Modelling the spatiotemporal growth dynamics of Haddock in Icelandic waters in RTMB

Growth is central to stock productivity and as such reliable estimates of growth are essential for assessing fish stocks, modelling ecosystems and developing science-based management plans. This thesis develops a hierarchical spatio-temporal von Bertalanffy growth model (VBGF) for Icelandic haddock (Melanogrammus aeglefinus), implemented in RTMB. Asymptotic length (L∞), growth rate (k), and theoretical age at zero length (t₀) are estimated jointly as correlated latent Gaussian random fields, with a non-stationary barrier SPDE constraining spatial correlation across Iceland's fragmented coastline, natural cubic splines for bottom temperature and depth, and a covariate for early-life nursery density. Individual length-at-age observations from the MFRI spring and autumn groundfish surveys (1985–2025) were used for fitting and validation. Nine candidate models were compared sequentially by AIC. The best-fitting model combined correlated AR(1) spatio-temporal fields on L∞ and k. Spatial and inter-annual structure accounted for larger improvements in fit than the measured environmental and density covariates combined. Estimated mean asymptotic length was 76.0 cm (spring) and 77.1 cm (autumn), with growth rates of 0.247 yr⁻¹ and 0.235 yr⁻¹. Nursery density was negatively associated with growth rate, reducing it by up to 15% across its observed range, and bottom temperature was positively associated with growth rate across the full range sampled. L∞ and k were strongly negatively correlated at both the baseline spatial scale (−0.932) and the inter-annual scale (−0.962), describing faster growth to a smaller asymptotic size inshore and slower growth to a larger asymptotic size offshore. The model predicted length-at-age accurately under spatial interpolation (R² = 0.913–0.937) and temporal forecasting beyond the training period (R² = 0.920–0.942). Forecast accuracy varied by year and declined over the later forecast horizon, consistent with the estimated two-to-three-year temporal persistence. The results show that haddock growth in Icelandic waters is spatially and temporally varying but predictable.