Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

A global model to forecast coastal hardening and mitigate associated socioecological risks

Abstract

Around the world, coastal urbanization continues to replace natural marine habitats with engineered structures, resulting in wholesale changes to shallow-water ecosystems and associated socioecological impacts. This process is expected to continue over the coming decades. The development of meaningful strategies to minimize future impacts requires an understanding of the rate at which ‘coastal hardening’ will take place regionally. Here we show that coastal infrastructure has replaced more than half (52.9 ± 4.9%) of the coastline associated with 30 global urban centres. The regional extent of coastal hardening is explained by eight predictor variables associated with shipping, boating, regional economies, populations and coastline length. Using a case study approach, we forecasted a 50–76% expansion of coastal infrastructure over a 25-year period. Our model can aid decision-makers to anticipate increases in coastal hardening, supporting identification and management of future threats to coastal ecosystems alongside social, economic and cultural objectives.

Main

In urbanized coastal areas around the world, seawalls, breakwalls, wharves, piers, pontoons and other structures (‘coastal hardening’) have replaced rocky reefs, sandy beaches, mudflats and mangroves. This process has been driven by the development of civic infrastructure, recreational and commercial shipping facilities, and the need for coastal protection1,2. Globally, the present (2018) footprint of engineered coastal and offshore structures (including ports, energy extraction infrastructure and telecommunication cables, among others) is estimated to be at least 32,000 km2, probably representing an underestimate2.

Engineered coastal structures have very different physical characteristics than the natural habitats they replace, often resulting in an extensive reduction in intertidal and shallow subtidal areas, and shifts from soft sediment to hard substrates and from natural (that is, rock, sand) to artificial (that is, concrete, plastic, steel) substrates3,4. These shifts can drive dramatic changes in functional profiles, productivity and ecosystem service provision of ecological communities5,6, with severe associated impacts. For example, the ‘trophic footprint’ (energy consumed) of 1 m2 of invertebrate-dominated engineered structures can negate the primary production of up to 130 m2 of surrounding coastal waters7. Globally, an estimated 1.0–3.4 million km2 of seascape has been modified by the installation of engineered coastal structures2.

Coastal hardening can also exacerbate the establishment and spread of marine non-indigenous species (NIS)4,8,9. Ports and marinas associated with coastal centres worldwide are vast artificial environments that represent important nodes in complex maritime transport networks, acting as sources and recipients of marine NIS10. These networks connect domestic and international coastal environments, including those associated with high ecological, economic, societal and cultural value11,12,13. Globally, the spread of marine NIS has contributed to dramatic, often irreversible ecological changes to coastal ecosystems14.

Coastal ecosystems are also heavily impacted by population growth. Substantial levels of future growth are expected and will be accompanied by additional coastal hardening, particularly in areas that require protection from the impacts of climate change15,16,17. These congested coastal spaces will increasingly require spatial planning and management to ensure a balance between ecological, economic and sociocultural objectives18. The ability to forecast the likely expansion of coastal infrastructure at the scale of regional jurisdictions or other spatial units would enable planners and decision-makers to anticipate potential impacts relating to habitat loss, changes in species pools, ecosystem services and social and cultural values, and develop proactive mitigation strategies19,20. A tool for this does not presently exist.

Here, we mapped coastal hardening associated with 30 international urban centres and, using machine-learning algorithms, developed a model to forecast the regional expansion of 4 globally common coastal infrastructure types. We applied this model to New Zealand as a case study location and show the model’s utility for anticipating regional distributions and future hotspots of socioecological risks over a 25-year period.

Results

We quantified the present-day extent (linear km) of coastal hardening associated with the regional coastlines of 30 global urban centres (Fig. 1). Coastal infrastructure accounted for approximately half (52.9 ± 4.9%; mean ± standard error) of the total regional coastline (up to 860 km per centre) around the locations examined (Fig. 2). Breakwalls were the most abundant type of coastal infrastructure (61 ± 3% of total regional infrastructure), followed by pontoons (14.5 ± 2.4%), wharves (14.1 ± 1.5%) and jetties (10.7 ± 1.3%). Overall, the maximum extents of breakwalls, pontoons, wharves and jetties found at any one study location were 254 km, 66 km, 74.5 km and 138 km, respectively (Fig. 2). Owing to their construction away from the shore, jetties and boating pontoons add ‘new’ rather than replace existing coastline (Supplementary Fig. 1). Across the locations mapped, these structures had added 13 ± 1.9% to the original (natural) coastline.

Fig. 1: Costal hardening extent determined for 30 global urban centres.
figure1

ae, Insets provide details on centres examined around North America (a, b), the United Kingdom (c), Australia (d) and New Zealand (e). Maps are from ArcGIS Living Atlas of the World.

Fig. 2: Present-day extent of coastal infrastructure associated with 30 international urban centres.
figure2

Bars represent the coastline distance (km) occupied by each structure type; per cent values indicate the artificial (hardened) proportion of the total coastline associated with the urbanized harbour.

Predictors of coastal urban infrastructure

We used machine-learning algorithms to examine the utility of 25 candidate variables associated with regional economy, population, trade and geomorphology as predictors of present-day coastal hardening (Table 1). The random forest ‘feature selection’ algorithm identified nine important or tentative variables, and these were used to develop a predictive model. Eleven variables were identified as unimportant, including four variables relevant to the effects of climate-induced increases in storm surges and sea-level rise (tidal range, per cent sand content of shoreline, per cent mud content of shoreline and mean significant wave height). Five variables were highly colinear (that is, variance inflation factor >3) and were excluded from subsequent analyses (Table 1).

