2.1 Challenge 1: access to landslide occurrence and impact records

Access to substantially complete landslide inventories is a challenge faced by the landslide community as a whole. Landslide inventory/database compilation methods (e.g. field surveying, remote sensing, archive) often have challenges surrounding completeness (e.g. Guzzetti et al., 2012) and may be spatially biased towards areas where landslides are easily observable or have affected humans, such as roads (e.g. Santangelo et al., 2010), which makes it difficult to definitively assess the scale of influence of transport network building upon landslide occurrence in comparison to surrounding areas.

In addition, the fact that the shift from hazard- to risk-based forecasting is relatively recent might explain why records of landslide impact on transportation networks are relatively short and unsystematic (Klose et al., 2014; Taylor et al., 2015). If impact is recorded at all, this may be done by different organisations (e.g. highways agencies) and only outline the primary physical impact rather than the secondary regional disruptions to the transportation network (Winter et al., 2016). As a result of incomplete records, it is likely that we are underestimating the influence of transport networks on landslide occurrence and the impact of landslides upon the transport network. Two papers in this special issue address this challenge of access to records:

• Voumard et al. (2018) attempt to address this challenge by developing a detailed database of landslides occurring on Swiss roads and railways over a 5-year period using digital news alerts. Their work particularly emphasises the importance of small landslides (<10 m³ volume of deposit) that occur along roads and railways, which may not meet thresholds (e.g. number of fatalities) for being recorded in other databases. Voumard et al. (2018) find along these road or railway transportation networks that small landslides (<10 m³) comprised 95 % of all events in their database and 75 % of the total direct cost of landslide impact to the transportation network each year. This finding departs from previously established landslide frequency-size statistics (e.g. Malamud et al., 2004), possibly indicating different processes governing landslide size along transportation networks compared to landslide processes at a broader regional scale. This work highlights the importance of systematically assessing all landslides that occur along roads or railways in a region and gives insight into the potential use of digital news for recording the impact of landslides.

• Ali et al. (2019) outline the construction of a landslide susceptibility map for the Karakoram Highway in Pakistan, using a novel approach to creating a landslide inventory from road clearance logs. Their research shows how in a data-poor environment, where access to substantially complete regional landslide inventories may not be possible, road corridors may be relatively data-rich for understanding landslide processes and constructing a susceptibility map in the road corridor. Ali et al. (2019) found that 60 % of the highway fell into high or very high landslide susceptibility classes, which suggests further management of slope stability is required for this key trade route connecting China and Pakistan to truly achieve sustainable development (Kreutzmann, 2009).

The work in this special issue addressing Challenge 1 shows how new methods to compile databases of landslide occurrence and impact can be applied in the Global South. Preliminary findings suggest that further understanding of landslide physical process within transportation corridors is required, which special issue papers in Challenge 2, in the next section, address.

2.2 Challenge 2: better understanding of links between landslide susceptibility and physical processes

A second challenge for understanding landslide–transport network interactions is that of understanding the links between landslide susceptibility and physical processes. Landslide susceptibility maps indicate where landslides are more or less likely to occur and thus underpin our understanding of transport network exposure to landslides. They can be created before a transportation corridor is built to aid planning or after building to understand which parts of the network are most exposed to landslides.

Reichenbach et al. (2018) show that since the 1980s there has been inertia in the input thematic variables and methods used, and often justification of methods is unclear, suggesting innovation is required. With regard to innovation in landslide susceptibility–transport network interactions, we argue that the treatment of transportation corridors as a separate susceptibility map is useful for understanding the local processes unique to transportation corridors. This means departing from the typical approach of developing regional-scale susceptibility maps where linear distance to roads is one of several thematic variables used.

In McAdoo et al. (2018) (addressing Challenge 3 in this special issue) they show that there is not a linear relationship between distance from a road and number of landslides observed. Reichenbach et al. (2018) argue that the physical processes and geomorphological settings do not follow linear patterns as we move away from roads. We argue the buffer zone around a transportation corridor is a unique and heterogeneous space, with some areas increasing landslide susceptibility (e.g. through poor drainage or slope modification) and other areas potentially decreasing landslide susceptibility (e.g. through remediation measures), and that these processes change over time. Four papers in this special issue attempt to address components of this complex challenge in different ways:

• Jeon et al. (2018) use 3D modelling to investigate the problem of leaky pipes causing subsidence in the cavity alongside railways, finding complex interactions between groundwater levels and distance between pipelines and roadbeds.

• Bordoni et al. (2018) develop a non-linear model of landslide susceptibility that explicitly models the connectivity of the landscape topography for allowing sediment to flow over a surface. For their study area, they found that the highest susceptibility portions of the road network appeared to be influenced by high sediment connectivity and steep slopes with a limited height, regardless of land use.

