Internal and Structural Variability in Global Urban Climate Projections in CMIP6

Abstract:

Climate change are projected to impose substantial impacts and climate-driven threats to urban areas. Effective development decisions and local actions to manage these risks rely on robust climate projections that are specific to built landscapes. A robust modeling framework to address uncertainty in local- or regional-scale climate change should include the roles of internal variability, model parameter and structural uncertainty, and scenario uncertainty. Quantitative attributions of the structure of these uncertainties have been typically done for non-urban surfaces at regional scales using multi-modeled (e.g. CMIP) grid cell means. Such analysis for local-scale urban areas, however, has never been achieved due to the near-universal lack of physical-based urban land parameterization in the state-of-the-art Earth system models (ESMs). This study assesses such variability structure through a novel multi-model urban emulator framework. We build a credible random forest-based urban climate emulator by utilizing the Community Earth System Model (CESM) simulation results. The emulator was then applied to 26 CMIP5 models to generate urban climatic responses. The structural variability was obtained by combining emulated results from different CMIP models, and the internal variability was assessed using the CESM-LE (large ensemble) simulations. Results show that there is an obvious structural variability of urban temperature change across CMIP models, ranging from 2-6 K, which aligns with that of background temperature change and increases with time. The structural variability of both urban and background temperature change also varies geographically and seasonally, and has a similar pattern with each other, with the Eurasia region demonstrating higher variability in both summer and winter. The internal variability of urban temperature change across CESM members, on the other hand, is consistent over the timescale of a century, ranging from 4-5 K, which also differs slightly in different regions and seasons, and accords with the background temperature variability.