Chiang Mai Urban Growth Model

This dataset includes the inputs and outputs of Chiang Mai Urban Growth Model. To predict the urbanization probabilities of Chiang Mai in 2050, a cellular automata-based urban growth model called SLEUTH was used. The details of SLEUTH model can be found in the Project Gigalopolis website (available at http://www.ncgia.ucsb.edu/projects/gig/About/about.html). Five growth parameters (i.e. Diffusion, Breed, Spread, Slope, and Road Gravity) are used by the SLEUTH model. As the name implies, the inputs are slope, land use, excluded area, urban area, transportation and hillshade. In this model the change in land use was not considered as the focus is only on the change of urban (i.e. built up) areas. The urban area inputs for four years (i.e. 1975, 1990, 2000, 2014) were obtained from the European Commission's GHS-BUILT datasets (available at https://ghsl.jrc.ec.europa.eu/ghs_bu2019.php). The slope and hillshade layers were obtained from the ASTER GDEM datasets (available at https://earthexplorer.usgs.gov/). Two inputs for road networks in 2010 and 2014 were obtained through digitization of Landsat images (available at https://earthexplorer.usgs.gov/). Excluded area includes the military zones, airport, religion institutions, golf courses and national parks, and it was extracted from Sangawongse and Kowsuvon (2011). Two different approaches, namely the brute force and genetic algorithm (GA), were employed for the calibration of growth parameters. The codes corresponding to these two models are also available in Project Gigalopolis website (available at http://www.ncgia.ucsb.edu/projects/gig/Dnload/download.htm). The Diffusion, Breed and Spread parameters are calculated as 1 in both calibration methods. Slope and Road Gravity parameters were calculated as 100 and 21 in brute force calibration, while they were calculated as 84 and 45 using GA calibration. This translates to a dominant road influenced growth in the city. During the calibrations, Lee-Sallee metric was used to measure the spatial urban fit. It was calculated as around 0.7 after the brute force calibration, and as around 0.75 after the GA calibration. Since a value between 0.3 and 0.6 is acceptable, both Lee-Sallee values indicate that the models can be considered successful. The outputs reveal that the highest probability of urbanization in brute force-based model is 39%, and it is 42% in GA-based model. [1] Sangawongse, S. and Kowsuvon, N. (2011) “Impact Assessment of Urbanization on Environmental Quality in Chiang Mai-Lamphun Valley, Northern Thailand” Japan Aerospace Exploration Agency (JAXA)-AIT Final Report.本数据集涵盖清迈城市增长模型的输入与输出数据。为预测清迈2050年的城市化概率，研究采用了基于元胞自动机（cellular automata）的城市增长模型SLEUTH。SLEUTH模型的详细信息可于Gigalopolis项目官网查询（访问链接：http://www.ncgia.ucsb.edu/projects/gig/About/about.html）。 该模型共设置5项增长参数，分别为扩散（Diffusion）、繁殖（Breed）、扩展（Spread）、坡度（Slope）与道路引力（Road Gravity）。顾名思义，模型输入数据包含坡度、土地利用、禁建区、建成区、交通设施与晕渲地形（hillshade）。本研究未考虑土地利用变化，仅聚焦于城市（即建成区）的扩张过程。 1975年、1990年、2000年及2014年共4期建成区输入数据，取自欧盟委员会GHS-BUILT数据集（访问链接：https://ghsl.jrc.ec.europa.eu/ghs_bu2019.php）。坡度与晕渲地形图层源自ASTER GDEM数据集（访问链接：https://earthexplorer.usgs.gov/）。2010年与2014年的两套道路网络输入数据，通过对Landsat影像进行数字化解译获取（访问链接：https://earthexplorer.usgs.gov/）。禁建区包含军事区域、机场、宗教机构、高尔夫球场与国家公园，其数据提取自Sangawongse与Kowsuvon（2011）的研究成果。 本次研究采用蛮力法（brute force）与遗传算法（GA, Genetic Algorithm）两种方法对增长参数进行校准。对应两种校准方案的代码同样可于Gigalopolis项目官网获取（访问链接：http://www.ncgia.ucsb.edu/projects/gig/Dnload/download.htm）。在两种校准方法下，扩散、繁殖与扩展参数的取值均为1。蛮力法校准得到的坡度参数与道路引力参数分别为100与21，而遗传算法校准得到的对应参数则为84与45，该结果表明研究区城市增长受道路因素主导。 在校准环节，采用Lee-Sallee指标（Lee-Sallee metric）衡量城市空间拟合程度。蛮力法校准后的Lee-Sallee指标值约为0.7，遗传算法校准后的该指标值约为0.75。由于0.3~0.6为可接受的取值区间，两项指标均证明本次校准模型具备有效性。 模型输出结果显示，基于蛮力法的城市化最高概率为39%，基于遗传算法的城市化最高概率为42%。 [1] Sangawongse, S. 与 Kowsuvon, N. (2011) 《泰国北部清迈-南奔谷地城市化对环境质量的影响评估》，日本宇宙航空研究开发机构（JAXA）-AIT最终报告。