Large-scale spatial population synthesis for Denmark

Abstract

The recent development in micro-based transport models is a major step towards an improved understanding of transport demand and its underlying drivers. By adapting a detailed geographical resolution level and a fine-grained social description of individuals it becomes possible to investigate distribution effects across social classes and geographical spaces, elements which were not possible to take into account until recently. However, the increasing amount of details comes at a cost. As the prediction-space is enlarged, models become increasingly dependent on the quality of inputs and exogenous model assumptions of which the formation of synthetic population forecasts is by far the most important one. The paper presents a coherent description of a large-scale population synthesis framework involving all relevant steps in the synthesis stages from target harmonisation, matrix fitting, post simulation of households and agents and reweighting of the final population. The model is implemented in the Danish National Transport Model and is aimed at predicting the entire Danish population at a very detailed spatial and social level. In the paper we offer some insight with respect to the propagation of sampling noise caused by the household simulation stage and a brief validation of the model when comparing a modelled 2015 population with observed data.

OLYMPUS-POPGEN: A synthetic population generation model to represent urban populations for assessing exposure to air quality

“…Several approaches can be implemented to model synthetic populations depending on the quantity, type and quality of data available [32]. Each has its own strengths and limitations, and we have explored the methods available to identify the most suitable and promising for our ambitions.…”

A synthetic population of Sweden: datasets of agents, households, and activity-travel patterns

“…Synthetic populations are crucial inputs for developing simulation models, particularly agent-based models, widely used in many research fields, such as transportation, land use, economics, and epidemiology. Although many countries have created synthetic populations (e.g., Switzerland [8] , Denmark [9] , and Germany [10] ), there is a lack of data representing Sweden's entire population to date. The data presented in the paper fills this gap.…”

Comparing Methods for Generating a Two-Layered Synthetic Population

“…The choice of appropriate method for generating a two-layered synthetic population is critical to transport modelers. This choice is partially driven by the amount, type and quality (representativeness and comprehensiveness) of available data (52). In order to facilitate this process and in addition to the method comparison presented herein, the decision tree shown in Figure 5 identifies the most suitable methods for generating a two-layered synthetic population, based on the following parameters: available input data (sample and marginals), number of attributes to be considered, and synthetic population size.…”