Warming seas around the UK and the wider north-east Atlantic could destabilise marine food webs, weaken ocean health and jeopardise food security by 2050, according to a four-year study.

Scientists who projected the impact of climate breakdown on four key groups of marine life crucial to maintaining ocean ecosystems in the region said their results challenged assumptions that in middle latitudes, such as the UK, marine biodiversity was likely to increase across the board.

Instead, their projections highlight a predator-prey imbalance that could “fundamentally reshape marine ecosystems” in unforeseen ways, they said.

The seas around the British isles, particularly the southern North Sea, have been identified as “marine climate hotspots”, one of 20 globally to have warmed the fastest over 50 years.

The new predictions, under a scenario of 3C (5.4F) of warming by 2100, found that while fish and seabed invertebrates may become more diverse by mid-century, key plankton groups lower in the food chain were likely to decline. This imbalance, the authors of the report said, could disrupt marine food webs and weaken the health and resilience of ocean ecosystems.

“What surprised us is that warming doesn’t increase or decrease biodiversity across the board,” said Elena Couce, a marine ecologist at the Centre for Environment, Fisheries and Aquaculture Science (Cefas), a British government agency, and co-lead author of the study.

“While some groups higher up the food chain, like fish, may gain species in parts of the north-east Atlantic, key plankton groups that underpin marine ecosystems are projected to lose species,” she said.

The research, led by Cefas and published in Science Advances, involved an international team of researchers from the University of East Anglia, the Marine Biological Association, UiT the Arctic University of Norway, the Royal Netherlands Institute for Sea Research and the Technical University of Denmark.

Pleurobrachia, a zooplankton genus known as the sea gooseberry. Crucial zooplankton could decline by about 7% by mid-century, the study found. Photograph: Hannah Lloyd-Hartley/Cefas

“When things get hotter, all the species from the equator are going to move north and south,” Cource said. “The UK is in the middle latitudes. So we expected diversity to go up, because you get species from the south that are more diverse than the species you lose to the north.”

Previous Cefas research looking at key fish species suggested warming in north-western European waters would result in climate “winners”, such as anchovies, seabass and sardines. Such conditions, however, are less suitable for climate “losers”, including Atlantic cod, haddock and coley.

The current study, focusing on ecosystems in UK waters and the wider north-east Atlantic including Ireland and France, found that by mid-century, fish and seabed invertebrate diversity could increase by about 12%, whereas phytoplankton and zooplankton diversity could decline by about 3% and 7%, respectively by 2050.

“If predator diversity rises while prey diversity falls, that has the potential to disrupt the food chain,” said Lily Greig, a marine ecologist at Cefas who also co-led the study. “Our results show that climate change could reshuffle marine biodiversity unevenly across the food web, and that gains in some groups do not necessarily mean the ecosystem as a whole is becoming more resilient.”

By identifying where biodiversity was changing fastest and which parts of the ecosystem were most at risk, the study could provide an “early warning system” of potential ecosystem change, the researchers said. The work has already been submitted to the government’s review of marine protected areas, which is looking at fishery management measures.

“Policymakers have a mandate to try to preserve or even increase biodiversity,” said Couce. “What we are trying to tell them is climate change is already changing biodiversity, and you need to take that into account before you can make other management plans.”

The scientists used multiple long-term datasets and advanced machine-learning techniques to determine the relationship between biodiversity and environmental change, which enabled them to predict climate-driven biodiversity change until 2100.