*4.1. Limitations and Assumptions*

This paper shows the application of two distinct traffic-modelling approaches with different levels of data availability.

The meso-scale traffic model used in Barcelona case study had very detailed traffic information across the city but the example does not represent the temporal aspect of a flood event. The impact assessment therefore only considers the effects during a flood event and not the recovery period after the event. In addition, within the Barcelona traffic model the criteria for determining hazard classification did not consider the length of the portion of the flooded road, which was an additional restriction applied to the Bristol case study, such that there could be an overestimate of flooded roads. Further assessment could investigate the reduction in perceived flood hazards with the length restriction included.

In contrast to Barcelona's traffic model, the Bristol model lacked both real traffic-count data and a pre-existing traffic model. Due to these challenges, the traffic model was built from the ground up using freely available Open Source software and data and deriving traffic flows from land-use classifications. As highlighted in Section 2.1.1 there were limitations with this approach when dealing with large traffic volumes and as such the simulations within the paper utilise relatively low traffic counts. For future assessment, a more detailed analysis of the performance of the network could be carried out with the aim of improving/optimising the network under standard dry weather conditions.

For further work and improvements, it would be of interest to see how the micro-scale model performs with real traffic-count data to determine both the volume of traffic over time and the routes/journeys taken by vehicles within the network and determine loss estimates and recovery times under flooded conditions accordingly. Moreover, where such data available, it would be of further interest to evaluate the effectiveness of land-use data for determining route distribution in comparison to real traffic data.

Additionally, in this paper, for the micro-scale simulation we have only considered the flood event occurring with a fixed duration (30 min) starting at one specific time (7 am) which is at the beginning of the rush hour scenario. The degree of disruption to traffic flows within the network would however be dependent upon both when the event occurs and for how long, therefore future work could examine the effects/sensitivity of the time of occurrence and duration parameters.

### *4.2. Verification of Results*

The costs of disruption to traffic flows to cities is generally quite high and within the UK, and has been shown to be within a range of 3–7% of the total accumulated estimated losses from flood events [27]. The flash floods and landslides that occurred as a result of high and prolonged precipitation in Catalonia in September 2006 resulted in the Consorcio de Compensaci´on de Seguros (CCS), the national insurance company paying out €55.9 million and resulted in bringing traffic to a standstill in Barcelona due to jams [28]. In the region of Co. Galway Ireland, the 2015/2016 floods were thought to have losses of €3.8 million of losses through traffic disruption [29]. The calculated losses for these events however are not limited solely to disruption of traffic as a result of standing water but also consider traffic light failures as in the case of Barcelona 2011 [5] and also the road closures due to potential risk to like from flooding. For example, the summer floods of 2007 resulted in the closure of the M1 in the UK for 40 h between junctions 31 and 41 due to the risk of a dam breach and the cost of this disruption alone was estimated to be £2.3 million [6].

The work outlined in this paper shows the potential of combining climate change data with flood mapping and traffic models as a means of assessing the possible implications change may have. For the Barcelona case study, the estimated losses with respect to the measured return periods seem to portray values within the orders of magnitude of similar climate events (as shown in Co. Galway flood event that was a 1 in 100-year event). As the data used for traffic model in the Bristol case study was assumed based on NRD data and with low traffic counts the estimated losses serve more as a benchmark/guide relative to the severity of input flood events to highlight the potential implications of climate change on traffic.

The model, however, highlights the impact of flooding is not limited to the period of time where there is standing water upon the roads surface but also post flood event as the network takes time to recover.