Table 1 Candidate predictor variables for coastal hardening and stages of analysis they were used in: exploratory analyses based on New Zealand case study data; development of global predictive model; and final model and forecasting

Random forest models fitted to the global dataset, comprising the 9 important or tentative predictors, returned 8 final variables that together best explained the present-day extent of coastal hardening across all 30 locations. For the model that incorporated all infrastructure types, these were, in decreasing order of relative importance: port vessel visits, number of marinas, port revenue, regional gross domestic product (GDP), council population, total coastline length, port cargo and container throughputs (Table 1 and Supplementary Box 1). The relative importance of predictors varied slightly between models for specific infrastructure types (that is, breakwalls, pontoons, wharves and jetties) but generally shipping- and boating-related variables had the highest relative importance. Model training achieved good correspondence between observed and predicted values for all models (R2 of 0.86–0.92). Final model fit was good overall, with root mean squared error equating to approximately 10% of the data range of the response variables. The best-performing models were derived for total coastal infrastructure extent and extent of pontoons and wharves, accounting for up to 54% of the variation in the data (Supplementary Box 1).

Forecasting future coastal hardening with a New Zealand case study

Regional economic, population and trade growth scenarios (25-year horizon) were applied to the final random forest models (Tables 1 and 2; Supplementary Data 1) to forecast the extent of hardening across New Zealand’s urban centres. The models predicted a total increase of 243 km (low-growth scenario) to 368 km (high-growth scenario) in coastal infrastructure by 2043 for New Zealand’s coastal urban centres combined, representing a 49–76% increase relative to 2018 (Fig. 3). The greatest absolute increases were predicted for breakwalls (104 to 161 km under low- to high-growth scenarios, respectively), equating to a relative expansion (relative to 2018 extent) of 31–48%. The highest relative increases were predicted for shipping wharves (125–191%) and jetties (119–197%), although these involve considerably smaller absolute expansions (44–75 km; Fig. 3a).

Table 2 Summary of growth predictions used to forecast the potential for regional increases in coastal infrastructure around New Zealand
Fig. 3: Coastal hardening forecast for New Zealand.
figure3

a, Forecasted increase in the extent of coastal infrastructure associated with New Zealand’s urban centres by 2043. Bars represent predicted absolute increase according to low-, moderate- or high-growth scenarios; percentages indicate predicted increase relative to 2018. b, Regional distribution of the forecasted change in coastal hardening extent for the New Zealand case study locations by 2043 (medium-growth scenario). Bars represent the average mean contribution of each regional centre to the total predicted domestic increase.

The greatest proportion (67%) of the total domestic increase in coastal hardening was predicted for Auckland, Tauranga and Lyttelton harbours, accounting for 14–26% of expansions in breakwalls, 19–42% in jetties, 20–26% in wharves and 11–35% in pontoons (Fig. 3b). Increases in the extent of coastal hardening by 2043 were also forecast for most other urban centres. By 2043, 52–61% (low- to high-growth scenarios) of the regional coastline associated with the New Zealand case study locations will be artificial (hardened), compared with 35% in 2018.

Coastal hardening and NIS risks

Since 2001, between 18 and 66 marine NIS have been detected in coastal waters around New Zealand’s urban centres by government-led nationwide baseline survey and targeted surveillance programmes. We found a significant relationship (R2 = 0.57, P < 0.01) between the 2018 extent of coastal hardening around coastal urban centres and the number of marine NIS detected in each (Fig. 4). Application of the regression model to the 25-year coastal hardening forecast for regions around New Zealand predicted greatest increases in NIS occurrence at Tauranga, Auckland, Lyttelton, Napier and Picton harbours. This outcome is almost identical to (1) our global model’s ranked regional distribution of total forecasted increases in coastal hardening and (2) the ranked domestic biosecurity risk scores derived for New Zealand’s international arrival ports via state-of-the-art risk models (Supplementary Table 1). There was also good correspondence between our model’s predictions and the domestic biosecurity risk scores for the lower-ranking locations.

Fig. 4: Relationship between present-day extent of coastal hardening in New Zealand case study locations and the numbers of NIS occurring around these locations.
figure4

Black circles depict case study locations; regression line (blue line) and 95% confidence intervals (shading) are overlaid.

Discussion

Our study shows that coastal hardening is consistently extensive around urbanized harbours and estuaries in global locations including North America, the United Kingdom, Australia and New Zealand. On average, over half (52%, or 2,135 linear km) of the total coastline associated with each of the 30 mapped locations is already artificial. We also show that eight variables associated with commercial shipping, recreational boating, regional economic status, population size and regional coastline length can robustly predict the extent of coastal hardening associated with urbanized harbours and estuaries. Variables related to shipping and boating had the highest relative importance.

The construction of novel habitats (jetties and pontoons) has increased the original coastline length by an average of 13%. It was recently estimated that construction of shipping ports around the world (of which there are at least 6,500) has replaced 4,500 km2 of coastal habitat2,21. In our study, port-related infrastructure represented on average only 14% of the overall extent of infrastructure associated with each urban centre, while the remainder consisted of civic structures and recreational boating facilities. Together, our study and others2,7 have demonstrated the massive extent of natural habitat loss from coastal hardening around global coastlines.

