539

1 INTRODUCTION

Seasonal reduction in Arctic sea ice has increased

interest in shorter Europe– Asia shipping corridors

through the Arctic. However, shorter geometric

distance alone does not imply navigability. Candidate

corridors must first satisfy physical constraints such as

land exclusion, sufficient water depth, and acceptable

sea-ice concentration. These constraints are especially

important in early-stage route screening, where the

aim is to determine whether a continuous feasible

corridor exists before applying vessel-specific

performance models, fuel-consumption calculations,

or operational risk assessment.

This paper addresses that first screening step. It

presents a reproducible graph-based workflow that

converts bathymetry and sea-ice concentration data

into feasibility masks and then evaluates route

continuity and route distance for a Rotterdam–

Yokohama trans-Arctic case study. The workflow is

intended as a navigation-feasibility layer: it identifies

candidate corridors and sensitivity to screening

assumptions, while leaving detailed operational

routing, ice-risk modelling, and vessel-performance

assessment to later analysis.

The methodological contribution is the transparent

integration of bathymetry and SIC constraints into a

reproducible graph-based feasibility workflow. The

paper does not claim that A-star search is new. Instead,

Bathymetry- and Sea-Ice-Constrained Feasibility

Screening for Trans-Arctic Navigation

A. Mohamed & X. Hu

Technical University of Munich, Garching, Germany

ABSTRACT: Arctic route planning requires an initial assessment of whether candidate corridors are physically

navigable before detailed voyage optimisation, vessel-performance modelling, or operational risk assessment is

attempted. This paper presents a reproducible feasibility-screening workflow for trans-Arctic navigation under

bathymetry and sea-ice concentration constraints. A regular pan-Arctic routing grid is constructed over 40–85◦N

and 20◦W–180◦E for a Rotterdam–Yokohama case study. GEBCO 2024 bathymetry is used to exclude land and

shallow-water cells, while CMEMS Arctic sea-ice concentration fields are used to exclude cells above selected ice-

screening thresholds. The remaining feasible cells are represented as a graph, and A-star search with geodesic

edge costs is used to compute shortest feasible routes. The contribution lies in the transparent integration of

bathymetry and sea-ice screening into a reproducible navigation-feasibility workflow, rather than in proposing a

new shortest-path algorithm. The case study evaluates a bathymetry-only baseline, date-dependent sea-ice

feasibility maps, seasonal route-distance behaviour, and sensitivity to sea-ice concentration thresholds. The

results show how route availability and route-distance ratio vary with the selected environmental screening

assumptions. The workflow provides a practical pre-routing layer for Arctic navigation studies and can support

later integration with vessel-performance, risk, and voyage-planning models.

http://www.transnav.eu

the International Journal

on Marine Navigation

and Safety of Sea Transportation

Volume 20

Number 3

September 2026

DOI: 10.12716/1001.20.03.02

540

it shows how standard graph-search methods can be

embedded in a clearly specified environmental-

screening procedure for Arctic navigation studies. This

is useful because many Arctic routing studies focus

directly on optimisation, weather routing, fuel

consumption, or operational decision support,

whereas a route must first be topologically feasible

under basic environmental constraints.

This paper addresses four practical research

questions:

− RQ1. Does a continuous trans-Arctic route between

Rotterdam and Yokohama exist on the selected 0.5◦

pan-Arctic grid when bathymetry constraints are

applied?

− RQ2. How does adding a SIC constraint change

route continuity and corridor selection relative to

the bathymetry-only baseline?

− RQ3. How do selected SIC thresholds and late-

summer dates affect route distance ratio and route

availability in the Rotterdam–Yokohama Arctic

case study?

− RQ4. What assumptions and limitations follow

from using this workflow as a strategic route-

feasibility layer rather than as complete operational

navigation guidance?

The paper is structured as follows. Section 2

positions the study relative to existing Arctic route-

planning work. Section 3 describes the routing domain,

bathymetry data, sea-ice data, and grid construction.

Section 4 defines the feasibility masks, graph

construction, edge costs, A-star search, and sensitivity

setup. Section 5 presents the route-feasibility and

sensitivity outputs. Section 6 discusses interpretation,

limitations, and use in future navigation and digital-

twin studies. Section 7 summarises the findings.

