基于空间计量模型的跨区公交通勤需求机理
Mechanism of cross？？line commuter demand based on spatial econometric model

为了揭示土地利用因素对跨区公共交通需求的影响，从空间角度出发，利用空间计量模型对北京通州与城区之间的跨区公交通勤需求开展研究。首先，利用公共交通IC卡数据获取交通小区出行通勤需求；其次，基于北京市兴趣点（point of interesting， POI）数据量化土地利用类型、交通设施、区位因素三类土地利用指标，其中为降低土地利用类型变量的冗余，利用因子分析方法进行变量降维，生成8个土地利用类型的综合变量；最后，在空间角度验证出行需求具有空间相关性的基础上，分别建立跨区通勤起点O及终点D出行需求与土地利用的空间自回归模型、空间误差模型及空间杜宾模型。通过模型比选，最终选取拟合最优的空间杜宾模型作为土地利用因素对公共交通需求关系模型。研究表明：跨区公共通勤出行O、D点全局Moran？？s ？？I？？指数分别为？？0.385？？、？？0.503？？，具有较强的空间相关性，有必要引用空间计量模型；轨道交通站点密度对于O、D点通勤跨区公共交通出行需求具有积极的引导作用，影响系数最大，分别为11.56、？？9.82？？；土地利用混合度的空间滞后变量对O、D 点通勤需求影响最大且为正相关，影响系数分别为0.51、0.68，即周边小区土地利用混合度较高时，将提升该小区的通勤需求。该研究验证了交通小区高强度的土地利用开发和丰富的公共交通资源对公共交通出行的正向引导作用，为土地利用与公共交通出行需求之间的互动关系提供了定量分析依据。
To study the impact of land？？use factors on cross？？line public transport commuter demand between Tongzhou and urban areas of Beijing，a spatial econometric model was used . First, public transport IC card data was used to obtain transport commuter demand. Second, the land？？use types, transportation facilities and location factors were quantified based on the POI (point of interest) data of Beijing. Variable dimensions were reduced in the factor analysis method to eliminate redundant variables regarding land？？use type, and eight comprehensive variables of land？？use types are generated. Finally, The spatial auto regressive model, spatial error model and spatial Durbin model of travel demand and land？？use were established for the trans？？regional commuting origin O and destination D. Based on veritfing that travel demands was spatially relevant. Through comparison, the spatial Durbin model was selected for modeling the relationship between land？？use and public transport demand. The results show that the global Moran？？s index of public cross？？line commuting for O and D point is 0.385 and 0.503. Spatial econometric models are necessary for strong spatial correlations. The density of rail transit stations plays an active role in guiding travel demand for public cross？？line commuting, yielding the largest influence coefficients of 11.56 and 9.82 for O and D point, respectively. The spatial lag variables for land？？use mixing degrees has the greatest positive correlation to commuter the demand, with influence coefficients 0.51 and 0.68 for O and D point, respectively. That means that commuter demand of the district is improved when the land？？use mixing degree is higher. The positive guidance of high？？intensity land？？use development and abundant public transport resources in transit travel has been verified from the perspective of spatial measurement. It provides a quantitative basis for the analysis of the interactive relationship between land？？use