matters; randomization reduces bias.
Nested sampling: use different sampling intensities for different organism sizes or densities within a larger sampling unit.
Before–after control–impact (BACI) designs: baseline data collected before an activity, control and impact areas compared to assess change; Megan to cover in more detail later.
Before fieldwork, you should decide on the spatial and temporal scope, sampling units, and replication strategy, guided by prior knowledge and pilot data where possible.
Approaches to allocate sampling effort
Random design (simple random): pick random coordinates within the study area; equal chance of selection.
Pros: unbiased sampling across area; simple to implement.
Cons: may yield clustering in some zones; may under-sample heterogeneous areas.
Stratified random design: partition area into strata (habitats) and allocate samples across strata, often proportionally to area (
Proportional allocation: allocate samples to strata in proportion to area size.
Notation: let $Ai$ be area of stratum i, total area , and total samples $n$; then
for each stratum i.
Example: 30% intertidal, 50% subtidal, 20% rocky reef with total $n = 10$ samples leads to
, , .
Pros: better representation of habitat heterogeneity; more efficient use of resources.
Considerations: some strata may be more variable; you can further weight sampling toward more variable strata (adaptive weighting) after pilot survey.
Unequal/optimized stratification (optimal allocation): allocate more samples to strata with higher variability to maximize precision.
Systematic sampling within strata: grid-based sampling at regular intervals; advantages in consistency, but risks aliasing if temporal or tidal cycles are not aligned with sampling times.
Systematic sampling pitfalls (aliasing): sampling at fixed intervals can misrepresent cyclical/periodic processes (e.g., tides, recruitment seasons) if not synchronized with these cycles.
Systematic sampling and equal opportunities: ensure each area has an equal chance of inclusion to preserve statistical validity; otherwise, randomization is necessary.
Preliminary sampling and site familiarization
Before large field campaigns, perform preliminary (pilot) sampling to:
Assess area and habitat diversity (identify distinct habitats, e.g., intertidal, subtidal, rocky reefs).
Gauge how many samples are needed to capture habitat heterogeneity and distribution patterns.
Check logistics: access routes, equipment needs, team size, budget, and time constraints.
Determine the size and density distribution of the target organisms; understand which sampling units are appropriate (e.g., core sizes for benthic sampling).
Pilot data informs decisions about stratification levels, sampling unit sizes, replication needs, and whether a more intensive approach is warranted in particular habitats.
Sampling units, spatial scale, and organism size considerations
The sampling unit should be representative of the ecosystem portion being studied.
Practical considerations:
How many sampling units to use within the area? More units generally improves precision and coverage.
The size of sampling units (e.g., quadrats, cores) influences the likelihood of capturing organisms of interest.
The organism of interest should typically constitute a small fraction of the sampling unit (rough guideline: about 5% or less of the unit size) to avoid oversampling a single patch and biasing results.
Examples by sampling unit type:
Sediment cores for small organisms (e.g., benthic invertebrates): choose core diameter and depth to match target organism size and density.
Larger organisms (e.g., echinoids) may require larger cores or different sampling methods to avoid zero or single captures in a core; poor sampling units reduce information on density and distribution.
Nested sampling considerations:
For scarce/mobile organisms, evaluate within larger quadrats; for abundant/patchy organisms, sample smaller areas within the larger area to capture variation without excessive effort.
Be cautious when extrapolating nested-nested results to broader regions due to zonation and habitat specificity.
How many samples? A practitioner-friendly view of power and replication
More samples generally increase precision and reduce SE, but budgets constrain replication.
A classic practical example (transect-based fish surveys):
Researchers test different transect lengths and numbers to determine a “sweet spot” where the standard error relative to the mean plateaus.
They observe a curve: increasing sampled area initially improves precision, but after a point, returns diminish (plateau).
Decision example from the slide: with transects of 25 m length and 5 m width, using 6 replicates achieved a plateau in SE/mean, suggesting sufficient design for balanced precision and effort.