2 RELATED WORK AND STUDY POSITIONING

Arctic route planning has been studied from several

perspectives, including route feasibility, ice-aware

routing, vessel-performance modelling, weather

routing, risk assessment, and decision support. Recent

review work emphasises that Arctic weather routing

depends on both ship-performance models and ice-

routing algorithms, and that the field still requires

transparent modelling choices and clearer treatment of

ice conditions (Liu et al., 2023). Route-View

demonstrates how big Earth data can be used in an

interactive Arctic route-planning system (Wu et al.,

2022). Other studies have addressed Arctic route

design using multi-objective formulations that

combine safety, economic, and environmental criteria

(Chen et al., 2023). Work on icebreaker-assisted routing

also shows that operational Arctic routing can require

additional constraints beyond simple distance

minimisation (Topaj et al., 2019). Economic and

operational studies of the Northern Sea Route further

show that distance reduction alone is insufficient for

assessing practical route attractiveness (Theocharis et

al., 2019).

2.1 Arctic route planning under environmental

constraints

Sea-ice concentration is commonly used as a first

indicator of ice accessibility. In operational navigation,

however, SIC alone is insufficient because vessel

capability, ice thickness, ridging, drift, visibility,

weather, icebreaker support, and regulatory

constraints also matter. The IMO Polar Code and the

POLARIS guidance link polar navigation to vessel

capability and ice-risk assessment rather than to sea-ice

concentration alone (IMO, 2016; PAME, 2024).

Bathymetry is similarly important but incomplete: a

minimum-depth screen is not a full under-keel-

clearance assessment and does not replace nautical

charts. In this paper, bathymetry and SIC are therefore

used as strategic screening constraints, not as complete

safety criteria.

2.2 Graph-based routing and shortest-path methods

Graph-based routing represents a navigable area as

nodes and edges. Once infeasible cells are removed,

shortest feasible paths can be computed by standard

graph-search algorithms. A-star search is appropriate

when a lower-bound heuristic is available. Here, the

heuristic is the geodesic distance from a node to the

destination. Since any feasible path through the graph

cannot be shorter than the direct geodesic distance to

the destination, the heuristic does not overestimate the

remaining distance. A-star is therefore used as an

efficient shortest-path implementation rather than as a

claimed methodological novelty.

2.3 Study positioning

Compared with full weather-routing, vessel-

performance, or ice-risk models, the present study

deliberately focuses on the pre-routing feasibility layer.

Weather-routing models use wind, waves, currents,

and sometimes ice to generate time-, fuel-, or safety-

oriented routes. Ice-routing models may use SIC, ice

thickness, ice charts, and vessel ice capability to

generate ice-aware or risk-aware routes. Vessel-

performance routing uses ship particulars, propulsion,

ice resistance, and metocean state to estimate speed

loss, fuel consumption, emissions, or estimated time of

arrival. Operational decision-support systems use

forecasts, traffic, operational rules, and vessel

constraints to support route advice.

The present workflow does not replace these

approaches. Its purpose is earlier in the analysis chain:

it tests whether a continuous candidate corridor exists

under specified bathymetry and SIC assumptions. Its

value is that the screening choices are explicit: routing

domain, grid resolution, bathymetry threshold, SIC

threshold, graph topology, edge costs, route metrics,

and sensitivity design.

3 DATA AND ROUTING DOMAIN

3.1 Routing domain and endpoints

The routing domain covers 40–85◦N and 20◦W–180◦E.

This domain captures the northern Europe–Asia

routing region and the main Arctic corridor space

541

considered in the study. The route endpoints are

Rotterdam and Yokohama. Rotterdam is represented

by an approximate port coordinate near 51.95◦N,

4.14◦E, and Yokohama by an approximate port

coordinate near 35.45◦N, 139.65◦E. Each endpoint is

mapped to the nearest feasible routing-grid node. The

endpoint pair is held constant across the bathymetry-

only and ice-constrained scenarios so that route

differences reflect the environmental constraints rather

than different origin–destination definitions.

