1 Introduction

Roads are critical infrastructure with high importance for social and economic well-being. In coastal areas, road networks are at risk of exposure to episodic flooding and permanent inundation from relative sea level (RSL) rise (Neumann et al. 2021). Evaluating the future timing and severity of flood and inundation exposure along national road networks informs actions and strategies to avoid or mitigate adverse social and economic consequences as future RSLs change local flood regimes.

Compared to coastal populations and other high-value socio-economic assets (e.g., buildings), road network impacts from extreme sea level (ESL) driven episodic flooding and RSL change at national or supra-national levels have received less attention. Nationwide studies in several countries have used one-dimensional flood maps to enumerate scenario-based (e.g., 1 in 100-year) exposure of roading components. In New Zealand (Paulik et al. 2020) and Norway (Breili et al. 2020), 1 m RSL rise could increase present-day national road exposure to 1 in 100-year and 1 in 200-year extreme sea-levels by more than 100% (~ 2856 km) and 162% (~ 1340 km), respectively. Despite the significant socio-economic implications, detailed investigations are often limited to small geographic areas (e.g., state, city, urban), whereby direct or indirect economic damage arising from service disruption and traffic congestion is simulated (Sadler et al. 2017; Jacobs et al. 2018). These evaluations demonstrate utility for local scale decisions on road construction investment and maintaining levels of service however, their complexity often restricts spatiotemporal scales required for large-scale road construction investment and management practices. In this present study, we formulate a spatiotemporal risk model approach to rapidly quantify economic losses for numerous episodic flood events and RSL scenarios that may affect national road networks.

Episodic flood risk evaluation at annual frequencies is a common practice used by catastrophe modelers in insurance and reinsurance sectors to inform risk-based asset insurance pricing. Annualised road economic losses from coastal flooding have received limited attention in academic studies, with investigations focusing on either deterministic impact using ‘what if’ scenarios, or location-specific probabilistic risk analysis. Deterministic studies often focus on low probability, high impact episodic flooding events (e.g., 1 in 100-year) under present-day and future RSL scenarios for specific years or decades (Lan et al. 2023). The direct and indirect economic impact of high probability nuisance flooding from projected tidal water level changes this century has been estimated for US road networks based on local tide gauge records (Sweet et al. 2018; Jacobs et al. 2018; Fant et al. 2021). These studies infer road exposure to inundation using the hourly empirical cumulative density function (CDF) of tide gauge water levels to estimate annual economic loss from physical road damage and service disruption. In the US, Fant et al. (2021) estimated the national road network’s present-day annual $1.3 to $1.5 billion indirect losses from traffic delays could increase to $220 to $260 billion in 2100 in response to projected RSL change for medium to high greenhouse gas emissions scenarios. Jacobs et al. (2018) also investigated tidal effects on inundation traffic delays on the US East Coast, observing that current (2010) ~ 160 million vehicle-hours could reach 3.4 billion by 2100 under a medium emission scenario. The annual economic losses caused by high-probability events are a significant concern for network operators and road service users. A holistic risk analysis, however, represents local variability of high (e.g. < 10 years) and low-frequency flooding events (e.g. > 100 years), with the latter exceeding the potential design life of road network components. Risk analyses considering local flood regime variance are critical to optimise road management interventions across different spatiotemporal domains and uncertain future RSLR projections.

This study evaluates direct economic risk from extreme sea levels and relative sea level (RSL) change for New Zealand’s national road network. A spatiotemporal risk analysis framework was designed and implemented in RiskScape, a spatial modelling tool for evaluating single or multi-hazard risk (Paulik et al. 2023). Road direct economic risk was calculated as monetary loss from physical damage and reported as exceedance probability loss (EPL) and expected average annual loss (AAL). Monetary losses and their uncertainty were computed using a quasi-random sampling approach, with losses enumerated for different Shared Socioeconomic Pathway (SSP) scenario projections of RSL change over 100 years in response to global mean sea level rise (GMSL) and local vertical land motion (VLM) (Naish et al. 2024). We estimate the timing and magnitude of direct economic risk change at a road component level in response GMSL alone and GMSL with VLM in combination, to enumerate risk change at national and regional levels in response to RSL change under future climate conditions. Our national risk evaluation provides insights at different jurisdictional levels on road management implications from future flood hazard regime response to local relative sea level change.

