Belgium: The Use of MATSim within an Estimation Framework for Assessing Economic Impacts of River Floods

With the history of river oods in Belgium and the signi cant probability that such events will again take place in the near future, assessment of both direct and indirect economic impact was deemed essential to allow formulation of an adequate policy program and e cient ood risk management. One proposal would assess ood risk at the micro-scale level: i.e., individual buildings for exposure analysis and direct economic damage estimation, individual companies for indirect economic damage estimation, 10 meter grid spacing for land-use modeling and individuals/ vehicles for transportation models. To enable this assessment, an integrated modeling framework combining di erent simulation theories from a multidisciplinary perspective is being developed. Figure 63.1 describes the procedure to measure the annual ood risk. A more detailed description of the whole modeling chain is available in Dewals et al. (2015). A basic modeling framework premise is that di erent spatial pattern ’families’ might in uence the damage intensity caused by river oods (e.g., land use change, transportation systems). In this chapter, we focus on how MATSim is being integrated into this overall framework, thus focusing on the TSA (Transport System Analysis) within the overall estimation procedure. For TSA, two con gurations (freight and passenger model) are distinguished. For the passenger model, a MATSim scenario is developed on a national scale to simulate travel demand at base year 2010 and its evolution during the following years. The main objective is to study the e ects of river oods on the transportation network and, consequently, on travel demand from an economic point of view. In addition, a freight travel demand model has been developed, to enable interactions between


Problem Statement
With the history of river oods in Belgium and the signi cant probability that such events will again take place in the near future, assessment of both direct and indirect economic impact was deemed essential to allow formulation of an adequate policy program and e cient ood risk management.One proposal would assess ood risk at the micro-scale level: i.e., individual buildings for exposure analysis and direct economic damage estimation, individual companies for indirect economic damage estimation, 10 meter grid spacing for land-use modeling and individuals/ vehicles for transportation models.To enable this assessment, an integrated modeling framework combining di erent simulation theories from a multidisciplinary perspective is being developed.Figure 63.1 describes the procedure to measure the annual ood risk.A more detailed description of the whole modeling chain is available in Dewals et al. (2015).
A basic modeling framework premise is that di erent spatial pattern 'families' might in uence the damage intensity caused by river oods (e.g., land use change, transportation systems).In this chapter, we focus on how MATSim is being integrated into this overall framework, thus focusing on the TSA (Transport System Analysis) within the overall estimation procedure.For TSA, two con gurations (freight and passenger model) are distinguished.For the passenger model, a MATSim scenario is developed on a national scale to simulate travel demand at base year 2010 and its evolution during the following years.The main objective is to study the e ects of river oods on the transportation network and, consequently, on travel demand from an economic point of view.In addition, a freight travel demand model has been developed, to enable interactions between

Data Collection
As inputs, MATSim requires a synthetic population (or travel demand) le, as well as the related transportation network.Unfortunately, no recent census is available for the rst input; the latest dates from 2001.To compensate, a synthetic population was derived from more recent travel surveys (e.g., Cornélis et al., 2012)

Network
The network data of Belgium, downloaded in 2015, is available online from the OSM server.It consists of 100 467 nodes and 232 715 links.Network quality is generally acceptable, according to many MATSim users, even if manual adjustment is necessary for speci c links.

Synthetic Population
Preparation of a synthetic population presents a signi cant challenge for this case study; only micro-data are available to enable population synthesis.From these partial views of the actual population, use of a Gibbs sampler enables the joint distribution (re-)construction.The outputs seem to be encouraging when comparing computed predictions to the reference dataset.Here, we propose testing the methodology by synthesizing some relevant variables for both transportation and urban systems simulations at the household level (see Figure 63.2).

Activity-Based Pattern Generation
A er the synthetic population has been generated, activity types, activity times and activity locations are generated and associated to the agents, using an activity-based pattern generator.Using a combined set of machine learning techniques, daily activity planners are generated for each agent.As shown in Figure 63.3, the model suggests some promising rst results.The activity-pattern generator is calibrated by using micro-data, such as activity travel diaries extracted from travel surveys.As outlined by Cools et al. (2011), uncertainties introduced by statistical distributions of random components in most activity-based models might be signi cant.Thus, some key indicators (e.g., sequences type proportions) will be investigated to measure micro-simulation error impact.

General Modeling Framework
In Saadi et al. (2014), the overall modeling framework is presented, as well as the integration of scheme components.This paper covers all concepts expected to be used in building the future MATSim scenario.Figure 63.4 is a partial view of the overall modeling framework being researched at the moment.

Modeling Network Disruption
As mentioned, this study also suggests modeling network inaccessibility occurring a er river oods.This approach assumes that link capacities subjected to river oods are reduced, depending on ood intensity.Given that damage is mainly a function of water depth, the idea is to intersect a steady-state inundation map with the transportation network or, at least, the area impacted by oods (Saadi et al., 2014).Then, an analysis extension will be achieved by including a time series of river oods for a better understanding of dynamic e ects: e.g., response to river oods propagation, return way and time to the new equilibrium point between transport supply and demand.

Next Development Steps
When the complete integrated agent-based transportation model is ready, combination with the land-use change CA (Cellular Automaton) based model proposed by Mustafa et al. (2014) will to allow more interactions between those two patterns.This connection will be the basis for an innovative micro-scale LUTI (Land-Use and Transport Interaction) model, allowing more accurate predictions about future river oods in uenced by di erent micro-scale patterns.