3.2 Bathymetry data

Bathymetry is taken from the GEBCO 2024 global grid

(GEBCO Bathymetric Compilation Group, 2024). The

bathymetry layer is used to classify cells as feasible or

infeasible according to a minimum-depth screen. The

reference bathymetric feasibility case uses a 50 m

minimum-depth threshold. This threshold is a strategic

screening value and not a navigation-grade under-

keel-clearance calculation.

3.3 Sea-ice concentration data

SIC fields are taken from the CMEMS Arctic Ocean Sea

Ice Reanalysis (Copernicus Marine Environment

Monitoring Service, 2023). For a selected threshold

SICmax, cells with SIC above the threshold are

excluded from the graph. The sensitivity figures

evaluate SIC thresholds of 5%, 10%, 15%, 30%, and

50%. The date-threshold analysis uses selected

hindcast dates between 15 June and 30 September 2018.

These cases are interpreted as feasibility screening

examples, not as climatological trend estimates.

3.4 Grid construction

The routing graph is constructed on a regular 0.5◦

latitude–longitude grid. Each grid-cell centre is treated

as a candidate node. SIC and bathymetry values are

assigned to the routing grid according to the gridded

data extraction procedure used in the routing scripts.

A node is retained only if it satisfies the active sea,

bathymetry, and SIC constraints. Edges connect

neighbouring retained nodes, and the cost of each edge

is the geodesic distance between the two node centres.

The same grid definition is used for the bathymetry-

only baseline and the ice-constrained cases.

4 METHODOLOGY

4.1 Workflow overview

The workflow converts gridded environmental data

into a constrained routing graph and then computes

route existence and shortest feasible distance. The

process has eight steps:

1. Load GEBCO 2024 bathymetry and CMEMS SIC

fields.

2. Define the routing domain over 40–85◦N and 20◦W–

180◦E.

3. Construct the regular 0.5◦ routing grid.

4. Apply the land–sea and 50 m bathymetry screens.

5. Apply the selected SIC threshold or date-threshold

feasibility mask.

6. Build the feasible routing graph from retained grid

nodes.

7. Run A-star path search with geodesic edge costs.

8. Report route existence, route distance, corridor

geometry, and sensitivity outputs.

The workflow is designed to make each screening

step traceable from input data to route output.

4.2 Feasibility masks

A grid cell is treated as feasible if it satisfies all active

constraints. The bathymetry constraint is

mini

dd

(1)

where di is the water depth at cell i, and dmin = 50 m in

the reference bathymetry screen. The SIC constraint is

( ) maxSICi t SIC

(2)

where SICi(t) is the SIC at cell i on date t, and SICmax

is the selected ice-screening threshold. A node is

included in the ice-constrained graph only if both

constraints are satisfied.

4.3 Graph construction and edge costs

Let G = (V,E) denote the feasible routing graph. The

node set V contains retained grid-cell centres, and the

edge set E connects neighbouring feasible nodes. An

edge is created only when both endpoint nodes are

feasible. The graph uses neighbouring grid cells as

candidate connections; diagonal connections are

included where both adjacent nodes are feasible. The

edge cost is the geodesic distance between node

centres:

( )

i j geo

c dist i j=

(3)

The route distance is the sum of all edge costs along

the selected path:

( )

()

i j P

D P c





(4)

where P is the ordered set of route edges. Route

distances are reported in nautical miles.

4.4 A-star route search and validity argument

A-star search is used to find the shortest feasible route

on the constrained graph. The evaluation function is

( ) ( ) ( )f n g n h n=+

(5)

where g(n) is accumulated distance from the origin to

node n, and h(n) is the geodesic distance from node n

to the destination. Because direct geodesic distance is a

lower bound on the distance of any feasible graph path

to the destination, this heuristic is admissible for

distance-minimising search. Under non-negative

geodesic edge costs, A-star is equivalent to Dijkstra

search with an admissible heuristic for the purpose of

distance-optimal routing on the constructed graph.

542

4.5 Scenario and sensitivity design

The scenario design contains five groups. First, the

bathymetry-only baseline tests whether a continuous

sea route exists when land and shallow-water cells are

excluded. Second, the seasonal SIC analysis tests how

the ice-safe sea area changes across selected 2018 dates.