2 Materials and methods

Flood risk in this study is the frequency and magnitude of direct tangible damage caused by spatio-temporal interactions between flood hazard processes (e.g., water depth) and road network components with an inherent vulnerability to damage when exposed to flood hazards. We analyse flood risk using the RiskScape multi-hazard model framework (Paulik et al. 2023), which is a modular and configurable system designed to evaluate coastal flooding exposure and impacts from extreme sea levels (ESLs) and sea level rise (SLR) under current and projected climate conditions. The model workflow steps used in this study are represented in Fig. 1, with input data, step functions and output data described in Sections 2.1 to 2.3.

Fig. 1
figure 1

Schematic representation of the spatiotemporal risk framework and model steps applied in this study

Full size image

2.1 Model step: input data

2.1.1 Hazard

Spatiotemporal maps of permanent tidal inundation and episodic flooding from extreme sea-levels (ESLs) for present-day and higher sea levels were obtained from Paulik et al. (2023). ESL flooding maps represent 2, 5, 10, 20, 50, 100, 200, 500 and 1,000 annual recurrence intervals (ARI). The mapping process involved extracting digital elevation model (DEM) raster cells situated below ESL and elevations using a static inundation mapping technique (Stephens et al. 2021). A comprehensive national DEM for coastal regions (Paulik et al. 2020, 2021), up to an elevation of 20 m above present-day mean sea levels, was established by amalgamating LIDAR DEMs resampled to a 10-meter resolution and employing a fully convolutional neural network (FCN) model to rectify vertical biases in the Shuttle Radar Topography Mission (SRTM) data (Meadows and Wilson 2021). Flood depth calculations were performed by computing the disparity between ESL water surface heights and land elevations for DEM grid cells. To ensure accuracy, only grid cells with a hydrologic connection to coastlines were considered, thereby minimizing the risk of overestimating inundation extents. Additionally, topographic protection structures like levees were identified from aerial imagery and incorporated into the analysis, albeit without detailed design-level information. Consequently, land protection was assumed up to ESLs corresponding to a 100-year recurrence interval at present-day mean sea level (MSL). This was consistent with statutory flood hazard risk management directed by the New Zealand Coastal Policy Statement (Department of Conservation 2010), which requires regional and local authorities to avoid increasing the risk of social, environmental and economic harm from coastal hazards, over at least a future 100-year timeframe. Future ESL flooding maps were generated incrementally for MSL increments ranging from 0.1 to 2 m, regardless of future projections of RSL rise.

New Zealand RSL projections for Shared Socioeconomic Pathways (SSPs) were obtained from probabilistic RSL projections estimated using the Framework for Assessing Changes to Sea-level (FACTS) from the IPCC Assessment Report 6 (Levy et al. 2023; Naish et al. 2024). These projections, incorporating VLM effects, were based on interferometric synthetic aperture radar (InSAR) data calibrated with campaign and continuous Global Navigation Satellite System (GNSS) measurements. In this study, we investigate land movement effects on direct economic risk timing and magnitude by selecting RSL projections with and without local VLM for SSP 1 to 5 over 100 years (2020–2120) relative to a 1995–2014 baseline period (0 m at year ~ 2005). RSL projections from Naish et al. (2024) were obtained for 7435 coastal sites from https://searise.takiwa.co/. Medium confidence climatic processes for each SSP scenario were represented by RSL curves, which were spatially joined to flood-exposed road segments to evaluate the future timing of expected economic losses.