Forecasts developed for a case study location (New Zealand) predict a considerable expansion in the extent of coastal hardening around domestic urban centres over the coming 25 years. Predicted rates of increase appear realistic when compared with historical expansions of coastal hardening recorded in New Zealand over the past 50–80 years (84–127%; Supplementary Fig. 2) and are consistent with recent strategic assessments by government and industry regarding expected domestic requirements for civic and transport infrastructure22,23,24. Interestingly, variables relevant to the effects of storm surges and sea-level rise were not recognized as important predictors of the present-day extent of coastal hardening. However, climate change and the increasing threat of sea-level rise might drive changes in management focus in the future25,26. For locations where this becomes the case, forecasts derived using our model might be conservative, in particular for breakwalls (structures frequently built to stabilize erosion-prone coastlines). Our current model assumes a business-as-usual scenario but understanding and incorporating potential shifts in socioeconomic priorities is an important future refinement as information becomes available.

The socioecological implications of further increases in coastal hardening are considerable. It will result in the loss of extensive remaining natural habitat (beaches, rocky reefs, mangroves, mudflats and others), with important consequences for shallow-water and coastal ecological communities, trophic processes, nutrient cycling and cultural/spiritual values6,13,27,28. For example, the capacity of ecosystems to store and degrade nutrients may be reduced through loss of sedimentary habitat5, while overall productivity might be enhanced due to the increase of hard substrates and the development of sessile biofouling communities on these. On average, each square metre of global coastal hardening infrastructure features 16.7 kg of sessile biomass that consumes the energy equivalent of 26 m2 of ocean surface primary production per day7. Further increases in the cumulative ‘trophic footprint’ of coastal hardening will deplete important resources for coastal marine ecosystems already under pressure from other processes such as pollution and harvesting2,29.

Further substantial increases in coastal hardening will also facilitate the establishment and spread of marine NIS4,9, exacerbating impacts on native ecosystem processes, the operations and productivity of aquaculture and other maritime industries and the integrity of cultural values30,31,32. Our case study models’ ranked predictions of (1) the domestic distribution of coastal hardening increases for the coming 25 years (‘hardening hotspots’) and (2) the associated increases in the occurrence of marine NIS both correspond well to independent modelling of domestic marine biosecurity risk associated with New Zealand’s main international arrival ports33. This highlights the utility of coastal hardening forecasts for identifying likely future ‘hotspots’ of NIS establishment, ideally in association with maritime trade models34.

Interpretation of forecasts

Our models indicate that regional rates of coastal hardening can be forecast based on present-day values and expected growth rates of only eight variables (port vessel visits, number of marinas, port revenue, regional GDP, council population, port cargo and container throughputs, and total coastline length). Collation of these data should be very achievable for most planning and management agencies around the world. Forecasts of regional rates of coastal hardening are best interpreted as potential for infrastructure expansions of a predicted magnitude. Depending on geological, industrial, political and societal factors this potential for growth may be realized in different ways: via (1) construction of new infrastructure, (2) urban expansion away from the coast, (3) increasing the efficiency of existing infrastructure or (4) outsourcing particular activities or services to other regions23,35. This interpretation allows for the consideration of location-specific circumstances when translating regional forecasts into actual expected growth for planning and mitigation purposes. For example, the current (2018) proportion of artificial coastline for the 30 urban centres examined varied from 5.4% to 83%. Those at the upper end of this range may be more likely to rely on strategies (2), (3) or (4) (above), while strategy (1) may be feasible for centres with larger proportions of remaining undeveloped coastline.

Minimizing the impacts of continued marine urban sprawl

Marine spatial planning (MSP) aims to manage ocean spaces by balancing multiple stakeholders and objectives. MSP is a growing field given the increasing global importance of the Blue Economy36 and the multiple pressures threatening marine systems37. Coastal development has historically occurred primarily in response to socioeconomic objectives, with less emphasis on ecological or cultural considerations38. MSP together with innovations in multifunctional design and biosecurity has the potential to manage future spread of built structures in the oceans to address different stakeholder needs18,39,40 and provide options for reconciliation of socioeconomic and ecological priorities38. The model developed here provides an important tool for MSP decision-making by enabling forecasts of regional rates of coastal hardening. This information, combined with innovations in ecological engineering, biofouling control and species monitoring tools, could support improved spatial risk assessment for conservation and biosecurity. Three important available avenues are discussed below.

Designing multifunctional coastal infrastructure

Multipurpose offshore platforms are becoming more widely implemented, with energy and aquaculture combining to co-locate systems and maximize benefit from a smaller ocean footprint41. A recent assessment38 proposes seven functions (or goals) that could be incorporated into the design of coastal structures in addition to traditional engineering functions: (1) maintenance of local native biota; (2) restoration of local biodiversity; (3) maintenance of regional biodiversity; (4) provision of educational and recreational opportunities; (5) maintenance of water quality; (6) facilitation of carbon storage; and (7) support of aquaculture and food production. We propose an additional, eighth function: the enhancement of cultural or spiritual values, where the design of infrastructure aims to restore or increase populations of species valued for traditional gathering, or to compensate for the impacts on other cultural values13. While the incorporation of multifunctionality into infrastructure development may not be able to mitigate trophic footprints or direct construction impacts, there is the potential to reduce the spatial extent of development if some functions are combined and offset impacts via the provision of ecological, societal, cultural or economic services18,42.