Third, the route-distance analysis evaluates how the

feasible route distance changes under selected late-

summer SIC fields. Fourth, the threshold-sensitivity

analysis varies the SIC upper bound across 5%, 10%,

15%, 30%, and 50%. Fifth, the date-threshold heatmap

combines selected dates and SIC thresholds to show

how route-distance ratio responds to both

assumptions.

The sensitivity analysis is parameter-based rather

than probabilistic. It varies SIC thresholds and selected

late-summer dates to test whether route availability

and route-distance ratio are robust to screening

choices. It does not quantify uncertainty in the

underlying SIC reanalysis, bathymetry data, or vessel

response.

4.6 Route metrics

The workflow reports route existence, route distance,

route-distance ratio, and route position. Route

existence indicates whether a continuous feasible path

connects Rotterdam and Yokohama. Route distance is

the total geodesic path length along the graph. The

relative distance change is computed as

()

100%

rel

D P D

−

 = 

(6)

where D(P) is the feasible route distance and DGC is the

direct great-circle distance between the endpoints.

5 RESULTS

The results show three main effects. First, the

bathymetry-only route provides a continuous sea-

route baseline between Rotterdam and Yokohama on

the selected 0.5◦ grid. Second, adding SIC screening

changes route availability and route-distance ratio

depending on date and threshold. Third, the date-

threshold heatmap shows that late-summer route

feasibility is not controlled by a single SIC threshold

alone; it also depends on the timing of the selected

hindcast field. In the feasible late-summer cases, the

ice-constrained route remains approximately 1.21–1.25

times the great-circle distance, indicating that

environmental feasibility imposes a non-negligible

detour even when a continuous Arctic corridor exists.

These outputs are interpreted as sensitivity-based

feasibility screens rather than deterministic operational

routes.

5.1 Bathymetry domain and sea-only feasibility

The bathymetry domain mask is shown in Figure 1.

This figure defines the land–sea and bathymetric

structure used to construct the feasible routing graph.

The bathymetry-only case removes land and shallow-

water cells while ignoring SIC. This scenario provides

a sea-route baseline and identifies where shallow

shelves affect route geometry.

Figure 1. Bathymetry domain mask used for the pan-Arctic

feasibility-screening workflow. The left panel shows GEBCO

2024 bathymetry on the routing grid, and the right panel

shows the land–sea mask and connected sea components

used to define the graph domain.

Figure 2. Bathymetry-only sea route between Rotterdam and

Yokohama on the reference 0.5◦ grid. This route is used as the

physical sea-route baseline before applying SIC constraints.

The bathymetry-only route remains continuous

across the selected domain. It is longer than the direct

great-circle reference because the graph must follow

feasible sea cells and avoid land and shallow-water

regions. This baseline is important because later ice-

constrained routes should be interpreted relative to an

already constrained sea route, not relative only to an

idealised great-circle line.

5.2 Bathymetric connectivity

Figure 3 shows how the bathymetric screen affects

domain connectivity. The figure indicates that

increasing the minimum-depth threshold reduces the

fraction of sea cells classified as depth-safe and changes

the number of connected components. This supports

the use of bathymetry as a first feasibility screen: even

before sea ice is considered, the selected depth

threshold influences which corridors remain connected

on the routing graph.

Figure 3. Connectivity response of the bathymetry-screened

routing graph. The figure shows how the fraction of depth-

safe sea cells and the number of depth-safe components

change with the selected minimum-depth threshold.

543

5.3 Seasonal sea-ice feasibility

The seasonal ice-area output is shown in Figure 4. For

the selected SIC threshold, the ice-safe sea area

increases from early summer towards late summer,

while the number of ice-safe components decreases.

This indicates that the ice-safe domain becomes more

connected as the summer season progresses.

Figure 4. Seasonal variation in ice-safe sea area and ice-safe

graph components for the selected routing domain and SIC

threshold. The result shows how the accessible sea area and

graph connectivity evolve during the 2018 summer season.