2.1.2 Exposure

Nationwide road network component information was obtained from Waka Kotahi NZ Transport Agency (NZTA). The object-based dataset represents continuous road segments between intersections. Road segments include several attributes for replacement value estimation. Roads were defined as eight classes based on the importance of their route and daily traffic volume (DTV). Classes included: national and high volume (15,000–35,000 DVT), regional (5,000–15,000 DVT), arterial (3,000–5,000 DVT), primary collector (1,000–3,000 DVT), secondary collector and access (200-1,000 DVT), and low volume (< 200 DVT). Road segment width is reported in meters (\(\:{R}_{LW}\)). Bridge object locations and attributes were not available for this study, nor could be accurately represented as road segments without corresponding platform heights above ground level. Road class unit cost rates per road width meter for 2020 NZD first quarter (\(\:{R}_{\$LWm}\)) were determined using the NZTA elemental cost manual (Waka Kotahi NZ Transport Agency 2021). We calculated road class replacement cost values per segment (R) using the general formula:

$$\:R=\:\frac{{R}_{\$LWm}}{{R}_{LW}}\:\:\:\:$$
(1)

2.1.3 Vulnerability

Road vulnerability to direct damage from episodic flooding was represented by damage curves. Synthetic curves were developed to evaluate relative damage as a dimensionless ratio representing the component ‘cost to repair’/’cost to replace’. The approach presented by Van Ginkel et al. (2021) was adopted to derive mean expected damage curves for New Zealand component classes exposed to high and low-intensity hydrodynamically driven processes (Fig. 2). Road damage from high-intensity processes (e.g., wave action, flow velocity, scour) was assumed from water depth above road level. Low-intensity curves were applied for water depths < 1 m or where roads border estuarine environments (Paulik et al. 2023). High-intensity curves default to other coastal areas where water depths > 1 m. These water depth thresholds were defined using coastal road damage observed from medium size (i.e., inundation depths < 5 m) tsunami events (Williams et al. 2020a, b, 2024), whereby corresponding flow velocity, hydrodynamic force and momentum flux intensities cause physical damage to road surfaces or embankments. Road damage curves demonstrate a monotonic trend with variable relative damage rates in response to increasing water depth and assumed road class sensitivity to other hydrodynamic processes.

Fig. 2
figure 2

Synthetic relative depth-damage curves used in this study for New Zealand road classes

Full size image

Road component damage estimation uncertainty occurs in response to variance in construction costs and damage response to hydrodynamic conditions represented in damage curves. Direct monetary loss uncertainty was computed using a quasi-random sampling approach. Uncertainty distributions were derived for each sampled road by (1) selecting a high-intensity or low-intensity curve based on sampled water depth, and (2) damage ratios were randomly sampled for percentiles within the z space between damage curves (i.e. high-intensity = p50, p75, p100; low-intensity = p0, p25, p50), (3) calculate direct monetary loss for each sampled damage ratio multiplied by a replacement cost value (R) sampled from random choice within the estimated replacement cost range. Assuming a truncated (0–1) normal distribution, 1000 quasi-random samples per road segment (Section 2.2) were derived within 2 standard deviations of the mean direct monetary loss. These uncertainty assumptions were necessary as detailed economic loss estimates from historic flooding events are not readily available.

2.2 Model step: geoprocessing and sampling

The geoprocessing and sampling steps transform model input data geometries to extract spatial information for consequence and risk analysis (Paulik et al. 2023). Model data was represented as raster grid (hazard), and vector line (roads), and polygon (jurisdictional regions) geometries. Road lines as ‘centrelines’ were converted to polygons by buffering the road line by its width in meters. Road polygons were then ‘cut’ as grid segments by the hazard grid geometry to sample spatial information from intersecting hazard grid and jurisdictional region polygons. The sampling process created a georelational coverage data file by converting relevant hazard and jurisdictional region information into indexed values at defined locations, e.g. road segment centroid. Indexed values were then extracted to the coverage data file. Lookup functions are then used access to hazard and jurisdictional region information for each road segment exposed to coastal flooding. This process facilitated both the calculation of descriptive statistics for consequence and risk analysis model step, and model result output reporting.

