Application of time-series analysis to urban climate change assessment

Emerging process-based urban management creates an urgent necessity to track the evolution of urban climate and assess the efficacy of existing plans and practices in building climate resilient cities. However, the complex interactions between urban surface and the upper atmosphere bring about non-linear, non-stationary, and multi-scale urban climate variations, making it challenging to isolate urban effects from other nature-induced variations. This study introduces two types of data-driven techniques, time series clustering and digital signal decomposition, to tackle the above challenge. Particularly, time series clustering is able to zone any study area into multiple geographical clusters according to the heterogeneity among temporal evolution trajectories of a given urban climate parameter, accounting for its non-linear and non-stationary variations. The clusters present diverse patterns of ascending, descending, or other complex temporal variations that can be used to identify both successes and failures of existing plans in combating adverse urban climate change. Digital signal decomposition can decompose the non-linear and non-stationary time series into multiple components with different time-scales, thereby offering the possibility to identify the part of long-term variation induced by the concerned urban effects. It can be used alone to identify the urban-induced climate change in a specific city, or in combination with time series clustering to recognize the similarities and disparities of urban-induced long-term trends across multiple clusters within a city. This study provides scientific basis and technical support for more robust and precise assessment of urban-induced local climate change, thereby empowers policymakers to develop targeted strategies to enhance climate-resilience.