The seasonal route-distance output is shown in

Figure 5. The figure shows that continuous ice-safe

routes are obtained only for selected late-summer dates

in the available output. The plotted route distances

remain longer than the great-circle reference, showing

that even when an ice-safe route exists, the

environmental masks impose a distance penalty

relative to an idealised direct route.

Figure 5. Seasonal route-distance response under the selected

SIC screening setup. The figure compares ice-safe route

distance, great-circle distance, and the route-to-great-circle

distance ratio for feasible late-summer cases.

5.4 Sensitivity to SIC threshold and date

The threshold-sensitivity output is shown in Figure 6.

This figure reports the response of ice-safe sea area and

graph connectivity to selected SIC thresholds on 15

September 2018. Lower thresholds represent more

conservative screening assumptions. Higher

thresholds allow more ice-affected cells to remain in

the graph and can change the number of connected ice-

safe components.

Figure 6. Sensitivity of ice-safe sea area and ice-safe graph

connectivity to selected SIC thresholds on 15 September 2018.

Lower thresholds represent more conservative ice-screening

assumptions.

Figure 7 shows the combined effect of selected date

and SIC threshold on the route-to-great-circle distance

ratio. The heatmap indicates that the distance ratio

varies with both the date of the SIC field and the

threshold used to define ice-safe cells. The strongest

variation appears across dates rather than across all

thresholds, which suggests that seasonal timing is a

dominant factor in this case-study output.

Figure 7. Date-threshold sensitivity output showing the

combined effect of selected date and SIC threshold on the

route-to-great-circle distance ratio. Blank cells indicate cases

where a continuous feasible route was not obtained in the

available output.

6 DISCUSSION

6.1 Interpretation of feasibility-screening results

A feasible route in this workflow means that a

continuous graph path exists through cells satisfying

the selected bathymetry and SIC thresholds. It does not

mean that the route is safe for all vessels, operationally

optimal, commercially attractive, or compliant with all

polar-navigation requirements. Conversely, an

infeasible route under a conservative threshold does

not imply that no vessel could navigate the region. It

means only that the selected screening assumptions do

not produce a continuous path on the selected grid.

The results also show why great-circle distance

alone is not a sufficient indicator of Arctic route

feasibility. Even the bathymetry-only route differs

from the direct geometric reference because land,

shallow water, and graph connectivity constrain the

feasible path. When SIC is added, route availability and

544

distance depend further on seasonal timing and

threshold choice.

6.2 Use in navigation and maritime-safety applications

For navigation and planning applications, the main

value of the workflow is early exclusion. A route

planner or researcher can first test whether a corridor

remains topologically connected under selected depth

and ice constraints. Only corridors that pass this

screening step need to be evaluated with more detailed

models for vessel speed loss, fuel consumption, ice

risk, weather exposure, regulatory restrictions, or

operational cost. In this sense, the workflow acts as a

pre-routing filter rather than a complete route-advice

system.

The workflow can also support comparative

studies. Because the same endpoint pair, grid, graph

construction, and route metric are held fixed, changes

in route distance can be attributed to changes in the

environmental screening assumptions. This makes the

approach useful for testing how sensitive Arctic route

feasibility is to selected bathymetry and SIC thresholds.

6.3 Limitations and assumptions

The main limitations are:

1. The SIC constraint uses concentration only. It does

not include ice thickness, ridging, floe size, ice drift,

compression, or ice-class-specific performance.

2. The 50 m bathymetry screen is a strategic threshold,

not a nautical-chart or under-keel-clearance

calculation.

3. The 0.5◦ grid is suitable for research screening but

not for operational navigation.

4. The ice analysis uses selected hindcast dates and

thresholds and should not be interpreted as a full

multi-year climatological route-availability

estimate.

5. The route is static at the routing-step level. It does

not yet sample evolving SIC along the route as a

function of vessel speed and arrival time.

6. Weather, waves, currents, traffic separation

schemes, icebreaker support, legal restrictions, fuel

consumption, emissions, and risk-cost tradeoffs are

not included in this feasibility layer.

7. A-star is justified by its admissible geodesic

heuristic, but the present implementation does not

include a separate empirical comparison against

Dijkstra search in the reported results.