2.3 Model step: consequence and risk analysis

The risk analysis step quantifies the monetary loss occurrence frequency for exposed elements based on hazard event probabilities. Direct monetary loss (\(\:{\text{R}}_{\text{L}}\)) were calculated using the unit loss method:

$${\text{R}}_{\text{L}}=\sum\limits_{i=1}^{N}{R}_{i}{\:f}_{Di}\left({WD}_{i}\right)$$
(2)

where for road segment i, \(\:{\text{R}}_{\text{L}}\:\)is enumerated based on replacement value \(\:R\) and sampled water depth \(\:WD\) determines the relative damage from a corresponding damage curve \(\:{\:f}_{D}\). \(\:{\text{R}}_{\text{L}}\) frequency was calculated as exceedance probability loss (EPL) and average annual loss (AAL). A Poisson model was first used to calculate the probability of occurrence for the period (T) representing each hazard recurrence interval. If for example a hazard event has an annual frequency (\(\:\lambda\:\)) of 0.01 (i.e. 100-year ARI) the probability of occurrence \(\:P\:\)in a given year is defined by:

$$\:P=1-exp\left(-\lambda\:*T\right)$$
(3)

\(\:P\:\)is then used to calculate EPL for the independent variable \(\:{R}_{L}\) and hazard event probability of occurrence \(\:P\:\)as follows:

$$\text{E}\text{P}\text{L}=\sum\limits_{i=1}^NR_L\left(P\right)$$
(4)

where \(\:{R}_{L}\left(P\right)\) is the direct damage annual probability of occurrence and N is the sum of damage road segments i exposed by the hazard event with probability of occurrence \(\:P\). A hypothetical loss curve is formed between \(\:P\) and \(\:\text{E}\text{P}\text{L}\) with a positive monotonic trend in response to decreasing \(\:P\). The expected AAL is then estimated using trapezoidal integration to compute the area under the curve:

$$\:\text{A}\text{A}\text{L}={\int\:}_{{P}_{\text{m}\text{i}\text{n}}}^{{P}_{\text{max}}}EPL\left(P\right)$$
(5)

where \(\:{P}_{\text{m}\text{a}\text{x}}\) represents the highest hazard event occurrence probability and \(\:{P}_{\text{m}\text{i}\text{n}}\) represents the lowest event occurrence probability. \(\:EPL\left(P\right)\) denotes the sum of monetary loss for a hazard event with the probability of occurrence \(\:P.\) AAL was then estimated by solving the integral in Eq. 5 using the trapezoidal Riemann sum approach. Finally, EPL and AAL were enumerated for 17th, 50th, and 83rd percentiles and reported at national and regional levels.

3 Results

In Section 3, direct economic loss from episodic flooding is reported at the national and regional level for over 100 years between 2020 and 2120. We report EPL and AAL for medium confidence SSP 2–4.5 and 5–8.5 RSL projections in response to global mean sea level rise (GMSL) alone and combined with local vertical land motion (VLM). We also focus on reporting 100-year ARI losses due to the high global attention for coastal flood risk studies (Jongman et al. 2012; Hallegatte et al. 2013; Vousdoukas et al. 2020; Bates et al. 2021).

3.1 Spatiotemporal economic loss change from low-frequency episodic flooding under future relative sea level change

New Zealand’s national road network is expected to experience an accelerating rate of economic losses from 100-year ARI coastal flooding toward the end of this century. Projected RSL change between 2060 and 2100 is expected to increase losses to 100-year ARI events by between 40 and 90% (p50) from GMSL rise alone, and 40–90% (p50) from VLM effects for SSP 2.4–5 and SSP 5.8-5, respectively. Expected 100-year ARI losses at 2100 may also occur 12 (SSP 2.4–5) and 8 (SSP 5.8-5) years earlier as VLM accelerates RSL rise later this century. After 2100, a slight reduction in VLM effects is observed whereby the SSP 5.8-5 p83 loss difference with GMSL rise alone decreases from 19 to 14% at 2120.

