Over the course of the IPCC's history — from the First Assessment Report (FAR) in 1990 to the Fifth Assessment Report (AR5) in 2014 — considerable progress has been made in our understanding of anthropogenic climate change. Key advances can partially be attributed to improvements in modelling capabilities, including resolution, addition of new or updated processes, and ultimately, computing power. However, model developments are inextricably linked to a simultaneous increase in observational analyses and a corresponding expansion of the observational network. Records are now arguably of sufficient length to assess the skill of early projections, and further track advances in understanding over the past 27 years of the IPCC.

Several studies have illustrated that, despite the relative simplicity of early climate models, large-scale temperature predictions from the FAR have been surprisingly consistent with observed changes, both in terms of the magnitude and spatial pattern of anthropogenic warming1,2. While an impressive feat, our understanding of other variables has seen a much greater evolution from the first to the fifth AR, some examples of which are illustrated in this issue. Ice-sheet models, for instance, now have the capacity to explicitly simulate dynamical imbalances, and as a result, projections of sea-level rise contributions are now in close agreement with observations (page 672). Similarly, it was not until AR5 that models had improved to the extent that overall cloud feedbacks were first quantified as being positive within the IPCC's uncertainty framework (page 674). Developments in methane sources and sinks (page 678) and quantification of agricultural yield impacts (page 680) offer further evidence of advanced understanding.

While highlighting a few instances of early successes and continued developments, the wealth of climate change literature clearly emphasises the progress made since the FAR. However, we cannot be complacent. Building on the examples above, key glaciological processes remain unquantified, individual cloud feedback terms continue to be highly uncertain, apportionment of methane sources and sinks challenging, and agricultural assessments focussed on crops, omitting impacts on other produce relevant to food security. As acknowledged by all these authors, more and continued observations are required to drive progress, both in terms of improved model validation and process understanding. In an Interview, Veronika Eyring, chair of the Coupled Model Intercomparison Project Phase 6 (CMIP6), further stresses the importance of increased collaboration between observationalists and modellers (page 684).

These are exciting times for climate science. As CMIP6 gets underway (page 684) in preparation for the IPCC's Sixth Assessment Report, a wealth of new model data will soon be available, further enhancing our understanding of the physics, impacts, and mitigation potential of anthropogenic climate change. However, continued progress relies on enhanced collaboration between various user groups and communities.