Interpreting a Trawl Survey with Confounded Sampling
A fisheries agency surveys a demersal stock with a trawl net over six years. In years 1 to 3, the same vessel and net are used, sampling 40 stations per year. In year 4 a new, larger net with a 15 percent higher catchability is introduced on a different vessel. From year 4, only 25 stations are sampled because of budget cuts, and sampling is shifted two weeks earlier in the season. Catch per unit effort (kilograms per hour) is: year 1, 42; year 2, 39; year 3, 36; year 4, 30; year 5, 33; year 6, 35. Independent egg surveys suggest spawning biomass fell steadily by about 4 percent per year throughout. A manager concludes the stock has recovered since year 4 because catch per unit effort has risen.
1. Critique the manager's conclusion, identifying each source of bias and its likely direction.
Read the worked answer
Critique. The manager reads the rise from 30 to 35 kilograms per hour as recovery. The survey is not comparable across the break at year 4, so the trend is confounded.
Sources of bias and direction.
- Gear change (year 4 onward): a net with 15 percent higher catchability should raise catch per unit effort by roughly 15 percent for the same abundance. This inflates years 4 to 6 relative to years 1 to 3, so the apparent recovery is partly an artefact.
- Vessel change: a different vessel tows differently, affecting speed, net spread and herding; direction is unknown but it breaks comparability.
- Reduced stations (40 to 25): fewer stations increase sampling variance, so year-to-year differences are less reliable and a rising trend could be noise.
- Earlier season: fish distribution and catchability vary seasonally; the direction depends on the species, but it is a systematic shift that can create a spurious trend.
- Contradictory independent data: the egg survey indicates a steady 4 percent annual decline in spawning biomass, which directly contradicts recovery.
Judgement. The manager's conclusion is not supported. The rise in catch per unit effort coincides with changes that would inflate it, and the independent egg survey points the other way. The most defensible reading is continued decline, masked by increased catchability.
2. Propose how the agency should analyse the data and redesign the survey so that future trends are defensible.
Read the worked answer
Analysis of existing data.
- Do not treat the series as a single continuous index. Analyse years 1 to 3 and years 4 to 6 as separate segments, and report the break explicitly.
- Apply a catchability correction to years 4 to 6 (divide by about 1.15) to make them roughly comparable, and present corrected and uncorrected values with uncertainty.
- Report confidence intervals; with 25 stations the intervals will be wide, so test whether the apparent rise is statistically distinguishable from zero.
- Calibrate against the egg survey, which is independent of gear, and use it as the primary abundance index if the trawl series is unreliable.
- Check for a vessel effect by comparing overlapping tows if any exist.
Survey redesign.
- Run old and new gear in parallel for at least one year (paired tows) to estimate the catchability ratio empirically rather than assuming 15 percent.
- Standardise vessel, gear, tow speed, season and station set; if change is unavoidable, maintain an overlap period.
- Use a stratified random design so reduced station numbers still give representative, unbiased estimates with known variance.
- Fix the seasonal window and do not shift it mid-series.
- Predefine indicators and decision rules so future managers cannot misinterpret an artefact as recovery.
Judgement. With calibration and standardisation, the survey can support defensible trends; without them, the egg survey should carry more weight.