Estimation of time-series heat fluxes at city scale, based on integration of geostationary and sun-synchronized satellite observations

Abstract

An Urban Heat Island (UHI) is a positive temperature anomaly over an urban area, where the temperature is significantly higher than the rural surroundings. In some cities, urban heat islands have negative effects on public health. This thesis investigates the UHI effect in terms of different energy and heat flux parameters including sensible heat and anthropogenic heat in Hong Kong, as a case study for other highly compact and urbanized world cities. Previous research applying remote sensing technology to the energy balance model can only derive heat flux images at a single time instance, due to the limitation of data provided by high spatial resolution satellites. Thus, this prohibits further research in the studying of UHI behavior for a whole day at city scale. A novel spatiotemporal data fusion algorithm, Hopfield Neural Network SPatio-tempOral data fusion model (HNN-SPOT), was developed in this research and was applied to enhance the spatial resolution of the time-series Himawari 8 satellite images from 2 km to 100 m spatial resolution. Therefore, it enables the study of diurnal pattern of different heat flux data during a day at city scale. Results showed that the HNN-SPOT algorithm was comparable with the Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM), the benchmark algorithm in spatiotemporal data fusion. HNN-SPOT can produce reliable fusion results (with correlation coefficient > 0.90) over consistent land cover. Lower accuracies with correlation coefficient around 0.40 were achieved when the land cover undergoes extensive spatial change and around 0.70 with extensive temporal change. Hourly energy balance parameters over Hong Kong at 100 m spatial resolution were generated and analyzed over the course of three days. Referring to the diurnal pattern of sensible heat flux, which is a suitable indicator of UHI effect, it was found that the difference of sensible heat flux between urban area and rural area in Hong Kong were higher from morning to afternoon and during nighttime, and became smaller from afternoon to evening. Over the urban areas in Hong Kong, higher nighttime sensible heat flux in warmer season and higher sensible heat flux in denser urban areas were discovered. These findings have a direct implication of the UHI effect in Hong Kong. This thesis is the first attempt to generate time-series heat flux data at city scale in a day using the energy balance model with remote sensing technology. It is hoped that the findings and results arising from this thesis can contribute to the new knowledge of the diurnal spatiotemporal behavior of UHI in a densely urbanized city under remotely sensed observations, and will be useful in decision making for urban planning and UHI mitigation.