Minimizing NIS populations on coastal infrastructure

Submerged coastal infrastructure, in particular structures associated with shipping and boating facilities, are attractive substrates for NIS and often function as hotspots for invaders4,8,10. Our model highlighted the potential for substantial future increases in the spatial extent of these structures (pontoons, wharves and jetties) around our case study area, highlighting a need for innovation in intervention measures. Novel strategies to minimize the establishment of NIS on coastal infrastructure include environmentally friendly antifouling solutions such as surface treatments, air-bubble based protection technology or augmentative biocontrol43,44,45,46. Reducing resident populations of problem organisms on key infrastructure will limit propagule production and colonization rates of locally moored vessels and, thereby, the frequency with which NIS or other problematic species are spread to further destinations.

Development of improved monitoring tools

Improved surveillance and monitoring tools can enhance our ability to detect adventive populations of marine NIS or other harmful (for example, pathogenic) species. Revolutionary advances continue to be made in the development, operationalization and cost-effectiveness of molecular tools for environmental monitoring and assessment47,48,49. Wide and effective implementation of such tools and prospective automation of molecular-based detections can help mitigate some of the risks associated with ‘unprotected’ infrastructure. Integration of technologies such as those described above with the model developed in this study could help prioritize locations where interventions are best directed.

Summary

Marine urban sprawl is expected to continue around the world2,28. The modelling framework described here facilitates forecasting of regional rates of increase in the world’s most common types of coastal infrastructure based on economic, population, transport and geomorphological predictors, and the identification of likely hotspots of future NIS incursions. The model can be adapted by global planners and decision-makers to manage increasingly congested ocean spaces more sustainably and to focus environmental solutions on prioritized structures and areas.

Methods

We quantified present-day coastal hardening in 30 global urban centres and developed a model to forecast regional increases in coastal hardening, using New Zealand’s urban harbours and estuaries as a case study. This model was developed based on observations from global coastal centres, using a five-step approach (Supplementary Fig. 3). First, we quantified the present-day extent of coastal hardening associated with 30 global coastal urban centres. Second, exploratory analyses for 14 of these locations were used to identify predictors of coastal hardening from a larger set of candidate variables. Third, data for the predictor variables identified in step 2 were collected for the remaining global locations. Fourth, a predictive model was constructed and validated using the overall dataset. Lastly, the model was applied to forecast coastal hardening for 14 New Zealand coastal centres by 2043 (a 25-year horizon), based on 3 expected growth scenarios for the economic, urban and demographic predictors identified in step 2.

Quantification of present extent of coastal infrastructure

We determined the spatial distribution and extent of coastal infrastructure associated with 30 global coastal centres. To develop a model with wide applicability, we ensured that the urban centres selected represented a large gradient in area, population and economic aspects. By necessity, our selection was also guided by the availability of reliable, high-quality data for predictor variables. Unfortunately, this precluded the incorporation of locations from coastal regions around Asia, Africa and South/Latin America, particularly those within developing nations. Our 30 global centres were located around Australia, New Zealand, the United Kingdom and North America. All centres were situated within protected harbours or estuaries, or protected by coastal islands, and comprised urban/civic, shipping and boating facilities (Fig. 1). The absence of model locations situated in some regions of the world mean that the final model is probably not applicable to all global regions. However, as data for those regions become available, our model can be readily updated to provide benefits for areas currently not included.

Coastlines associated with each of the targeted locations were digitized based on 0.5 m resolution or higher satellite and aerial imagery (Land Information New Zealand’s imagery basemaps, Esri’s World Imagery (2014–2018; DigitalGlobe, councils and Marine GeoLibrary)) using ArcGIS Pro (version 2.4.0). Where urban centres extended into coastal areas beyond the harbour’s or estuary’s entrance, digitization was completed to the jurisdictional boundary. For each location, we mapped the extent of (1) natural coastline and (2) four main types of artificial structure: breakwalls, shipping wharves, jetties and floating pontoons (Supplementary Table 2 and Supplementary Fig. 1). While pontoons are floating structures, they have a stationary footprint and 50-year service lives. Floating marinas are consented as permanent installations and hence included as coastal infrastructure in our study. Geodesic length of each mapped element was calculated using New Zealand Transverse Mercator projection and relevant Universal Transverse Mercator projection zones for international regions. For large structures such as wharves and piers, the outer perimeter was delineated, while smaller structures such as jetties were represented by single lengths. For marina pontoons, the total length included their central walkway and all associated finger wharves. Natural coastline was traced along estimated high tide lines. For the 14 New Zealand locations, the surface area of all natural and artificial coastline elements was also calculated using Land Information New Zealand’s bathymetry data and standardized equations (details in Supplementary Box 2). Correlation analysis determined a strong association between linear and surface area extent of coastal artificial structures around New Zealand harbours (Pearson’s r = 0.90; P < 0.05). Based on this correlation, digitization of global model locations quantified only the linear extent of coastal infrastructure.

