Regional seas are exceptionally vulnerable to climate change, yet are the most directly societally important regions of the marine environment. The combination of widely varying conditions of mixing, forcing, geography (coastline and bathymetry) and exposure to the open-ocean makes these seas subject to a wide range of physical processes that mediates how large scale climate change impacts on these seas' ecosystems. In this paper we explore these physical processes and their biophysical interactions, and the effects of atmospheric, oceanic and terrestrial change on them. Our aim is to elucidate the controlling dynamical processes and how these vary between and within regional seas. We focus on primary production and consider the potential climatic impacts: on long term changes in elemental budgets, on seasonal and mesoscale processes that control phytoplankton's exposure to light and nutrients, and briefly on direct temperature response. We draw examples from the MEECE FP7 project and five regional models systems using ECOSMO, POLCOMS-ERSEM and BIMS_ECO. These cover the Barents Sea, Black Sea, Baltic Sea, North Sea, Celtic Seas, and a region of the Northeast Atlantic, using a common global ocean-atmosphere model as forcing. We consider a common analysis approach, and a more detailed analysis of the POLCOMS-ERSEM model. Comparing projections for the end of the 21st century with mean present day conditions, these simulations generally show an increase in seasonal and permanent stratification (where present). However, the first order (low- and mid-latitude) effect in the open ocean projections of increased permanent stratification leading to reduced nutrient levels, and so to reduced primary production, is largely absent, except in the NE Atlantic. Instead, results show a highly heterogeneous picture of positive and negative change arising from the varying mixing and circulation conditions. Even in the two highly stratified, deep water seas (Black and Baltic Seas) the increase in stratification is not seen as a first order control on primary production. The approaches to downscaled experiment design and lessons learned from the MEECE project are also discussed.