以深度學習模型探討亞熱帶不同都市景觀尺度之熱島特性比較

The Comparison of Thermal Properties of Subtropical Heat Islands at Different Urban landscape Scales by Deep Learning

不同都市造就不同都市景觀，都市景觀又進一步影響到不同都市景觀尺度上空的熱環境的分佈。因此本研究透過不同土地使用之測站與不同氣象因子下的大數據，比較不同都市景觀尺度的熱島特性。
研究整合氣象水文研究資料庫地面與垂直探空的數據，以深度學習演算來歸納不同尺度的都市景觀型態的結構樣態，透過瞭解三維結構的熱分佈樣態，比較大都會與小型城鎮的都市熱島結構與舒適度的差異。
深度學習演算法證實都市景觀尺度與型態對熱環境的差異影響至垂直結構，而雲量與風速在不同都市型態與季節皆會有不同緩解舒適度的作用。研究修正氣溫垂直遞減率每上升100m下降0.6℃，夏季板橋下降0.54℃與花蓮下降0.53℃，冬季修正為板橋下降0.43℃與花蓮下降0.42℃，結果呈現冬季溫度偏離理論值斜率的現象比夏季更大，顯示冬季熱島現象更為顯著；花蓮的風速較板橋來得高，因此在疏導都市熱，都市化程度較低的花蓮較有優勢；高雲量的陰天在夏季能阻擋太陽輻射，尤其在小型城鎮的花蓮又更加明顯；而小型城鎮在冬天高雲量時，輻射熱被雲吸收，反射地面的輻射熱，熱量蓄積在大氣雲下的溫室中，更容易加熱地面，因此雲量高反而會緩解冷不舒適，顯示小型城鎮花蓮比板橋對雲量因素更為顯著敏感。

Different urban patterns create different urban landscapes, and urban landscapes would further affect the distribution of thermal environments above different urban landscape scales. This study compares the characteristics of heat islands at different urban scales with big data under different meteorological factors. The study integrates ground and vertical-sounding data from the data band for atmospheric & hydrologic research. It uses deep learning algorithms to summarize the urban patterns at different scales. By understanding the heat distribution patterns of three-dimensional structures, to compare metropolitan and small towns in urban heat island structures and comfort.
The deep learning algorithm proves that the difference between the scale and type of urban landscape affects the thermal environment to the vertical structure, and the cloudiness and wind speed will have different comfort relief effects in different urban types and seasons. The research corrected the temperature vertical lapse rate to decrease by 0.6°C for every 100m increase. Banqiao decreased by 0.54°C in summer, and Hualien decreased by 0.53°C. In winter, it was corrected to reduce by 0.43°C in Banqiao and Hualien by 0.42°C.
The results show that the slope of the temperature deviation from the theoretical value is greater in winter than in summer, indicating that the heat island phenomenon in winter is more significant; The wind speed in Hualien is higher than that in Banqiao, so Hualien with a lower degree of urbanization has an advantage in relieving urban heat; cloudy days with high cloudiness can block solar radiation in summer, especially in Hualien, a small town. When the cloudiness is high in winter, the radiant heat is absorbed by the clouds and reflects the radiant heat from the ground, and the heat accumulates under the atmospheric clouds, which makes it easier to heat the ground. Therefore, the high cloud cover will alleviate the cold discomfort, which shows that the small town of Hualien is more significantly sensitive to cloudiness factors than Banqiao.