National economic losses from 100-year ARI events are primarily driven by arterial and access road damage (Fig. 3). Under SSP 2–4.5 and SSP 5–8.5, the expected losses to these road classes increase by over 50% between 2060 and 2100 when local VLM is considered in projected RSLs. Similar proportional loss increases are observed for secondary collector roads. National and high volume roads could expect the lowest absolute losses and proportional loss change for 100-year ARI events between 2020 and 2120. At the national level, downward VLM caused expected losses for most road classes to accelerate relative to GMSL alone after 2060 (Fig. 3).

Fig. 3
figure 3

National projections of combined direct economic loss (EPL) for road class exposure to 100-year ARI coastal flooding

Full size image

The Auckland, Waikato, Hawkes Bay, and Canterbury regions account for 60% of direct economic losses for 100-year ARI events at 2060 under SSP 2.4–5 and SSP 5.8-5 RSL projections, lowering slightly to 58% by 2100. Waikato observes the highest regional losses to 100-year ARI events up to 2060, with Canterbury showing a significant acceleration in losses thereafter (Fig. 4). Upward VLM slows the rate of loss increase from 100-year ARI events in Waikato and Bay of Plenty regions, with expected losses from GMSL rise at 2100 may occurring up to 5-years later. Downward VLM may exacerbate RSL rise in Auckland, Canterbury, Hawkes Bay, Tasman and Wellington after 2060 for SSP 2.4–5 and SSP 5.8-5, increasing 100-year ARI event losses (p50) by 25–85%. Several regions (Gisborne, Taranaki and Southland) demonstrate relatively low loss potential (< NZD 20 million) in response to local RSL change toward the end of this century.

Fig. 4
figure 4

Regional projections of combined direct economic loss (EPL) for road class exposure to 100-year ARI coastal flooding

Full size image

3.2 Expected annual economic loss change from episodic flooding events under future relative sea level change

Direct economic average annual loss (AAL) could exceed NZD $86 million ($72–112 million) and NZD $119 million ($96–164 million) by 2100 in response to GMSL change under medium confidence SSP 2–4.5 and 5–8.5 respectively (Fig. 5). Downward land motion accelerates AAL after 2060, adding a further 12% (−4–24%) to 15% (4–21%) at 2100. The VLM effectincreases the national expected AAL to NZD $97 million ($71–112 million) and NZD $138 million ($99–199 million) by the end of this century.

Fig. 5
figure 5

National projections of combined direct economic average annual loss (AAL) for road class exposure episodic coastal flooding

Full size image

Access roads combined with primary and secondary collector roads serving populations of less than 10,000 people contribute on average between 47% and 50% of AAL each year during this century under SSP 2–4.5 and 5–8.5 RSL projections. This equates to NZD $40 million ($34–53 million) and NZD $57 million ($45–79 million) for SSP 2–4.5 by 2100. Arterial roads further contribute 20–22% of AAL in any one year during this period. Higher volume national and regional roads connecting population centres exceeding 30,000 people comprise 7–12% of national AAL at 2060, decreasing slightly (~ 2%) by 2100. Similar proportional AAL increases occur for access, primary collector, and secondary collector roads suggesting local transfer of economic losses.

The expected AAL increase later this century is highly variable at regional level. Waikato observes a 4 to 6-fold AAL increase between 2020 and 2100 under SSP 2–4.5 and 5–8.5 RSL projections, primarily driven by GMSL (Fig. 6). In this period, arterial roads account for 30–33% of AAL each year. Several regions including Auckland, Canterbury, Hawkes Bay, Tasman and Wellington demonstrate accelerating AAL after 2060 in response to GMSL rise and downward VLM. In Canterbury, the 2 to 3-fold AAL increase each year between 2060 and 2100 is driven by damage to secondary collector and arterial roads (~ 50%). Canterbury further accounts for ~ 60% of AAL for national roads during this period for RSL projections under SSP 2–4.5 and 5–8.5 scenarios. Several regions (e.g., Gisborne, Southland, West Coast) observing low AAL increases < NZD $5 million by 2100 for SSP 5–8.5 RSL showed local VLM has minimal influence on future road damage from episodic coastal flooding.

