建構都市垂直發展預測模式-以臺北市內湖區與南港區為例

NCUIR > Engineering College > Graduate Institute of Civil Engineering > Electronic Thesis & Dissertation > Item 987654321/87174

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/87174

Title:	建構都市垂直發展預測模式-以臺北市內湖區與南港區為例
Authors:	楊鈞棠;Wang, Chun-Tang
Contributors:	土木工程學系
Keywords:	垂直都市成長;土地利用與地表覆蓋變遷;隨機森林模式;Urban development;Land use and land cover change (LULC);Vertical urban growth;Random forest algorithm;Taipei City
Date:	2021-10-28
Issue Date:	2021-12-07 15:17:28 (UTC+8)
Publisher:	國立中央大學
Abstract:	都市土地覆蓋受到人類活動影響而產生快速的變動，因此過去研究常使用不同的都市發展驅動因子和模型來模擬和預測都市成長趨勢。隨著都市地區的人口增加，為了在稀少的都市土地增加可以使用的空間，近年來在亞洲地區，許多都市開始往垂直方向發展。臺北市亦出現垂直成長現象，位在臺北市東側的兩個行政區－內湖區與南港區，由於基隆河截彎取直工程和老舊產業遷移，出現許多閒置的土地，隨著政府開發政策和民間廠商的進駐，成為臺北市快速成長的區域。若僅監測與模擬單一水平方向的都市成長，容易低估都市的成長速度及忽略都市的結構變化。　　本研究希望分析與模擬內湖區與南港區2002至2017年間的都市垂直成長現象，透過多個年期臺北市一千分之一比例尺地形圖，取得建物空間分布和建物樓層數，將建物依樓層數分為三個高度等級進行統計，探討這兩個行政區是否出現垂直成長和空間結構改變的現象。接著使用隨機森林模式(Random Forest, RF)與多個都市發展驅動因子建立都市的建物成長分布和變遷模型，驗證模式的正確性與時效性和了解影響都市建物分布的因子。結果顯示，內湖區在2002-2017年有明顯的垂直方向都市成長，區內的容積量與中高樓層建物所占比例皆有明顯成長；南港區則是在2007年以後容積量與中高樓層建物面積出現明顯增加。隨機森林模式模擬各年期都市整體建物分布的AUC值皆0.9以上，其中對都市建物分布較重要的因子主要為自然環境和交通類別。利用模型建立的各年期三個等級的建物分布機率地圖，經過不同年期地真資料的驗證，發現AUC值從原本的0.9以上，隨著預測時間的拉長逐漸下降，且中高樓層建物預測效力下降的幅度比較大。2012-2017年建物面積變遷幅度較大，利用隨機森林模式模擬，其預測正確性約為70%。整體而言，各等級分布機率地圖在預測十年後的建地分布時，AUC值仍保持在0.8以上，代表本研究所建立的模式對於十年內的建物分布預測有一定的效力，所建立的建物分布潛勢地圖可做為未來都市發展趨勢的參考。;The land cover of the urban area has changed rapidly due to human activities. As the urban population keeps increasing, some cities, especially cities in East Asia, start to grow vertically to create more available space in the limited land. Taipei city, the capital city in Taiwan, can also find this phenomenon. In recent years, many construction companies and high-tech companies have entered Neihu and Nangang, two districts in the east of Taipei, and developed tall buildings that reshape the urban structure significantly since 2000. In this research, to analyze and model the vertical urban growth in Neihu and Nangang from 2002 to 2017, this study produced the building data using 1/1000 scale topographical maps to calculate the building area and total floor area and class the building height into three categories to find out the growth trend in the two districts. Then, the random forest (RF) algorithm is selected to model the building growth based on the three-building height classes, and the outcomes were verified by topographical maps. The results show that the total floor area and the tall buildings had increased both in Neihu and Nangang since 2002, showing the urban growth in the vertical direction is significant. The RF generates the probability maps of the potential distribution and change of tall buildings in the study area, and the modeling outcomes were assessed by AUC values in the tested years which are all higher than 0.9. About 70% of changed building height in 2007-2012 are correctly predicted by the RF model. The model reveals that the natural environmental factors and traffic factors are more important to affect the distribution of buildings with different heights. However, when predicting future building height, the AUC decreases to 0.8 when the predicting year is more than 10 years. This study gives an example indicating the applicability of RF to predict future urban structure in the vertical dimension which can greatly help the decision-making of land development for government and urban planners.
Appears in Collections:	[Graduate Institute of Civil Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	133	View/Open

社群 sharing

Loading...