PURSESEINE CATCH ESTIMATES DETERMINED FROM OBSERVER GRAB SAMPLES
PURSE-SEINE CATCH ESTIMATES DETERMINED FROM OBSERVER GRAB SAMPLES CORRECTED FOR SIZE SELECTIVITY
MFCL PURSE-SEINE CATCH DATA — STANDARD PRACTICE LOGSHEETS SKJ ↑ ANNUAL CATCH ESTIMATES YFT+BET ↓ GRAB SAMPLES UNCORRECTED Total Catch of SKJ + YFT + BET SKJ ↓ YFT ↑ BET ? S_BEST GRAB SAMPLES UNADJUSTED UNCORRECTED 1 x 1 x Month YFT and BET in YFT+BET SKJ ↑ YFT+BET ↓ S_BEST ADJUSTED 1 x 1 x Month SKJ ↑ YFT ↓ BET ↓ MFCL INPUT DATA MFCL Area x Quarter SKJ ↑ YFT ↓ BET ↓
GRAB SAMPLES CAN BE CORRECTED FOR SIZE SELECTIVITY BIAS GRAB SAMPLES UNCORRECTED SKJ ↓ YFT ↑ BET ? PAIRED SPILL AND GRAB SAMPLES Size Selectivity Bias GRAB SAMPLES CORRECTED SKJ YFT BET
MFCL PURSE-SEINE CATCH DATA — ALTERNATE PRACTICE LOGSHEETS SKJ ↑ YFT ↓ SKJ ↑ YFT+BET ↓ ANNUAL CATCH ESTIMATES GRAB SAMPLES UNCORRECTED Total Catch of SKJ + YFT + BET SKJ ↓ YFT ↑ BET ? S_BEST GRAB SAMPLES UNADJUSTED UNCORRECTED 1 x 1 x Month YFT and BET in YFT+BET SKJ ↑ ↑YFT+BET YFT ↓ ↓ PAIRED SPILL AND GRAB SAMPLES Size Selectivity Bias S_BEST GRAB SAMPLES ADJUSTED CORRECTED 1 x 1 x Month SKJ YFT BET SKJ ↑ YFT ↓ BET ↓ MFCL INPUT DATA MFCL Area x Quarter SKJ ↑ YFT ↓ BET ↓ MFCL INPUT DATA ADJUSTED MFCL Area x Quarter SKJ YFT BET
Status of Paired Grab and Spill Samples on Purse Seiners From 2008 to mid 2010 • Data for 17 trips were used in analysis presented at SC 6 (2010) • Data for 4 trips were unusable • Still missing data for 4 trips in Solomons in 2009 Mid 2010 to now • Seven trips completed: US x 2, KR x 2, TW x 3 • Data from 3 are unusable: KR x 2, TW x 1 • Data from 4 should be usable. Next? • EU = Spain : No response to proposal following SC 6 • Japan : Currently under negotiation • Others : More KR? TW? New FSM, Marshalls, Vanuatu? China, Philippines? • Data Collection Officer will be recruited to implement paired samples
Progress with the design for collecting paired samples proposed in SC 5 -ST-WP 6 Vessel Nationality / Arrangement Target # Trips End of 2010 FSM Arrangement 8 15 (all in PNG) China 2 0 Japan 6 0 Korea 8 2 bad New Zealand 2 4 bad Philippines 2 0 Solomon Islands 2 4 missing Chinese Taipei 8 1 bad + 2 good United States of America 8 4 Vanuatu 2 0 EU and EPO-based fleets 2 0 TOTAL 50 25 + 7 bad = 32
Status of Analyses of Paired Grab and Spill Samples 2010 • Grab sample selectivity bias was estimated from paired samples collected during 17 trips. • Historical grab samples were corrected for selectivity bias. • MFCL catch data were generated with species compositions determined from corrected grab samples. 2011 • Estimates of selectivity bias will be updated, pending the availability of data. • MFCL catch data will be updated. • MFCL length data will be generated with grab samples corrected for selectivity bias.
Grab Sample Size Selectivity The length frequencies determined from the spill samples (red) and the grab samples (green) show that the grab samplers miss the smaller fish.
The paired spill and grab samples can be used to quantify the selectivity bias in grab samples. See ST-WP-2, equation (1).
Data covering very large fish are insufficient to reliably estimate the selectivity bias, but the use of cubic splines suggests that it is similar to very small fish.
In 2010 the estimates of selectivity bias were used to correct the species compositions determined from grab samples collected from 1995 to 2009, which were then be used to adjust the purseseine catch data. MFCL input catch data, 1995 -2009, unadjusted (left) and adjusted with grab samples corrected for selectivity bias (right):
Discussion • The estimation of selectivity bias assumes that the species compositions determined from spill samples are unbiased. Comparison with cannery receipts and port sampling of landing categories for trips for which both spill samples and port samples were collected may give us more confidence in this assumption. • During the 17 trips analysed, grab samples were collected by 14 observers. Additional paired samples with different grab samplers are required to obtain estimates of selectivity bias that are representive of the population of observers.
Discussion (continued) • Additional paired samples are required to obtain reliable estimates of selectivity bias for fish less than 35 cm and greater than 54 cm; paired samples of large fish in unassociated schools would be particularly useful. • The analyses have assumed that selectivity bias depends only on the size of fish; other variables — such as species, school association and the observer — should also be examined as the data permit; again, paired samples from unassociated schools would be particularly useful.
- Slides: 13