gadget3: An evolution of the gadget modelling framework

Fisheries systems vary widely in ecological processes, data availability, and management objectives, meaning that no single modelling structure is universally appropriate. Modellers frequently face limitations when adapting standardised frameworks to represent specific biological processes or to incorporate novel data types, while custom models tailored to individual systems can be difficult to reuse, maintain, or extend. Here we present gadget3, a new process-orientated modelling framework designed to address these challenges in modelling fish populations and fisheries dynamics. gadget3 is a complete re-implementation of the original GADGET framework, retaining its modular philosophy while extending and generalising its capabilities. gadget3 takes advantage of the R ecosystem of model optimisation utilities, particularly the Template Model Builder (TMB) library, enabling efficient parameter estimation through automatic differentiation and allowing advanced statistical features such as random effects. We outline the structure of a gadget3 model, which is assembled from modular components representing biological processes, fleet interactions, and likelihood components. Through the use of metaprogramming and composable functions, users can implement alternative formulations or entirely new processes while maintaining compatibility with the broader modelling structure. Examples are presented demonstrating how model behaviour can be adapted through custom formulations, allowing the incorporation of system-specific biological relationships, environmental drivers, or alternative recruitment and mortality functions. Finally, we discuss ongoing and planned developments, including expanded modelling capabilities, improved diagnostics, and enhanced support for complex multi-species and ecosystem applications. Together, these features position gadget3 as a flexible and extensible framework for developing next-generation fisheries and ecosystem models.