Destination Earth Climate Change Adaptation Digital Twin (Climate DT) is setting up an operational simulation framework that enables to produce bespoke climate and impact-sector information, supporting evidence-based decision-making and fortified societal resilience. Climate DT combines global kilometer-scale climate models and a range on impact-sector applications in a unified operational workflow, producing both multi-decadal climate projections and storyline simulations at the scales of the impacts of climate change and extreme events. This imposes new demands for the evaluation of simulation quality. The question is, which reference data are adequate to evaluate the rich process-level variability present in these new-generation models. Here, we advocate the use of raw Earth observations for assessing physical fidelity of the modelled fine-scale variability, operating exclusively in the observation-space.</p> <p>The Climate DT framework (<a href="https://platform.destine.eu/" target="_blank" rel="noopener">https://platform.destine.eu/</a>) enables unique evaluation of the simulation quality using Earth observations directly, as the model output data is available run-time in a near-native grid. This is in stark contrast to traditional model-space evaluation using archived simulations and gridded reference data. Climate DT simulations are evaluated in observation-space prior to any spatio-temporal truncation, exposing their process-level variability for inter-comparison with raw Earth observations and bridging the gap between modelling and observing the Earth system. The synergy aspect here is the extensive sharing of observation modelling infrastructure with data assimilation in numerical weather prediction.</p> <p>This article presents the Climate DT concept to assess simulation quality with Earth observations and showcase it through the lens of synoptic surface observations. The evaluation covers the simulated mean, trend, variability, and extremes of historic simulations from 1991 to 2014 of the IFS-NEMO, IFS-FESOM, and ICON models. Annual cycles of 2‑metre temperature, and (to a somewhat lesser extent) humidity and 10‑metre wind speed are generally well-simulated. However, these quantities contain also significant process-level weaknesses, such as too weak process level variability at diurnal, intramonth, and interannual time-scales. The evaluation results indicate that while there is still work to be done to improve the Climate DT simulation models, it is a novelty that they can now be examined at such level of detail and objectively interfaced with raw Earth observations containing the corresponding process-level imprints. Furthermore, since climate change adaptation mostly occurs at local level, the raw observation-based evaluation informs directly the scales of interest.