A surface runoff mapping method for optimizing risk assessment on railways

Abstract

Railways are critical infrastructures for the transportation of people and goods and network failures must be controlled in order to maintain safety and to limit economic losses. The railway network is exposed to natural hazards and particularly to intense pluvial runoff. Due to the complexity of the phenomenon, management of risks induced by pluvial runoff raises technical and scientific issues. An innovative method for runoff susceptibility mapping, called IRIP for "Indicator of Intense Pluvial Runoff", has been created and adapted to the railway context. The objective of this paper is to evaluate the relevance of the mapping method and to provide application advice. The mapping method is evaluated by comparison with the results of a hydraulic diagnosis, on a 20 km railway line, using quantitative and qualitative comparisons. On the basis of contingency tables, probabilities of detection (POD, railway sections exposed and detected by IRIP) and false alarm ratios (FAR, railway sections detected by IRIP whereas they are not exposed) are computed. POD range from 94 to 100% and FAR range from 20 to 26%. Then spatial information provided by the maps is compared with field observations and recommendations. It is shown that the mapping method can bring substantial contribution to risk identification and that the IRIP method can allow pushing forward the current risk reduction methods. Thus, the surface runoff maps open up new opportunities to manage surface runoff, such as targeting mitigation actions at the origin of the hazard in partnership with the other territory stakeholders.

Evaluation of a model for mapping intense pluvial runoff hazard using proxy data of runoff-related impacts. Application to the railway context

“…The IRIP model is briefly described here, but more details can be found in the literature (Dehotin and Breil, 2011;Lagadec et al, 2018). The present description is mainly taken from Lagadec et al (2018) that retained improvements proposed by 15 Lagadec (2017) to the IRIP model.…”

“…On the other hand, indirect information of runoff-related impacts can be more easily available as runoff may have damaging consequences such as flooding of buildings or of transport networks (roads or railways), mud flows, erosion, landslides. Information about these impacts can be collected and reported based on various media: post-event surveys to collect the location of impacts on infrastructures or on transport networks (Versini et al, 2010b;Naulin et al, 2013;Defrance et al, 2014;Lagadec et al, 2016bLagadec et al, , 2018, insurance claims on buildings or infrastructures (Moncoulon et al, 2014;15 Le Bihan et al, 2017), analyses of the press and social media (Llasat et al, 2013;Saint Martin et al, 2018;Petrucci et al, 2019) or citizen science (Gourley et al, 2010;Le Coz et al, 2016). All these data are referred to as "proxy data" in this paper.…”

“…Such approach has been extended for the evaluation of non-quantitative prediction model such as the IRIP model by Lagadec et al (2016b). It was further improved by Lagadec et al (2018) using a comparison with expert judgement, taking into account the vulnerability of the railway. However, these evaluations remained quite qualitative.…”

A method to use proxy data of runoff-related impacts for the evaluation of a model mapping intense storm runoff hazard: application to the railway context

“…On the other hand, indirect information on runoff-related impacts can be more easily available, as runoff may have damaging consequences such as flooding of buildings or of transport networks (roads or railways), mudflows, erosion, or landslides. Information on these impacts can be collected and reported based on various media: post-event surveys to collect the location of impacts on infrastructures or on transport networks (Versini et al, 2010b;Naulin et al, 2013;Defrance et al, 2014;Lagadec et al, 2016bLagadec et al, , 2018, insurance claims on buildings or infrastructures (Moncoulon et al, 2014;Le Bihan et al, 2017), analyses of the press and social media (Llasat et al, 2013;Saint Martin et al, 2018;Petrucci et al, 2019), or citizen science (Gourley et al, 2010;Le Coz et al, 2016). All these data are referred to as "proxy data" in this paper.…”

“…Such data have been used for the evaluation of quantitative flash-flood forecasting models (e.g. Gourley et al, 2010;Defrance et al, 2014;Javelle et al, 2014;Saint-Martin et al, 2016), road cutting warning models (Versini et al, 2010b;Naulin et al, 2013) or flooding impact models on buildings (Le Bihan et al, 2017). The evaluation is based on criteria that are used for the evaluation of meteorological or hydrological forecasts (WWRP/WGNE Joint Working Group on Forecast Verification Research, 2015), i.e.…”