• Tseng et al. (2018) develop a mountain road susceptibility map for typhoon-triggered landslides in Taiwan and also investigate the characteristics of landslides in their inventory in relation to roads. Their findings highlight a number of processes: (i) landslides tend to be smaller in size within 100 m from the road, (ii) landslides occurring above mountain roads (i.e. between the road and the mountain ridge) tend to occur close to the road but did not increase in number after the typhoon, (iii) landslides occurring below mountain roads (i.e. between the road and the stream divide) tend to be more active following the typhoon. Changes to landslides below roads are attributed to changes in drainage from the road, a growing issue widely observed in other climate and geomorphologic contexts (Penna et al., 2014). More work is needed in this area, for which Tseng et al. (2018) provide methods to quantify and analyse different processes acting within the locality of roads.

• Meneses et al. (2019) also support our argument that the zone around roads needs special consideration. They constructed two landslide susceptibility maps: one with detailed local data and one with generalised land cover from CORINE, finding that the susceptibility map with the detailed land cover datasets performed better.

The papers in this special issue addressing Challenge 2 highlight the complex, site-specific processes determining landslide susceptibility in transportation corridors and propose a range of approaches for different settings. Special issue papers addressing Challenge 3, in the next section, build on Challenge 2 by discussing the unique setting of mountain roads.

2.3 Challenge 3: approaches to deal with the unique setting of mountain transport networks

Mountainous areas are unique in terms of their transportation networks and thus require different approaches from transportation research on urban and low-lying areas. This uniqueness stems from steep topography making building more challenging – resulting in a lower-density population served by a less dense network of smaller roads. The propensity of steeper topography to landslides (and future climate change affects landslide triggering) amplifies the problem of landslide–transport network interactions, as a network with less redundancy increases the likelihood of isolation in the event of a landslide disruption. Three of the papers in this special issue (McAdoo et al., 2018; Schlögl et al., 2019; Sudmeier-Rieux et al., 2019) address the challenge of mountain transportation networks:

• Schlögl et al. (2019) and Sudmeier-Rieux et al. (2019) emphasise the dependence of local communities on local transportation networks at both ends of the spectrum of economic development in their studies of Austria and Nepal, respectively.

• Schlögl et al. (2019) illustrate the potential for using agent-based traffic models to understand the impacts to mountain communities from road blockages at a fine spatial scale, revealing issues such as long daily commutes and dependence on specific industries (e.g. tourism).

• McAdoo et al. (2018) observe that in Nepal, where there are areas with more agricultural development (and thus a greater density of informal, minor roads), there are twice as many landslides within 100 m from the road network compared to in non-agricultural areas.

Each of these three special issue papers (McAdoo et al., 2018; Schlögl et al., 2019; Sudmeier-Rieux et al., 2019) highlights the issue that regional, major roads are given greater research attention, but local roads may be built and maintained poorly, challenging the livelihoods of local communities. Their work highlights the need to bring together research on economic development, road building and landslides and understand the bi-directional relationships between these factors. In the next section, Challenge 4, special issue papers present commentary and potential methods for looking at how management and development of mountain roads may change in the future.

2.4 Challenge 4: future drivers of landslide–transportation network interactions

The fourth research challenge we highlight is better understanding of future drivers of landslide–transportation network interactions. Three papers in this special issue address this challenge:

• Sudmeier-Rieux et al. (2019) (invited commentary) discuss trade-offs between expanding the road network of Nepal for economic development and the associated increase of landslide risk. There was a 1200 % increase in the local road network  between 1998 and 2016 in Nepal (DoLIDAR, 2016), and Sudmeier-Rieux et al. (2019) indicate that many of these roads are poorly constructed and become seasonally blocked by landslides, perhaps resulting in overall losses to the community reliant on those roads. Sudmeier-Rieux et al. (2019) highlight the ongoing decentralisation of power and the China Belt and Road Initiative as governance opportunities to improve practices of road construction.

• Li et al. (2019) highlight the rapid expansion of highways in China and the associated issue of exposing bare soil and subsequent erosion. Citing previous work by Wang (2006) for every kilometre of highway built, 0.05–0.07 km² of soil is exposed, causing 450 t of soil loss per year in China (Chen et al., 2010). Combined with rapid expansion of the road network and climate change, the contribution of highway building to soil loss could be significant. Using the Revised Universal Soil Loss Equation (RULSE), Li et al. (2019) show that under 20-year return period conditions, 7.1 % of slopes are considered high or extremely high risk of soil erosion.

• Schlögl and Matulla (2018) use a downscaled climate projection of rainfall to examine the risk of rainfall-triggered landslide events to the road and rail network of central Europe. They show that for low-lying areas, the change in rainfall (and thus landslide triggering) is relatively low. However, areas with higher elevation may experience at least 14 additional rainfall triggering events per year, and these changes can be expected in the near future. Using the case study area of the Black Forest mountains, they show that areas of steep topography tend to have lower density transport networks with less redundancy (i.e. alternative routes in the case of a landslide) in addition to higher landslide susceptibility and a greater increase in the number of landslide-triggering rainfall events, and thus this should be prioritised for planning.

The papers addressing Challenge 4 in this special issue highlight the competing demands (and potential feedbacks) between road building, economic development, climate change and landslides. Evidence in these papers indicates that integrated planning is required to ensure that the benefits of road building are not outweighed by increased impacts from landslides, especially considering that other papers in this special issue have highlighted the need for further understanding of the physical processes driving landslides within transportation corridors.