Factors influencing the quality of macroscopic travel demand models in person transport

出行需求模型是交通规划的重要工具，服务于交通政策决策的准备。它们应在数量上以及在运输过程和相互依存关系方面适当地复制现实情况。为了评估出行需求模型的质量，建模实践最常见的做法是在选定的横截面上比较观察到的和建模的交通量。一方面，这是因为交通量可以很容易地调查。另一方面，交通量代表了许多规划任务的决定性值，并被决策者用来确定特定状态的影响。由此产生的问题很少被讨论。旅行时间和旅行距离分布以及距离依赖的模式分割份额也越来越多地被考虑用于旅行需求模型的验证。然而，关于输入数据和建模假设如何影响模型质量的主要陈述尚未得到。本项目的目的是确定影响宏观出行需求模型质量的因素，并量化所选因素的重要性。影响模型质量的因素包括输入数据（土地利用数据、网络数据、行为数据）的精度、交通供给（网络的详细程度）和出行需求（交通分析区域、人群、出行目的、出行方式）的分割、对旅行者决策过程建模的假设以及模型的数值准确性。为了评估旅游需求模型的质量，对文献中提到的质量度量进行了分类和评价。然后将质量测度应用于实践中常用的宏观出行需求模型。所选的因素系统地变化，以便得出对这些因素重要性的陈述。为了衡量模型的质量，需要参考数据以具有代表性的方式揭示旅行者的实际行为。为此，使用了来自家庭调查的地理编码数据，并生成了来自虚拟现实的数据。因此，发表了关于影响因素的声明以及衡量和确保旅行需求模型质量的建议，为旅行需求建模领域的进一步研究提供了动力，并为运输建模的实践提供了建议。

Travel demand models are an important tool of transport planning serving the preparation of transport policy decisions. They should appropriately replicate the reality in quantitative terms as well as with regard to the transport processes and the interdependencies. For assessing the quality of a travel demand model the modeling practice most commonly compares observed and modeled traffic volumes at selected cross sections. One the one hand this results from the fact that traffic volumes can be easily surveyed. On the other hand traffic volumes represent the decisive value for many planning tasks and are employed by decision makers to determine the impacts of a specific state. Problems resulting from this are rarely discussed. Increasingly also travel times and travel distance distributions as well as distance dependent modal split shares are considered for the validation of travel demand models. However, principal statements how input data and modeling assumptions influence the model quality are not yet available.Objective of this project is to identify the factors influencing the quality of macroscopic travel demand models for person transport and to quantify the importance of selected factors. The model quality is influenced by the precision of the input data (land use data, network data, behavioral data), by the segmentation of the transport supply (level of detail of the network) and the travel demand (traffic analysis zones, person groups, trip purposes, transport modes), by the assumptions for the modeling of the decision processes of travelers as well as by the numerical exactness of the model.In order to assess the quality of travel demand models quality measures referred to in literature are classified and evaluated. The quality measures are then applied to macroscopic travel demand models commonly used in practice. Selected factors are varied systematically in order to derive statements on the importance of these factors. To measure the quality of a model reference data are required revealing the actual behavior of travelers in a representative manner. For this geocoded data from household surveys are used and data from a virtual reality are generated.As a result statements on the influencing factors and recommendations for measuring and ensuring the quality of travel demand models are issued offering impulses for further research in the field of travel demand modeling and advice for the practice of transport modeling.