Climate Change - Evidence, Attribution, Projections and Remedies

Narrative

The surface temperature record from five independent datasets, HadCRUT5, NASA GISTEMP, NOAA GlobalTemp, Berkeley Earth, and ECMWF ERA5, converges on 2024 as the warmest year in the 175-year instrumental record, at approximately 1.55 to 1.62 degrees C above the 1850 to 1900 pre-industrial baseline depending on dataset. Berkeley Earth's analysis, combining 23 million monthly station measurements with roughly 500 million ocean observations, estimated 2025 as the third warmest year at 1.44 plus or minus 0.09 degrees C, still well above previous decades. The 11 warmest years on record have all occurred in the last 11 years, and the DCENT dataset released in 2024 now offers a fully dynamically consistent uninterpolated product alongside the established records. A key contested boundary involves SST homogenisation: Menemenlis et al. (2025) show that satellite-era SST trends differ substantially across products, with MERRA2 recording 0.176 degrees C per decade and ERA5 0.142 degrees C per decade, meaning global temperature trend estimates carry structural uncertainty from SST dataset choice.

Physical proxies reinforce the surface record. Arctic sea ice maximum extent in March 2025 reached 14.33 million square kilometres, the lowest in the 47-year satellite record, 1.31 million square kilometres below the 1981 to 2010 average; the oldest ice category, ice more than four years old, has declined by more than 95 per cent since the 1980s. Ocean heat content in the upper 2000 metres broke records again in both 2024 and 2025: from 2023 to 2024 upper-ocean heat content increased by 16 zettajoules, roughly 40 times world annual electricity generation. Svalbard glacier mass loss in the record-warm summer of 2024 exceeded all previous estimates by a wide margin, with sea-level rise contribution from Svalbard alone comparable to that of the entire Greenland ice sheet in that single year.

Attribution of observed warming to fossil-fuel combustion is anchored by the Global Carbon Project's annual carbon budget series. The 2024 budget (Friedlingstein et al., Earth System Science Data, 2025) documented atmospheric CO2 reaching 423 ppm in 2024, with fossil fuel emissions continuing to rise and human-induced warming at 1.36 degrees C. A companion Nature paper revised the natural land sink substantially downwards and increased net land-use change emissions, reducing the apparent buffer on remaining carbon budgets. The IPCC AR6 Synthesis Report (2023) attributed 0.8 to 1.3 degrees C of the observed 0.95 to 1.20 degrees C warming to anthropogenic emissions. The principal scientific debate about ECS remains active: CMIP6 models produce ECS values ranging from 1.8 to 5.6 degrees C, while new observational cloud-feedback analyses published in 2025 to 2026 point towards the higher end. Scafetta (2025, Gondwana Research) argues via arXiv preprint that models overstate ECS and solar influences, representing a minority empirical position contested by mainstream attribution literature.

Tipping-point research has hardened since Armstrong McKay et al. (Science, 2022) identified 16 tipping elements, with five already possible at current warming and four, including Greenland and West Antarctic ice sheets, widespread permafrost thaw, and coral reef die-off, becoming likely above 1.5 degrees C. AMOC research remains contested: CMIP6 models suggest no 21st-century collapse under medium scenarios but Drijfhout et al. (2025) and van Westen et al. (Science Advances, 2024) provide physics-based early warning signals indicating the AMOC is on a tipping trajectory. On remedies, a 2026 Earth System Dynamics paper finds stratospheric aerosol injection (SAI) could increase global land carbon storage by 5.9 per cent and protect the Amazon under SSP585, while a 2025 npj Climate Action economic analysis estimates SAI side-effect harms at zero to 809 billion USD annually. The Oxford Open Climate Change systematic review (2024) identifies three classes of SAI uncertainty, scenario, model, and natural variability, which remain poorly quantified for policy purposes.