作　　者： ; ; ; ;

机构地区： 中山大学地理科学与规划学院

出　　处： 《地理与地理信息科学》 2016年第5期29-36,128,共9页

摘　　要： 城市土地利用及其变化对城市环境有着重要影响。很多学者已经结合元胞自动机和机器学习算法对城市扩张进行了相关的模拟研究,但针对复杂的多类土地利用相互变化过程的研究仍然较少。该文提出了一种基于随机森林算法的多类元胞自动机(RFA-CA)模型,并将其用于模拟和预测复杂的多类土地利用变化。该模型使用随机森林算法提取元胞自动机的转换规则,并计算了各空间变量的重要性,在东莞市2000-2014年土地利用动态模拟结果中,Kappa系数和整体精度分别为0.73和84.7%。针对每一种土地利用类型,计算了影响东莞市土地利用变化的各空间变量的重要性,结果显示,交通、区位因素对东莞市土地利用变化格局的形成有重要影响。文中引入的POIs邻近因素反映了城市空间开发程度的高低,同样对多类土地利用格局的形成具有重要作用。 Urban land use information plays an important role in urban environment. China＇s land resources are under increasing pressure due to the rapid development of ec：onomic growth and urbanization process. Many previous studies have focused on urban area expansion by integrating Cellular Automata （CA） and M achine Learning （ML） algorithms. However, simulation of multiple land use changes using CA model is difficult because there are numerous spatial variables and the change types have to be settled.This paper proposes a new method to simulate spatiotemporal Complex multiple land uses by using Random Forest Algorithm （ RFA） based on CA model. One significant advantage is that this algorithim can reduce error upper limit of the gen-eralized model. RFA- CA model can extract the complex land use conversion rules and measure the im prtance of the spat ial variables, wThich could explain the influence of variables in different land use change. We apply RFA- CA model on simulating and predicting the dynamics of multiple land uses in Dongguan during 2000- 2014（ Dongguan is a modern city in South China with rapid ecmomic development in recent 20 years） .The result sho＇ws that RFA- CA model has high accuracy ＇with Kappa 0 72 and overall accuracy 84 7%. Additionally, when compared to ANN- CA model, the results are also improved to varying degree. Through measuring the importance of some spatial variables including POIs variables which are introduced in our study, we obtain different regular for each land use type. Taking all land uses as a whole, we find that the traffic, position and a part of POIs factors play an important role in forming Dongguan s land use pattern.

关 键 词： 多类元胞自动机 随机森林算法 土地利用变化 变量重要性

领　　域： [经济管理]