顾及地类转换差异的城市空间扩展元胞自动机模型及应用研究
作者简介:于明明(1990-),男,河北石家庄人,硕士生,研究方向为城市扩张模拟。E-mail: mmyunited@mail.csu.edu.cn
收稿日期: 2017-04-30
要求修回日期: 2017-08-25
网络出版日期: 2018-01-20
基金资助
国家自然科学基金项目(41171326、40771198)
Cellular Automata Model of Urban Spatial Expansion Considering the Differences of Land Types Transition and its Application Research
Received date: 2017-04-30
Request revised date: 2017-08-25
Online published: 2018-01-20
Supported by
National Natural Science Foundation of China, No.41171326, 40771198.
Copyright
元胞自动机模型已经成为城市空间扩展模拟研究的重要方法之一,并得到广泛应用。然而,现有的城市扩展元胞自动机模型仍存在不足。由于元胞状态设置较为简单,从而使模型转换规则中对不同用地类型向城市用地转换的差异与强度考虑不够。基于此本文在元胞自动机模型的框架下,设计了多元结构的元胞状态及转换规则,提出了顾及地类转换差异与强度的城市扩展元胞自动机模型。在计算非城市用地向城市用地转换的转换概率时,该模型考虑了3个方面的概率:① 地形地貌、经济发展等城市发展的驱动因素对城市用地扩展的影响概率,该概率采用logistics方法进行计算;② 邻域元胞的用地类型对中心元胞转换概率的影响,该概率采用扩展摩尔型方法进行计算;③ 不同类型的非城市用地(本研究中包括耕地、林地和裸地3种类型)向城市用地转换的强度,该概率由模拟基期土地利用数据与目标年份土地利用数据的叠加,得出不同类型的非城市用地在此时间段内向城市用地转换的规模,进而确定不同类型的非城市用地向城市用地转换的强度。最后,将以上3种概率的乘积作为元胞转换的概率。通过转换概率与转换阈值的对比判断中心元胞是否在下一个阶段转换为城市用地。经过迭代计算,不断增加城市用地元胞的数量。当模拟城市用地的结果与目标年份的城市用地规模差值在一定的范围内时停止模拟,得出最终结果。模型构建完成后,本文以长株潭城市群核心区为例进行了模拟实验。以2001年该地区的土地利用数据为基期数据,模拟2010年该地区的城市用地规模和空间分布。研究结果表明,根据本文提出的模型模拟的城市扩展结果与真实数据相比具有较高的一致性。模拟结果正确率达到68.66%,比基于传统logistics回归的元胞自动机模型的模拟精度提高了4.25%,Kappa系数为0.675。该模型较好地模拟了长株潭城市群核心区城市扩展,在城市空间扩展模拟中具有较好的适应性与有效性。
于明明 , 曾永年 . 顾及地类转换差异的城市空间扩展元胞自动机模型及应用研究[J]. 地球信息科学学报, 2018 , 20(1) : 48 -56 . DOI: 10.12082/dqxxkx.2018.170132
The cellular automata model has become one of the important methods of urban spatial expansion simulation. However, the existing cellular automata model of urban expansion still has some shortcomings. The cell state setting is relatively simple. Differences and strength of land types conversion are not enough in the conversion rules. In this paper, the cellular state and conversion rules of multivariate structures are designed under the framework of cellular automata model, and the urban extended cellular automata model considering the difference and intensity of land conversion is proposed. In the calculation of the conversion probability of non-urban land to urban land, this model takes into account the probability of three aspects: (1) For the impact probability of topography, economic development and other factors of urban development on urban land expansion, we used the logistics approach to calculate this probability. (2) The impacts of land types of neighborhood cells on the convergence rates of central cells. We use the extended molar method to calculate this part of probability. (3) The conversion intensity of different types of non-urban land (i.e. cultivated land, woodland and bare land) into urban land. The calculation of this part is to get the conversion scale of different types of non-urban land into urban land during the period of the base year and the target year by simulation of the superposition of the land use data in this period. Then, we further determine the conversion intensity of the different types of non-urban land into urban land. Finally, the product of the above three probabilities is used as the probability of cell transformation. We used the conversion probability and the conversion threshold to determine whether the central unit would be converted into urban land in the next stage. The number of urban land cells would be increased after the iteration calculation. When the difference between the results of simulated urban land and the size of urban land of the target year was in a certain range, we stopped the simulation and get the final results. The results show that the proposed model can capture urban expansion in the study area with sgood adaptability. The accuracy of the simulation results is 68.66%, which is 4.25% higher than that of the cellular automata model based on the traditional logistics regression. The Kappa coefficient is 0.675.
