Climate change can cause property damage and deaths in cities. City-scale climate projections are essential for making informed decisions towards climate change mitigation and adaptation at city levels. This study aims at developing ensemble projections of temperature extremes at the city-level and quantifying the contributions of various factors to the resulting uncertainty of the ensemble projections. The city of Toronto will be used here as an example to demonstrate the effectiveness of the proposed research framework. In particular, the stepwise cluster analysis (SCA) model will be used to perform climate downscaling to three GCM datasets (GFDL, IPSL, and MPI) under three emission scenarios (RCP2.6, RCP4.5, and RCP8.5) in order to generate city-level climate projections for the city of Toronto. The SCA model is demonstrated to be capable of capturing the inter- and intra-annual variations of the daily maximum, mean, and minimum temperatures in the studied city. The results suggest that mean temperatures in Toronto are projected to increase at the rate of 0.15 and 0.5 °C/decade under RCP4.5 and RCP8.5, respectively, while no significant warming trend is detected for RCP2.6. In terms of temperature extremes, extreme warm events are projected to increase while extreme cold events decrease under all emission scenarios. The decrease in the heating demand is two to four times larger than the increase in the cooling demand, indicating a decrease in the city’s total energy use. The projected warming might be beneficial for the urban growers because of the significant increases in the growing season length and growing degree days; however, the residents of the city of Toronto are likely to experience simultaneous increases in the intensity, duration, and frequency of heatwave events in future summers. Because of the warming, coldwave events in winters are likely to become less frequent and be shorter in duration, but their intensity is expected to increase significantly. Through decomposition of the resulting uncertainty of the ensemble projections, emission scenario is found to be the dominant factor for the uncertainty associated with urban climate projection.

气候变化可能导致城市财产损失和死亡。城市规模的气候预测对于做出明智的决策以在城市一级缓解和适应气候变化至关重要。这项研究的目的是开发城市一级极端温度的整体预报，并量化各种因素对整体预报的不确定性的贡献。此处以多伦多市为例，说明所提出的研究框架的有效性。特别是，逐步聚类分析（SCA）模型将用于在三种排放情景（RCP2.6，RCP4.5和RCP8.5）下按顺序将气候缩减到三个GCM数据集（GFDL，IPSL和MPI）生成多伦多市的城市级气候预测。SCA模型被证明能够捕获所研究城市的每日最高，平均和最低温度的年际和年内变化。结果表明，在RCP4.5和RCP8.5下，多伦多的平均温度预计将分别以0.15和0.5°C / decade的速度增加，而RCP2.6没有发现明显的变暖趋势。就极端温度而言，在所有排放情景下，预计极端温暖事件将增加，而极端寒冷事件将减少。供热需求的减少量是供冷需求量增加量的二到四倍，表明城市的总能源使用量有所减少。由于生长季节长度和生长天数的显着增加，预计的变暖可能对城市种植者有利。然而，在未来的夏天，多伦多市的居民可能会同时增加热浪事件的强度，持续时间和频率。由于变暖，冬季的冷浪事件可能会变得不那么频繁且持续时间较短，但预计其强度会大大增加。通过分解整体预测的不确定性，发现排放情景是与城市气候预测相关的不确定性的主导因素。但其强度预计会大大增加。通过分解整体预测的不确定性，发现排放情景是与城市气候预测相关的不确定性的主导因素。但其强度预计会大大增加。通过分解整体预测的不确定性，发现排放情景是与城市气候预测相关的不确定性的主导因素。