交通驱动城市混合土地利用形态演化机理与发展预测研究

Literature review shows that most of past studies concentrate on how mixed land uses impact travel behavior and study results prove that mixed land-use plays an important role in reducing long-distance trips and those based on private vehicles. However, no study has been found to quantify and forecast the “stimulating effect” of the transport system to mixed land use pattern (including its range, variety and intensity), which is an important scientific question to be answered in the process of planning mixed land-use and improving the efficiency of urban land-use. In the proposed research, a large number of variables related to transport system and mixed land use will be collected and important factors shaping urban mixed land-use patterns will be systematically examined and their impact mechanism will be separately analyzed through System Dynamics (SD) models. Specifically, structure equation (SE) models will be used to categorize the factors identified above into endogenous and exogenous variables of land-use-transport system and quantify the driving force of the exogenous varialbes to the endogenous variables and to the development of mixed land-use pattern. In addition, a series of models, ranging from contiuous nested logit (CNL), artificial neural network (ANN) and supporting vector machine (SVM) will be developed based on identified endogenous variables above and their forecast results will be further adjusted based on the quantified stimulating effect from the exogenous to endogenous variables obtained from the SE models, in order to systematically and accurately forecast mixed land-use pattern under different levels of transport supply. The models developed will be tested in a set of microsimulations at a fine spatial resolution (e.g., parcel or block level), in order to visually demonstrate the impact of transport supply and the corresponding dynamic development process of mixed land-use pattern. The results from the proposed research are expected to provide a set of advanced theories and forecasting models for better planning/designing mixed land uses in the “New Urbanization” process of China.

现有研究大多聚焦混合土地利用形态对出行行为的影响，缺乏交通如何影响混合土地利用形态、量化与预测交通驱动作用方面的理论研究，而这也正是合理规划我国混合土地利用、提升大中城市交通与土地利用效率的关键科学问题。本研究将基于交通和混合土地利用相关空间大数据，通过系统动力学理论系统地分析城市混合土地利用形态的影响要素与作用机理，通过结构方程辨析混合土地利用系统内外生变量并量化城市交通供给水平等外生变量对内生变量及城市混合土地利用形态演化的驱动作用；构建基于土地利用系统内生变量的连续巢式Logit、神经网络、支持向量机等城市微观（如地块或街区水平）混合土地利用智能体预测模型，并通过基于结构方程的交通综合驱动混合土地利用作用机理演化模型修正预测结果，实现在不同交通供给水平下城市混合土地利用形态的科学预测；通过混合土地利用微观动态仿真，形象地表现城市混合土地利用形态的具体变化过程和交通驱动效果。

城市混合土地利用是我国“土地资源缺乏”国情对大中城市建设发展的客观要求，也是实现“新型城镇化”国家战略的主导模式。然而，现有研究大多聚焦于混合土地利用形态对城市出行行为的影响，缺乏城市交通影响混合土地利用形态机理以及交通驱动作用下混合土地利用形态预测方面的研究。本项目首先在分析混合土地利用形态特征的基础上，构建了系统动力学模型，对混合土地利用形态影响因素与作用机理进行了研究；其次，提出“时变”可达性的概念，并分析了“时变”可达性与各类土地利用的相互关系，揭示了城市交通驱动混合土地利用形态演化机理。再次，利用地理加权回归和深度神经网络构建了综合交通驱动作用下的土地利用形态预测模型。同时，基于PECAS整体规划建模理论与框架，构建了武汉市、武汉市长江新城以及长江经济带PECAS模型，并对未来数十年的土地利用形态与客货运输需求进行了预测。最后，在地块的微观层面上，对城市土地利用精细分类进行建模研究，并对土地利用形态空间开发进行了仿真实证研究。以上研究成果为我国城市与区域的整体规划与管理奠定了坚实的决策支持工具基础，对于我国城市与区域的可持续发展具有重要的应用价值。