Fig. 1 Model of conversion intensity from different land-use types to urban land图1 不同土地利用类型向城市用地转换强度模型 |
Fig. 2 Location of the study area图2 实验区位置图 |
Tab. 1 Experimental data and its sources表1 实验数据及其来源 |
数据类型 | 数据名称 | 数据来源 |
---|---|---|
栅格数据 | 土地利用数据(2001、2010年),坡度、海拔数据 | 土地利用数据由相应时间段的Landsat TM数据解译获得,遥感影像来源于中国地理空间数据云,中国科学院计算机网络信息中心国际科学数据镜像网站 |
矢量数据 | 交通道路信息(高速、铁路、国道、省道、城市主干道、城市次干道) | 国家基础地理信息中心 |
生活服务信息(宾馆酒店、车站码头、各级政府、汽车服务、商业网点、学校、医院、银行、邮政、电信)、工作便利信息(企事业单位、政府机关)、休闲娱乐信息(水上洲、水域、休闲娱乐场所、旅游景点、运动场馆、植被层) | 湖南省城市电子地图 | |
社会经济数据 | 人口密度、城市化率、人均 GDP、全社会固定资产 | 湖南省统计年鉴、长沙市统计年鉴、湘潭市统计年鉴和株洲市统计年鉴以及《长株潭城市群生态绿心地区总体规划(2010-2030)》、《长株潭城市群区域规划(2008-2020 年)》 |
注:本文的栅格数据分辨率均为30 m×30 m |
Fig. 3 Land use data图3 土地利用数据 |
Fig. 4 Conversion intensity from different land-use types to urban land图4 不同用地类型向城市用地转换强度 |
Tab. 2 The driving factors and weights of urban land表2 建设用地驱动因子及权重 |
驱动因子 | 权重 | 驱动因子 | 权重 |
---|---|---|---|
生活服务(β1) | -2.641 | 房价(β9) | 4.627 |
坡度(β2) | 1.734 | 城市化率(β10) | 0.527 |
海拔(β3) | 5.434 | 到国道距离(β11) | 1.674 |
工作便利(β4) | 5.442 | 到高速距离(β12) | 0.842 |
到桥梁距离(β5) | 1.243 | 到城市干道距离(β13) | -2.517 |
人口密度(β6) | 5.205 | 到城市环线距离(β5) | 0.654 |
人均GDP(β7) | 0.654 | 到省道距离(β6) | 1.981 |
固定资产(β8) | 0.340 | 常数(β0) | -7.172 |
Fig. 5 Spatial distribution of actual urban land and modeling urban use in 2010图5 2010年实际城市用地与模拟城市用地空间分布 |
Fig. 6 The spatial distribution of actual new urban land use and modeling urban land use from 2001 to 2010图6 2001-2010年城市用地实际空间扩展及模拟结果 |
Fig. 7 Overlay results of simulation and status Quo of urban land in 2010图7 2010年城市用地扩展模拟结果与实际分布的差异 |
Tab. 3 Confusing matrix of the simulation results of 2010表3 2010年模拟结果混淆矩阵 |
模拟 | ||||
---|---|---|---|---|
不转变 | 转变 | 正确比/% | ||
实际 | 不转变 | 5821601 | 123389 | 97.92 |
转变 | 125267 | 274435 | 68.66 | |
总精度/% | 96.08 | |||
Kappa系数 | 0.675 |
Fig. 8 The traditional simulation results of logistic-based cellular automata model图8 传统基于logistic的元胞自动机模型模拟结果 |
The authors have declared that no competing interests exist.
[1] |
[
|
[2] |
[
|
[3] |
[
|
[4] |
[
|
[5] |
[
|
[6] |
[
|
[7] |
[
|
[8] |
[
|
[9] |
|
[10] |
[
|
[11] |
[
|
[12] |
|
[13] |
[
|
[14] |
|
[15] |
|
[16] |
|
[17] |
|
[18] |
[
|
[19] |
[
|
[20] |
[
|
[21] |
[
|
[22] |
[
|
[23] |
[
|
[24] |
[
|
[25] |
[
|
[26] |
[
|
/
〈 | 〉 |