Choice of case study area

We chose New Zealand as a case study area to identify potential predictors of regional increases in coastal hardening for several reasons. First, it is a maritime and highly urbanized50 country featuring numerous coastal centres of varying size located in or around protected harbours and estuaries. Second, New Zealand’s economy is heavily reliant on shipping, and it is a popular destination for international and domestic cruising yachts. As such, New Zealand’s coastal urban centres generally feature shipping and boating infrastructure, which are thought to be strong drivers of coastal development and the spread of non-indigenous marine species2,9. Finally, excellent datasets are available for New Zealand relating to a wide range of potential predictor variables for coastal hardening, via resources offered by regional or central government agencies, industry groups or science organisations.

Identification of candidate predictor variables

We identified a total of 25 economic, demographic and geomorphological/environmental variables as potential predictor variables of the present-day extent of coastal hardening (Table 1). These candidate variables were selected because (1) they had been identified as drivers of urban expansion in previous studies15,51, (2) they had been identified as key variables for the classification of New Zealand’s estuarine and coastal marine environments52 and (3) in the case of geomorphological variables, they had direct relevance to expected effects of climate change. Using the 14 New Zealand case study locations, exploratory analyses were then undertaken to identify a constrained subset of potentially important predictors before collecting these data for the remaining global urban coastal centres (Table 1).

Random forest models were the primary statistical technique used to (1) explain present-day extent and (2) predict future growth of coastal infrastructure (below). Random forests are a non-parametric, machine-learning algorithm method that uses classification and regression trees to classify data or make predictions53,54. They bootstrap data to construct multiple trees grown with a randomized subset of the predictors and often outperform traditional regression approaches55. Two approaches were used to identify the important variables amongst the candidate predictors. First, the relative importance of the 25 candidate variables was examined using random forest ‘feature selection’ (R package Boruta56) in relation to the present-day extent of (1) the total extent of coastal infrastructure and (2) the extent of each of the four different structure types (that is breakwall, wharf, jetty and pontoon) for the 14 case study locations. This approach utilizes an algorithm designed as a ‘wrapper’ around the random forest classification algorithm that iteratively removes variables that are proven by a statistical test (based on z scores) to be less relevant than random probes. Secondly, correlations between the remaining variables were then identified using variance inflation factors and sequential removal of predictors with a variance inflation factor <3 (ref. 57).

Construction and validation of a global predictive model

Data for the 9 variables identified as important predictors in the exploratory analysis (Table 1) were obtained and collated for all 30 global model locations (Supplementary Data 1). Random forest models were then fitted using the R software packages randomForest58 and Caret59 for the total linear extent (km) of coastal infrastructure and for each of the four infrastructure types. Models were validated using 10-fold repeated cross-validation. Final models were selected based on the root mean squared error, enabling optimal model parametrization regarding number and depth of trees. Final models included 8 predictor variables (Table 1) and were used to forecast the extent of coastal hardening around New Zealand by 2043, a time horizon of 25 years at the time of the analyses.

Forecasting future coastal hardening with a New Zealand case study

Using government (for example, Stats NZ, regional councils and central government agencies), industry (for example, port and shipping companies and representative industry bodies) and financial (for example, banks and economic institutions) publications and databases (Supplementary Data 1), the final models were used to forecast the change in coastal hardening (total extent and extent of the four infrastructure types) around New Zealand based on expected economic, demographic and trade-related information. Forecasting was undertaken for a 25-year time horizon (2043) using low-, moderate- and high-growth scenarios identified from the predictor sources. For many predictors (for example, population increases, economic growth and some shipping trade variables), expected annual growth rates have been published for the modelled time horizon. For some other variables (for example, growth in recreational vessel marinas) published information included past average annual growth rates or shorter-term future growth rates. In those instances, annual compound growth was calculated and extrapolated to the 25-year horizon (Supplementary Data 1). The variance around the forecasted coastal infrastructure extent was calculated using infinitesimal jack-knifing. This technique works by omitting each observation (leave-one-out) and rerunning the estimate with the remaining data. Variances were square-root transformed to provide a standard deviation for each data point (Supplementary Box 1).

The outlook by industry and government at the time of model development suggested small to considerable annual growth relating to ports (0.7–3.2%), marinas (0.25–0.65%) and economic productivity (2.9–6.9%) for all of New Zealand’s urban centres examined and with considerable ranges between low and high growth-scenario predictions. Expected annual population growth was usually positive (averaging 0.5% across regions, with a maximum of 1.9% for Auckland) but low- and moderate-growth scenarios predicted population declines of up to 0.1–0.9% p.a. (predicted annual growth rate) for some areas on the North Island’s east and the South Island’s west coasts (Table 2 and Supplementary Data 1). All the information utilized for model and forecast development was issued before the onset of the COVID-19 pandemic.

Relationships between coastal hardening and NIS occurrence

We used linear regression analysis to examine potential associations between the degree of coastal hardening (cumulative extent of all infrastructure types mapped) and the number of marine NIS occurring around coastal urban centres in our case study region. Validated datasets from government-funded national baseline and target surveillance programmes (2001 to ongoing) provided the number of non-indigenous marine species reported from 12 of the 14 New Zealand model locations in 2018 (ref. 60) (Supplementary Data 1). These numbers were regressed on the extent (linear km) of coastal hardening associated with each of the 12 locations.