Fig. 6
figure 6

Regional projections of combined direct economic average annual loss (AAL) for road class exposure episodic coastal flooding

Full size image

4 Discussion

This study contributes to growing global evidence of accelerating direct economic loss from relative sea-level change later this century. High-value built-asset risk from episodic coastal flooding often evaluates global mean sea level rise (GMSL) independent of local vertical land motion (VLM). Our national study has demonstrated that VLM causes high spatiotemporal variance in expected economic loss occurrence during this century. Under SSP 2–4.5 and SSP 5–8.5 RSL projections, expected average annual loss (AAL) at 2100 for New Zealand’s roads could occur up to 10 to 20 years earlier when VLM is considered with GMSL. Interregional variance showed downward land motion may cause the expected AAL at 2100 up to 20 years earlier in Hawkes Bay and Wellington and 5 years later in Waikato and Bay of Plenty from upward motion. The variable episodic coastal flood risk timing and magnitude estimated this century demonstrates the importance of local VLM trajectory on the rate of RSL change (Wöppelmann and Marcos 2016). Despite recent advancements in quantifying VLM along global coastlines, VLM control on economic risk from either permanent inundation or episodic flooding is often inferred from changing hazard exposure of land (Blackwell et al. 2020; Sherpa et al. 2023; Ohenhen et al. 2023). Our findings propose a more routine consideration of VLM in RSL change in spatiotemporal coastal flood risk evaluation for roads and other horizontal or vertical components of critical infrastructure networks.

Changing coastal flood regimes in response to future RSLs will stress national and subnational road asset management budgets, particularly when serving asset repairs from multiple climatic and non-climatic hazard phenomena. Since 2009 the national annual emergency works budget for all hazards in New Zealand has ranged between NZD $100 million to NZD $160 million, with actual expenditures ranging from NZD $86 million to NZD $733 million (Waka Kotahi NZ Transport Agency 2024). We demonstrate that due to RSL change, average annualised losses caused by episodic coastal flooding alone could absorb the present-day national emergency works budget by 2100. This signals a need for decisions on continued higher investment in expected repair costs due to flood regime change or interventions to counteract these costs. As several regions (i.e., Canterbury, Hawkes Bay, Waikato) accounted for over half of the annualised loss change this century, the spatiotemporal risk model framework should be extended to evaluate local and national economic benefits created from adaptation interventions on specific high-risk roads in these regions. Ideally, such analyses should evaluate indirect costs and risk posed by other hazard phenomena exposing regional road networks to realise the potential benefits from coastal road protection (Argyroudis et al. 2020).

The future timing, location, and rate of RSL change along national coastlines is inherently uncertain, particularly toward the end of this century (Naish et al. 2024). Considering a broad range of episodic flood events and RSL scenarios is critical for infrastructure network operators to develop adaptation strategies that adapt components and services to future risks across time and space. While this study delivered a component-level approach to model direct economic risk from episodic flooding for a national infrastructure network, evaluating risk over a 100-year period is highly uncertain for several reasons. Direct economic losses from tidal inundation were not evaluated in the present study. Tidal inundation is the focus of several indirect road loss studies to date, and has been demonstrated to cause a potential 170-fold increase in economic loss by 2100 in the USA (Fant et al. 2021). Our national road network data did not represent future road component change in flood-prone areas including road development, redesign or relocation to avoid or mitigate tidal inundation or new construction to service future coastal development and population growth expected during this century. These phenomena occur beyond multi-year or decadal network management plans and are rarely evaluated through quantitative models. We contend however, the present model approach is extensible to incorporate future road network and component design and vulnerability to simulated hazard conditions under RSL-driven tidal and episodic flooding regime change.

