Continuing with my Climate Change in Fisheries Series (CCFS), here is my 6th on Incorporating climate scenarios in fisheries research.
I had never met Dr Inna Senina before, but I knew about her SPC work and impressive expertise in modelling population dynamics.
Her reputation is not undeserved. Her presentation was impressive and methodologically delivered, with the great graph above illuminating the “Conceptual model of population dynamics” that made a lot of sense to me and tied up various ideas in an elegant image.
She elucidated how the model estimates tuna habitats, biomass distributions, and abundance and explained the differences among the four target tuna species. She also discussed existing uncertainties in modelling tuna population dynamics, the projected biomass redistributions under climate change, and related implications for the Pacific Island Countries and Territories.
The presentation was structured to:
explain the dynamic model for tuna population dynamics,
describe how tuna abundance and spatial distributions are estimated, and
describe the climate change impacts on Pacific Islands' and WCPFC tuna stocks as predicted by quantitative models.
In the first part, the conceptual view of the model was provided and discussed the complexity of the minimal model, enabling a description of tuna population dynamics and the study of the impacts of climate change on tunas. The importance of including spatial dynamics was noted as a description of reproduction, mortality, and movement, all of which are influenced by environmental conditions such as primary production, temperature, oxygen, food resources, and the absence of predators. It was noted that different behaviours of juveniles and adults dictate the need to consider age and life stage structure. On the other hand, how the model could be simplified by considering a two-dimensional space was presented, focusing on three distinct layers based on the micronekton vertical distribution. The model presented feeding habitat as the accessible forage biomass that drives tuna movement. It also explained the different types of movements, including directional and non-directional movement and their effect on tuna distributions. Real-world examples were used to illustrate the model’s performance.
The importance of parameterising the model informing dynamic rates and habitat parameters from observations, validating the model, and addressing potential uncertainties was highlighted. The use of different types of data in estimating abundance, spatial distribution, and movement of tuna species was discussed. The presentation highlighted the limitations of each data type, such as biases in fisheries data and the difficulty to accurately estimate spatial distribution and movement from catch and length frequency data alone. The impact of data on the predicted distribution and movement of a species, using the model of skipjack tuna was illustrated as an example, and discussed the uncertainties in the estimation process, including data coverage, ocean forcing variables and model structure.
The effects of climate change on the four main Pacific tuna populations was discussed, and its implications for Pacific Islands countries and territories. She mentioned using different forcing models to predict the warming under RCP 8.5 and 4.5. The biomass and catch decreases under RCP 8.5 and have less severe impacts under RCP 4.5. The model showed that stocks were stable for the first half of the century but decreased in late 2040, particularly for the skipjack and yellowfin tunas. Using the skipjack example, she attributed this to the reduced availability of mesopelagic micronekton, which migrates at night to the surface and represent a key food resource of skipjack tuna in the warm pool area. The uncertainties in predicting the effects of climate change on tuna populations was discussed, emphasising the importance of including uncertainties in projecting temperature and primary production. Despite model uncertainties, agreement between different models on distributional shifts suggests that changes are inevitable. Highlighting the need to collaborate with others to address these general questions.
Discussion
It is currently unknown weather light levels or availability within the ecosystem will change. It is also not known what the long-term changes to the depth of the thermocline may be.
Due to the accumulated uncertainties, the projections stop in 2050 as the further into the future you go the larger the compounded error. In addition, most of the interest is currently for the closer time frames, i.e. up to 2050.
The biggest source of uncertainty is ocean forcing bias and the other large unknown is the observations of the biology many of which are missing. The model is age-structured and starts at age 0 to capture recruitment trends. The survival of larvae is predicted and information on recruits are included from the fisheries data. The larval distribution data are predicted as the larvae are distributed by currents and influenced by temperature and the density of other plankton, as well as micronekton at the surface, which results in larval mortality. Larval observations are sparse and getting more observations would be beneficial for the model. In addition, observing the fish movement for a longer age window than we currently have would be helpful. For example. for bigeye tuna only fish aged 1-2 are tagged in large numbers and the model would benefit from data for older fish.
The modelling can be EEZ-specific, but if there is strong connectivity between the EEZ and outside, it's not useful. The model can be easily separated along obvious natural boundaries, such as the equator. Once a model is developed, one can zoom into the EEZ, but if the EEZ is small, then it may not be that beneficial.
Model timeframe relies on a projection window of the modelled data, so the time scale for a modelled period can be shortened e.g. if the forcing data are available at a 1 year level then the model can be refined down to a single year. The model does get validated by looking back in time to see how well it goes when compared to existing ocean data.
SEAPODYM will be extended to the Atlantic and Indian Oceans, and once that is done, interocean data can be compared.
If the fish population is reduced by fishing, the prey limiting aspects of the model are still valid as the model is linked to the absolute size of the biomass. Fishing vessel acoustic data can be used in the models, but they are very sparse so are not widely used currently, but they could be used in future if more data become available.
Currently, all the modelling is done on tuna, but modelling more spatially restricted species, such as non-migratory species, could be included in the future.
Conclusions
Despite model uncertainties, agreement between different models on distributional shifts suggests that it is not a question of ‘IF’ the tuna biomass will shift due to climate change from the Pacific SIDS EEZs but ‘WHEN’ and ‘TO WHAT EXTENT’.
Moderate redistributions of tuna under a lower-emissions scenario indicate that reductions in greenhouse gas emissions, in line with the Paris Agreement, would provide a pathway to sustainability for tuna-dependent Pacific Island economies.
Quantitative (Predictive) modelling of fish population dynamics requires data to observe all modelled dynamic processes and a realistic description of the tuna environment on historical, decadal, and climate timescales.
Ongoing and future work is dedicated to reducing uncertainties linked to the model structure and parameter estimation and to providing better quantification of uncertainties related to climate modelling.