We then applied the derived relationship between coastal hardening and NIS occurrence to the predicted regional increases in coastal hardening (25-year timeframe, see above). This provided estimates of the number of new/additional NIS that might be expected for each of the New Zealand locations given the predicted increases in coastal hardening. It was not our intention to use this relationship to make empirical predictions of the likely emergence of new NIS. Instead, we wanted to explore any correspondence of these results to recent quantitative risk modelling commissioned by the New Zealand government that examined the relative importance of New Zealand coastal centres as recipients of new NIS based on shipping patterns and associated transport of organisms in ballast water and via hull fouling33. We thus ranked our case study locations in accordance with: (1) their relative contribution to the total predicted increase in domestic coastal hardening (New Zealand-wide by 2043; Fig. 3), (2) the number of new NIS expected based on the regression equation and our coastal hardening forecasts and (3) each centre’s ‘biosecurity risk score’ (ranked) derived from the risk models by Hatami et al.33. We then examined the correspondence of regional rankings according to (1), (2) and (3) to assess how our predictions compare with a sophisticated biosecurity risk analysis based on biofouling and ballast water discharge models.

Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

All data collected and analysed during this study are included in the Supplementary Information and a spatial geodatabase available at https://github.com/ofloerl/Coastal-Hardening.

Code availability

No custom computer code or algorithm was used to generate results. The software environment R was used for all data exploration, statistical analyses and model development described in this manuscript using freely available R packages. The R workflow is at https://github.com/ofloerl/Coastal-Hardening.

References

  1. 1.

    Dugan, J., Airoldi, L., Chapman, G. & Walker, S. in Treatise on Estuarine and Coastal Science Vol. 8 (eds Wolanski, E. & McLusky, D.) 17–41 (2011).

  2. 2.

    Bugnot, A. B. et al. Current and projected global extent of marine built structures. Nat. Sustain. 4, 33–41 (2020).

    Article  Google Scholar 

  3. 3.

    Connell, S. D. Floating pontoons create novel habitats for subtidal epibiota. J. Exp. Mar. Biol. Ecol. 247, 183–194 (2000).

    CAS  Article  Google Scholar 

  4. 4.

    Glasby, T., Connell, S., Holloway, M. & Hewitt, C. Nonindigenous biota on artificial structures: could habitat creation facilitate biological invasions? Mar. Biol. 151, 887–895 (2007).

    Article  Google Scholar 

  5. 5.

    Heery, E. C. et al. Identifying the consequences of ocean sprawl for sedimentary habitats. J. Exp. Mar. Biol. Ecol. 492, 31–48 (2017).

    Article  Google Scholar 

  6. 6.

    Scherner, F. et al. Coastal urbanization leads to remarkable seaweed species loss and community shifts along the SW Atlantic. Mar. Pollut. Bull. 76, 106–115 (2013).

    CAS  Article  Google Scholar 

  7. 7.

    Malerba, M. E., White, C. R. & Marshall, D. J. The outsized trophic footprint of marine urbanization. Front. Ecol. Environ. 17, 400–406 (2019).

    Article  Google Scholar 

  8. 8.

    Dafforn, K. A., Glasby, T. M. & Johnston, E. L. Comparing the invasibility of experimental “reefs” with field observations of natural reefs and artificial structures. PLoS ONE 7, e38124 (2012).

    CAS  Article  Google Scholar 

  9. 9.

    Airoldi, L., Turon, X., Perkol-Finkel, S. & Rius, M. Corridors for aliens but not for natives: effects of marine urban sprawl at a regional scale. Divers. Distrib. 21, 755–768 (2015).

    Article  Google Scholar 

  10. 10.

    Hayes, K. R., Inglis, G. J. & Barry, S. C. The assessment and management of marine pest risks posed by shipping: the Australian and New Zealand experience. Front. Mar. Sci. https://doi.org/10.3389/fmars.2019.00489 (2019).

  11. 11.

    Floerl, O., Inglis, G., Dey, K. L. & Smith, A. The importance of transport hubs in stepping-stone invasions. J. Appl. Ecol. 46, 37–45 (2009).

    Article  Google Scholar 

  12. 12.

    Kaluza, P., Kolzsch, A., Gastner, M. T. & Blasius, B. The complex network of global cargo ship movements. J. R. Soc. Interface 7, 1093–1103 (2010).

    Article  Google Scholar 

  13. 13.

    Aguirre, D. et al. Loved to pieces: toward the sustainable management of the Waitematā Harbour and Hauraki Gulf. Reg. Stud. Mar. Sci. 8, 220–233 (2016).

    Article  Google Scholar 

  14. 14.

    Molnar, J. L., Gamboa, R. L., Revenga, C. & Spalding, M. D. Assessing the global threat of invasive species to marine biodiversity. Front. Ecol. Environ. 6, 485–492 (2008).

    Article  Google Scholar 

  15. 15.

    Seto, K. C., Güneralp, B. & Hutyra, L. R. Global forecasts of urban expansion to 2030 and direct impacts on biodiversity and carbon pools. Proc. Natl Acad. Sci. USA 109, 16083–16088 (2012).

    CAS  Article  Google Scholar 

  16. 16.

    Neumann, B., Vafeidis, A. T., Zimmermann, J. & Nicholls, R. J. Future coastal population growth and exposure to sea-level rise and coastal flooding—a global assessment. PLoS ONE 10, e0118571 (2015).

    Article  CAS  Google Scholar 

  17. 17.

    Kulp, S. A. & Strauss, B. H. New elevation data triple estimates of global vulnerability to sea-level rise and coastal flooding. Nat. Commun. 10, 4844 (2019).

    CAS  Article  Google Scholar 

  18. 18.