Economic risk models at national and supra-national levels require a trade-off between data resolution and model accuracy. Improved access to road network spatial data through government or crowd-sourced mapping initiatives such as OpenStreetMap (OSM) facilitates ‘object-based’ approaches for large-scale flood risk analysis (Koks et al. 2019; Van Ginkel et al. 2021). Our spatiotemporal model simulated road direct economic loss at meter scale for enumeration from road segment to national levels. This approach supports both national and local risk analysis practices such as future ‘hotspot’ identification (Thacker et al. 2017), and cost-benefit analysis (CBA) to identify optimal structural and non-structural adaptation interventions to meet community service expectations (Fant et al. 2021). Model interoperability with local dynamic flood inundation models and/or network service flow models to simulate indirect economic loss from road service disruption (Pant et al. 2018) will deliver a more accurate and holistic national and subnational road network risk.

Several limitations must be acknowledged for future improvement of the model approach presented. Judgment-based vulnerability models were used as historical data was absent to establish empirical relationships between flood hazard processes, physical road component damage, and economic loss. Similarly, reliable economic loss estimates from network exposure to historic events analogous to the simulated ARI scenarios limited opportunities for local validation of modelled loss estimates. We then emphasise the current study’s purpose to create a new understanding of national road network risk to episodic flooding and sea level rise over the next century, using a general-purpose spatiotemporal model and consistent hazard, exposure and vulnerability input data. Further model performance evaluation is encouraged through post-event flood damage data collection and validation, while the modular approach supports input data ‘localisation’ to ensure community investment in high-resolution local models and data is utilised within a national coastal flood risk analysis.

5 Conclusions

This nationwide study evaluated the future direct economic risk to episodic coastal flooding for New Zealand’s road network. A new spatial risk analysis framework modelled monetary loss from physical road damage at meter scale as exceedance probability loss (EPL) and average annual loss (AAL). The model produced monetary losses for nine annual recurrence interval (ARI) extreme sea levels and twenty-one time-independent sea level rise scenarios. This approach facilitated direct economic loss change estimation for episodic coastal flooding over a selected 100-year period (2020–2120) in response to global mean sea level rise (GMSL) and local vertical land motion (VLM).

The future timing and magnitude change of 100-year ARI and AAL economic losses for the national road network were calculated for medium confidence Shared Socio-economic Pathways (SSP) 2–4.5 and 5–8.5 scenarios. Our national 100-year ARI losses at 2100 estimated for GMSL alone may occur up to 12 years earlier as local downward VLM accelerates RSL rise later this century. Expected national road AAL at 2100 could also occur 10 to 20 years earlier. Approximately half of expected national economic losses in any one year occur for access and collector roads serving populations of less than 10,000 people. Local upward land motion could delay expected regional level road AAL from GMSL by 5-years (Bay of Plenty and Waikato) at the end of this century, or advance AAL by up to 20 years (Hawkes Bay and Wellington). Regional loss variability warrants a need for local and national road and other critical infrastructure network asset managers to consider VLM controls on future RSL change in economic risk evaluations of episodic coastal flooding.

The spatiotemporal model approach in this study is extensible for future flood risk evaluations of critical infrastructure networks. Input data resolution for flood hazard, element exposure (i.e., bridges), and vulnerability (i.e. damage curves) available or developed for specific risk contexts can be easily migrated into the model workflow, replacing lower-resolution data. The object-based model can also be extended to operate with network service flow models that simulate indirect economic loss from road service disruption and, evaluate optimal component-level adaptation interventions to minimise economic and service loss. Future efforts to improve modelled economic loss prediction accuracy should satisfy the information resolution requirements for risk research or practice activities, and ensure validation is possible using empirical data or expert-knowledge of economic losses sustained from analogous historical flood events.