Time Series Analysis And Simuation Of The Surface Urban Heat Island

The Urban Heat Island (UHI) effect has become one of important topics of the global and regional climatic change research. Studying timescale characteristics of the UHI makes it possible to graspe the intrinsic variation of UHI and figure out the contribution of each factor. At present, many studies analyzing the driving mechanism of UHI are based on remote sensing data and take limited number of factors in consideration. Meanwhile, qualitative analysis of UHI timescale characteristics is still a preference. The research in this paper lies on the urban land surface heat island which is observed by satellite remote sensing in Beijing region. The UHI timescale model is built aims at finding the evolution characteristics of monthly and daily UHI. The inherent mechanism of UHI is quantitatively analyzed through driving and simulating the mechanism of daily UHI. Summary of the content in the dissection are:(1) A disaggregation method is generated to upscale the spatial resolution of land surface temperature (LST) to lkm spatial resolution which keeps four monthly LST datasets being consistent from 2001 to 2012. Verification result shows that the upscaled datasets have high accuracy and meets the needs of the research of UHI’s temporal scale well. Furthermore, the monthly UHI intensity time series are constructed based on these datasets.(2) To have a quantitatively analysis of the time series characteristics of UHI intensity in different frequencies, and quantize the influencing mechanism the weather factors effect on UHI, Classification Seasonal Decomposition model and X-11-ARIMA model are applied to separate each component of time series. High frequency components of daily UHI intensity time series are also acquired.(3) The factors for UHI are divided into four categories:land surface features, climate background, weather phenomena, and human activities. These factors are parameterized by datasets from remote sensing and monitoring stations. Support Vector Regression (SVR) method is used to simulate the correlation between daily UHI intensity high-frequency components and correlated factors. The results show that SVR obtained best performance when Linear kernel function are chosen and partial high-frequency factors are averaged to five-day values in this study. The RMSEs of four groups of daytime and nighttime samples in SVR simulation both have modest value less than 1.5K.(4) Finally, the relationship between each factor and daily high-frequency UHI is obtained and analyzed by taking scenario analysis based on SVR. Then the contribution of every factor is quantified, which helps to reduce the UHI and minimize the extent of UHI.