Implication for planning:
Use pilot data to perform a power-like analysis to figure out how many transects and what length yield robust estimates without over-sampling.
Power analysis and power curves help identify diminishing returns and justify the chosen sample size.
Multi-species sampling and nested designs
When multiple species are measured within a single sampling unit, consider nested sampling strategies:
For rare/mobile species, sample within a larger quadrat or area to capture sparse occurrences.
For common/dense organisms (e.g., barnacles on a rocky shore), sample within smaller areas to manage counting effort and avoid oversampling dense patches.
Cautions with nested designs:
Patchy distributions and zonation can limit extrapolations; patterns observed in a small area may not hold in other zones.
Different habitats (exposure, depth, gradient) may host different communities; hierarchical sampling should respect these boundaries.
Practical rules and planning takeaways
Before fieldwork:
Gather existing site knowledge and prior data to inform sampling strategy (avoid reinventing the wheel if the site is well-studied).
If knowledge is lacking, perform preliminary sampling to estimate habitat distribution and organism variability.
Verify the suitability of sampling units (size, density, distribution) and the feasibility of replication given time and budget.
General design guidance:
Large-scale spatial variation often benefits from stratification by habitat type.
Verify that sampling units are appropriate for the target organisms and that the sampling effort aligns with the study’s aims.
Plan replication thoughtfully; more replication generally improves precision, but budgets and logistics are real constraints.
Maintain randomization to avoid biases; randomization should be used in selection of sampling units or within strata when feasible.
When to use which approach:
If area is homogeneous and uniform, simple random design may suffice.
If there are distinct habitats or gradients, stratified random with proportional allocation is advantageous.
If there are known periodic patterns (tidal cycles, seasonal recruitment), consider systematic designs with timing adjusted to capture those patterns or use intensified sampling around expected events.
If time-series or management interventions are involved, BACI-like designs may be appropriate to detect changes.
The practical message: there is no one-size-fits-all design; the best approach uses pilot data, knowledge of habitat heterogeneity, and a clear link between the question and the sampling framework.
Design pitfalls to avoid (summary)
Ignoring prior site knowledge; assuming homogeneity when it is not.
Over-reliance on a single sampling method or unit size that misses key organisms or habitats.
Systematic sampling without accounting for periodic processes (aliasing) and with a risk of missing important phases.
Underestimating the importance of randomization and replication; biased site selection or too few replicates inflate error and bias.
Extrapolating results from patchy/habitat-specific patterns to broader regions without considering zonation.
What’s next in the course (context)
Megan will cover monitoring in disturbed habitats and impacted sites.
Later lectures will delve into experimental design specifics and data analysis for the field data collected.
The Rocky Shore data analysis labs will give hands-on practice with the concepts discussed here.
Quick glossary of terms from today
Census: a survey of all individuals in a population; often used in social sciences.
Random design: each sampling unit has an equal probability of being selected.
Stratified random design: area divided into strata (habitats), with random sampling within each stratum; allocation can be proportional to area or optimized by variance.
Proportional allocation: allocate sampling effort to strata in proportion to their area: .
Optimal allocation: allocate more samples to strata with greater variability to maximize precision.
Systematic sampling: sampling at regular intervals (e.g., every x meters); risks aliasing if not aligned with periodic processes.
Systematic yet randomized: introduce random starting points to reduce bias.
Nested sampling: using different sampling scales for different targets within a larger sampling unit.
BACI design: Before-After-Control-Impact design to assess environmental impacts of activities.
Accuracy: closeness of estimates to the true value.
Precision: variability of repeated measurements (tightness around the mean).
SE (standard error): for the mean; describes precision of the estimate.
Patchiness and zonation: uneven distribution of organisms across habitats; important for interpretation of nested designs.
Aliasing: misinterpreting cyclic patterns as real trends due to sampling at inappropriate times or intervals.
Pilot sampling: pre-survey sampling to inform methodology, logistics, and feasibility.