Ageing Outputs in Stock Assessments in Queensland - Focus on Fisheries Concerns Moving the Technology Forward

Abstract

The Queensland Department of Agriculture and Fisheries, Australia (DAF) routinely collects length, age, and reproductive information for 15 species (Lates calcarifer, Polydactylus machochir, Lutjanus erythropterus, L. malabaricus, L. carponotatus, L. sebae, Lethrinus nebulosus, Platycephalus fuscus, Acanthopagrus australis, Sillago ciliata, Pagrus auratus, Glaucosoma scapulare, Mugil cephalus, Scomberomorus commerson, and S. munroi), processing in the order of 10,000 otoliths each year. Queensland has a well-established protocol for fish ageing. The exact ageing technique (whole vs. thin section) differs between species, but for all species includes reference collections, annual reader training and qualification prior to expert reading of current samples, centralized data recording and quality control measures (e.g., precision in the form of repeat reads, IAPE, and bias). In 2015, DAF completed a ‘proof of concept’ study to apply Near Infrared Spectroscopy to age fish (http://www.frdc.com.au/Archived-Reports/FRDC%20Projects/2012-011-DLD.pdf). Results were promising and indicated that NIRS could provide cost-efficiencies in routine fish ageing. However, Queensland has not adopted NIRS to routinely age fish due to the moderate cost-efficiency savings and issues with age accuracy and precision. In particular, the fisheries end-users desired to understand what NIRS measures in fish otoliths that is correlated with age and whether the correlation was a function solely of time or was confounded by growth. NIRS is a secondary method of determination and relies on the accuracy of the reference samples, in this case the age based on expert visual interpretation of otoliths. Therefore, age samples used in developing NIRS calibration and validation models should be well understood for their inherent properties as these will influence and perpetuate through the NIRS predictive models. Inherent properties include age precision, accuracy and bias, as well any physical/chemical changes to the otolith that occur from differences in post-mortem handling (e.g., fresh vs. frozen) which may affect the NIR spectra. Additionally, fisheries biologists/scientists should be included in the spectra acquisition, processing and calibration model development, thus expanding the understanding of fisheries end-users of NIRS age estimates such that these estimates would not be viewed as ‘black box’ estimates. Predicted age of a fisheries species is not the end-product but is used in numerous subsequent analyses. Age estimates (as age frequencies) are used in catch curve analysis and hindcasting recruitment indices of cohort/year-class strength. ‘Old fish’ are often poorly estimated for age using NIRS and could be treated as outliers or as relatively unimportant because they occur at low frequency in fishery samples, fishery harvests or biomass estimates. However, such ‘outliers’ are important as they indicate that the NIRS model doesn’t fit as it should and we should strive to understand why the model doesn’t fit rather than exclude outliers (without good reason) from the calibration/validation dataset. Old fish (or the lack of them) can be an important indicator of age-truncation in heavily fished species and therefore should be important in NIRS model-fitting considerations. Young fish also may be poorly estimated for age using NIRS. In fisheries with highly variable recruitment, accurately aged young fish can indicate likely pulses in recruitment to the exploitable biomass, which may have implications for setting harvest limits (e.g., quotas). Age estimates are used in growth curves and are fundamental to stock assessment. Age (and ageing error) are just one of multiple data inputs into stock assessment. Precision and accuracy associated with NIRS estimates of fish age should be documented, but it should be kept in perspective that other data inputs to stock assessments also have error (and assumptions and biases). The Strategic Initiative
development team have wisely included fisheries assessment scientists in operationalizing FT-NIRS for age estimation.