    Lombard, A. T. et al. Practical approaches and advances in spatial tools to achieve multi-objective marine spatial planning. Front. Mar. Sci. https://doi.org/10.3389/fmars.2019.00166 (2019).

  19. 19.

    Pelling, M. & Blackburn, S. Megacities and the Coast: Risk, Resilience and Transformation (Routledge, 2013).

  20. 20.

    Sutton-Grier, A. E., Wowk, K. & Bamford, H. Future of our coasts: the potential for natural and hybrid infrastructure to enhance the resilience of our coastal communities, economies and ecosystems. Environ. Sci. Policy 51, 137–148 (2015).

    Article  Google Scholar 

  21. 21.

    Keller, R., Drake, J., Drew, M. & Lodge, D. Linking environmental conditions and ship movements to estimate invasive species transport across the global shipping network. Divers. Distrib. 17, 93–102 (2011).

    Article  Google Scholar 

  22. 22.

    How Can We Meet Increasing Demand for Ports in the Upper North Island? A Report for the Upper North Island Strategic Alliance (PricewaterhouseCoopers, 2012).

  23. 23.

    Ernst & Young Port Future Study. A Report Prepared for Auckland Council (Auckland Council, 2016).

  24. 24.

    NZIER Bigger Ships—Past, Present and Future Implications for New Zealand Supply Chains (New Zealand Economic Research Institute, 2017).

  25. 25.

    Hino, M., Belanger, S. T., Field, C. B., Davies, A. R. & Mach, K. J. High-tide flooding disrupts local economic activity. Sci. Adv. 5, eaau2736 (2019).

    Article  Google Scholar 

  26. 26.

    United Nations Review of Maritime Transport 109 (United Nations Conference on Trade and Development, 2019).

  27. 27.

    Ferrario, F., Iveša, L., Jaklin, A., Perkol-Finkel, S. & Airoldi, L. The overlooked role of biotic factors in controlling the ecological performance of artificial marine habitats. J. Appl. Ecol. 53, 16–24 (2016).

    Article  Google Scholar 

  28. 28.

    Firth, L. et al. Ocean sprawl: challenges and opportunities for biodiversity management in a changing world. Oceanogr. Mar. Biol. 54, 189–262 (2016).

    Google Scholar 

  29. 29.

    Mayer-Pinto, M. et al. Functional and structural responses to marine urbanisation. Environ. Res. Lett. 13, 014009 (2018).

    Article  Google Scholar 

  30. 30.

    Bannister, J., Sievers, M., Bush, F. & Bloecher, N. Biofouling in marine aquaculture: a review of recent research and developments. Biofouling 35, 631–648 (2019).

    CAS  Article  Google Scholar 

  31. 31.

    Colautti, R. I., Bailey, S. A., van Overdijk, C. D. A., Amundsen, K. & MacIsaac, H. J. Characterised and projected costs of nonindigenous species in Canada. Biol. Invasions 8, 45–59 (2006).

    Article  Google Scholar 

  32. 32.

    Mazur, K., Bath, A., Curtotti, R. & Summerson, R. An Assessment of the Non-market Value of Reducing the Risk of Marine Pest Incursions in Australia’s Waters (Australian Bureau of Agricultural and Resource Economics and Sciences, 2018).

  33. 33.

    Hatami, R. et al. Improving New Zealand’s Marine Biosecurity Surveillance Programme Biosecurity New Zealand Technical Paper No. 2021/01 (Ministry for Primary Industries, 2021).

  34. 34.

    Sardain, A., Sardain, E. & Leung, B. Global forecasts of shipping traffic and biological invasions to 2050. Nat. Sustain. 2, 274–282 (2019).

    Article  Google Scholar 

  35. 35.

    Monios, J., Bergqvist, R. & Woxenius, J. Port-centric cities: the role of freight distribution in defining the port-city relationship. J. Transp. Geogr. 66, 53–64 (2018).

    Article  Google Scholar 

  36. 36.

    The Ocean Economy in 2030 (Organisation for Economic Co-operation and Development, 2016).

  37. 37.

    Halpern, B. S. et al. Recent pace of change in human impact on the world’s ocean. Sci. Rep. 9, 11609 (2019).

    Article  CAS  Google Scholar 

  38. 38.

    Dafforn, K. A. et al. Marine urbanization: an ecological framework for designing multifunctional artificial structures. Front. Ecol. Environ. 13, 82–90 (2015).

    Article  Google Scholar 

  39. 39.

    Diggon, S. et al. The marine plan partnership: Indigenous community-based marine spatial planning. Mar. Policy https://doi.org/10.1016/j.marpol.2019.04.014 (2019).

  40. 40.

    Noble, M. M., Harasti, D., Pittock, J. & Doran, B. Understanding the spatial diversity of social uses, dynamics, and conflicts in marine spatial planning. J. Environ. Manag. 246, 929–940 (2019).

    Article  Google Scholar 

  41. 41.

    Abhinav, K. A. et al. Offshore multi-purpose platforms for a blue growth: a technological, environmental and socio-economic review. Sci. Total Environ. 734, 138256 (2020).

    CAS  Article  Google Scholar 

  42. 42.

    Jacob, C., Buffard, A., Pioch, S. & Thorin, S. Marine ecosystem restoration and biodiversity offset. Ecol. Eng. 120, 585–594 (2018).

