We present a bottom trawl survey (BTS) field experience carried out off the Portuguese Continental shelf to test two sampling designs proposals previously analysed by simulation which implement a hybrid random-systematic and a systematic sampling strategy. We used a common base regular grid covering the survey area and overlapped it with the existent random design to build the hybrid design while the systematic design added a set of regular locations at smaller distances creating four denser sampling areas. We use hake (Merluccius merluccius) abundance and model-based geostatistics to compute measures such as mean abundance, µ, and the 95% percentile, p95, which summarise the areal behaviour; coverage of the prediction confidence interval, ξ, to assess the adequacy of the model; and a modified generalised cross validation index, p, to evaluate prediction precision. The hybrid design showed a lower coefficient of variation for µ (11.89% against 13.25%); a slightly higher coefficient of variation for p95 (11.31% against 11.09%); similar ξ (0.94); and lower π (16.32 against 18.82). We conclude that the hybrid design performs better, our procedure for building it can be used to adjust BTS designs to modern geostatistical techniques, and the statistics used constitute valuable tools for assessing BTS performance.
|