Most contemporary stock assessment models include age structure, although some include size instead of age, particularly for hard-to-age species such as crustaceans and shellfish. Few include both size and age because of computational or data requirements. Most contemporary stock assessments for the important tuna stocks are age-structured, although some use biomass dynamics models and a few length-based models have been developed. Including age (or length)-structure is important for many tuna stocks because different fleets catch widely different aged tuna, which may differ from the ages represented by the index of abundance and spawning biomass.
Most tropical tuna stocks have little, if any, reliable age data. Therefore, there is no time series of age data as used in VPA or Statistical Catch-at-Age analysis, unless length composition data is converted into age using cohort slice or similar method. Similar issues arise with other data, such as tagging, are available by length, but not age. Aging data is becoming more available for tropical tunas, but controversy remains on it reliability for all species. The lack of aging data and the possibility that population (e.g., natural mortality, growth, and movement) and fishery processes are length-based, suggest that length-based models may be more appropriate.
Age-based models that are fit to length-composition data typically assume a constant function for length-at-age (e.g., the normal distribution) and therefore length-base processes do not change the distribution of length-at-age. However, strong length-based processes, particularly selectivity when fishing mortality is high or a minimum legal size is used, can affect the distribution of length-at-age and should be taken into consideration. This requires modelling both age and length dynamics (the model has to explicitly keep track of the age and length of fish increasing the dimensions of the model) considerably increasing the computational and data requirements. Although, approximations such as platoons in Stock Synthesis can reduce the computational demands.
Spatial variation in growth is apparent in tuna length data and may need to be accounted for in stock assessment models. Assuming the same growth curve in current spatial tuna stock assessments can result in different depletion levels among areas simply due to fitting to the length-composition data. However, modelling spatial variation in growth within age-structured spatial stock assessment models becomes complicated since the length-at-age changes as a fish moves within the model from one area to the other. Modelling both age and length dynamics can allow for fish to retain the same age and length as it moves. However, estimating the growth parameters and determining if growth is age-based (genetic) or length-based (environmental) may be difficult.
It is not clear if length-based dynamics are important for tuna stock assessment.
Do we need age-length dynamics (e.g., length based processes such as fishing mortality influence the distribution of length-at-age, or for modelling spatial changes in growth as fish move)?
Will age-length models greatly increase computational times and are there efficient methods to implement age-length models (in RTMB)?
Can a fully length-based model substitute for a age-length model?
Are approximations such as platoons good enough?
How do age-length models interact with spatial models (block transfer or fine scale spatial temporal dynamics assessment model)?
Can you turn off the length (age) dynamics and still be as efficient as an age (length) only model?
Will FIMS adequately implement age-length models in the medium term or do we need to create a RTMB model?
Akselrud, C.I.A., Punt, A.E., Cronin-Fine, L. 2017. Exploring model structure uncertainty using a general stock assessment framework: The case of Pacific cod in the Eastern Bering Sea. Fisheries Research, 193: 104–120. https://www.sciencedirect.com/science/article/abs/pii/S0165783617300814?via%3Dihub
Þór Elvarsson, B., Woods, P.J., Björnsson, H., Lentin, J., Thordarson, G. 2018. Pushing the limits of a data challenged stock: A size- and age-structured assessment of ling (Molva molva) in Icelandic waters using Gadget. Fisheries Research, 207: 95-109. https://www.sciencedirect.com/science/article/abs/pii/S0165783618301759
McGarvey, R., Feenstra, J.E., Methot, R.D., Doering, K. 2024. Performance gains from incorporating dynamic numbers by length-within-age in fishery assessment models. Fish. Res. 276, 107039 https://www.sciencedirect.com/science/article/pii/S0165783624001036
Punt, A.E., Akselrud, C.A., Cronin-Fine, L. 2017. The effects of applying mis-specified age- and size-structured models. Fisheries Research, 188: 58–73. https://www.sciencedirect.com/science/article/abs/pii/S0165783616304027?via%3Dihub
Taylor, I.G., Methot, R.D. 2013. Hiding or dead? A computationally efficient model of selective fisheries mortality. Fisheries Research, 142: 75-85. https://www.sciencedirect.com/science/article/pii/S016578361200269X
Zhang, F. Cadigan, N.G. 2022. An age- and length-structured statistical catch-at-length model for hard-to-age fisheries stocks. Fish and Fisheries, 23: 1121–1135. https://onlinelibrary.wiley.com/doi/10.1111/faf.12673