    Article  Google Scholar 

  43. 43.

    Hopkins, G. A. et al. Continuous bubble streams for controlling marine biofouling on static artificial structures. PeerJ 9, e11323 (2021).

    Article  Google Scholar 

  44. 44.

    Vucko, M. J. et al. Cold spray metal embedment: an innovative antifouling technology. Biofouling 28, 239–248 (2012).

    CAS  Article  Google Scholar 

  45. 45.

    Atalah, J., Newcombe, E. M., Hopkins, G. A. & Forrest, B. M. Potential biocontrol agents for biofouling on artificial structures. Biofouling 30, 999–1010 (2014).

    CAS  Article  Google Scholar 

  46. 46.

    Airoldi, L. et al. Emerging solutions to return nature to the urban ocean. Ann. Rev. Mar. Sci. 13, 445–477 (2021).

    Article  Google Scholar 

  47. 47.

    Keeley, N., Wood, S. A. & Pochon, X. Development and preliminary validation of a multi-trophic metabarcoding biotic index for monitoring benthic organic enrichment. Ecol. Indic. 85, 1044–1057 (2018).

    CAS  Article  Google Scholar 

  48. 48.

    Zaiko, A., Pochon, X., Garcia-Vazquez, E., Olenin, S. & Wood, S. A. Advantages and limitations of environmental DNA/RNA tools for marine biosecurity: management and surveillance of non-indigenous species. Front. Mar. Sci. https://doi.org/10.3389/fmars.2018.00322 (2018).

  49. 49.

    Cristescu, M. E. Can environmental RNA revolutionize biodiversity science? Trends Ecol. Evol. 34, 694–697 (2019).

    Article  Google Scholar 

  50. 50.

    Chakravarthy, K., Charters, F. & Cochrane, T. The impact of urbanisation on New Zealand freshwater quality. Policy Q. 15, 17–21 (2019).

    Article  Google Scholar 

  51. 51.

    Gittman, R. K. et al. Engineering away our natural defenses: an analysis of shoreline hardening in the US. Front. Ecol. Environ. 13, 301–307 (2015).

    Article  Google Scholar 

  52. 52.

    Hume, T. M., Snelder, T., Weatherhead, M. & Liefting, R. A controlling factor approach to estuary classification. Ocean Coast. Manag. 50, 905–929 (2007).

    Article  Google Scholar 

  53. 53.

    Breiman, L. Random forests. Mach. Learn. 45, 5–32 (2001).

    Article  Google Scholar 

  54. 54.

    Prasad, A. M., Iverson, L. R. & Liaw, A. Newer classification and regression tree techniques: bagging and random forests for ecological prediction. Ecosystems 9, 181–199 (2006).

    Article  Google Scholar 

  55. 55.

    Olden, J. D., Lawler, J. J. & Poff, N. L. Machine learning methods without tears: a primer for ecologists. Q. Rev. Biol. 83, 171–193 (2008).

    Article  Google Scholar 

  56. 56.

    Kursa, M. B. & Rudnicki, W. R. Feature selection with the boruta package. J. Stat. Softw. 36, 1–13 (2010).

    Article  Google Scholar 

  57. 57.

    Zuur, A. F., Leno, E. N. & Elphick, C. S. A protocol for data exploration to avoid common statistical problems. Methods Ecol. Evol. 1, 3–14 (2010).

    Article  Google Scholar 

  58. 58.

    Liaw, A. & Wiener, M. Classification and regression by randomForest. R News 2, 18–22 (2002).

    Google Scholar 

  59. 59.

    Kuhn, M. et al. caret: Classification and Regression Training (CRAN, 2019); https://CRAN.R-project.org/package=caret

  60. 60.

    Ministry for the Environment & Stats NZ. New Zealand’s Environmental Reporting Series: Environment Aotearoa 2019 (Ministry for the Environment, 2019).

Download references

Acknowledgements

This research was funded by New Zealand Ministry for Business, Innovation and Employment research grants C01X1511 (O.F.) and CAWX1904 (O.F.) and internal resources provided by the authors’ institutions. We thank D. Lohrer (National Institute of Water and Atmospheric Research) for access to the Coastal Explorer estuary classification database, A. Growcott (New Zealand Ministry for Primary Industries) for access to government biosecurity risk modelling and P. South (Cawthron Institute) for valuable feedback on an earlier version of the manuscript.

Author information

Affiliations

Authors

Contributions

O.F. led the development of the original concept of this manuscript, with assistance from all co-authors. M.C. and L.F. led the development and implementation of spatial mapping and analysis. J.A. and R.M. led the development of analytical pipelines. O.F. and M.C. led the data-gathering effort. O.F., R.M., A.B.B., K.A.D. and A.Z. led the preparation of the manuscript.

Corresponding author

Correspondence to Oliver Floerl.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Peer review information Nature Sustainability thanks the anonymous reviewers for their contribution to the peer review of this work.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Supplementary methods and model diagnostics, Figs. 1–3, Boxes 1 and 2 and Table 1.

Reporting Summary

Supplementary Data 1

Model predictor variables and case study growth predictions input data.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Floerl, O., Atalah, J., Bugnot, A.B. et al. A global model to forecast coastal hardening and mitigate associated socioecological risks. Nat Sustain (2021). https://doi.org/10.1038/s41893-021-00780-w

Download citation

Further reading

Search

Quick links