6.4 Future work

Future work should extend the workflow in four

directions. First, time-dependent routing should

sample SIC fields according to the estimated arrival

time at each route segment. Second, uncertainty

analysis should vary SIC thresholds, bathymetric

thresholds, departure dates, and grid resolution

systematically. Third, the graph layer should be

coupled with vessel-performance, fuel-consumption,

emissions, and ice-risk models. Fourth, the feasibility-

screening layer should be integrated into a voyage-

planning digital-twin framework that separates

environmental feasibility, route choice, performance

prediction, and operational risk.

7 CONCLUSION

This paper presented a reproducible feasibility-

screening workflow for trans-Arctic navigation under

bathymetry and SIC constraints. The workflow uses

GEBCO 2024 bathymetry, CMEMS Arctic sea-ice

concentration, a 0.5◦ pan-Arctic grid, a 50 m

bathymetric screen, and A-star pathfinding with

geodesic edge costs. The case study considers a

Rotterdam–Yokohama route within 40–85◦N and

20◦W–180◦E, including a bathymetry-only baseline,

ice-feasibility outputs, seasonal route-distance outputs,

and sea-ice-threshold sensitivity figures.

The paper treats the workflow as a transparent

environmental-constraint layer rather than as a new

routing algorithm or operational navigation tool. This

positioning is useful for navigation studies because the

value of the workflow lies in explicit assumptions,

reproducible feasibility masks, route-continuity

screening, and sensitivity to ice-threshold choice. The

workflow can support future dynamic voyage

planning, maritime-safety analysis, and Arctic routing

digital-twin development once time dependence,

uncertainty, vessel performance, and risk models are

added.

ACKNOWLEDGEMENTS

This work was supported by the Technical University

of Munich.

DATA AND CODE AVAILABILITY

Bathymetry data are taken from the GEBCO 2024 global grid

(GEBCO Bathymetric Compilation Group, 2024). Sea-ice

concentration fields are taken from the CMEMS Arctic Ocean

Sea Ice Reanalysis (Copernicus Marine Environment

Monitoring Service, 2023). The scripts and configuration files

used to construct feasibility masks and perform route

searches are available from the corresponding author upon

reasonable request and will be deposited in an open

repository after manuscript acceptance.

APPENDIX A. SUPPLEMENTARY ROUTE MAPS

AND SENSITIVITY RESULTS

Figure A.8. Depth profile along the reference sea-only route

and distribution of depths in route cells compared with all

sea cells.

545

Figure A.9. Supplementary route-distance sensitivity to the

selected SIC threshold for the late-summer case. Solid line:

ice-constrained route; dashed line: bathymetry-only sea

route; dotted line: great-circle distance.

Figure A.10: Supplementary ice-feasibility route map for 1

September 2018 under the selected screening setup. The

figure illustrates the corridor obtained after classifying grid

cells by SIC threshold.

REFERENCES

Chen, A., Zhang, Y., Wang, J., and Liu, X. (2023). Arctic route

planning and navigation strategy. Journal of Marine

Science and Engineering, 11(7), 1308.

doi:10.3390/jmse11071308.

Copernicus Marine Environment Monitoring Service. (2023).

Arctic Ocean Sea Ice Reanalysis. Copernicus Marine

Service Marine Data Store. doi:10.48670/mds-00336.

GEBCO Bathymetric Compilation Group. (2024). The

GEBCO 2024 Grid: a continuous terrain model of the

global oceans and land. doi:10.5285/1c44ce99-0a0d-5f4f-

e063-7086abc0ea0f.

International Maritime Organization. (2016). Guidance on

methodologies for assessing operational capabilities and

limitations in ice. MSC.1/Circ.1519.

Liu, Q., Zhang, Y., Li, X., and Huang, L. (2023). Arctic

weather routing: a review of ship performance models

and ice routing algorithms. Frontiers in Marine Science,

10, 1190164. doi:10.3389/fmars.2023.1190164.

Protection of the Arctic Marine Environment. (2024). Polar

Operational Limit Assessment Risk Indexing System.

PAME Arctic Shipping Project documentation.

Schroder, C., Reimer, N., Jochmann, P., and Rinke, A. (2017).

Environmental impact of exhaust emissions by Arctic

shipping. Ambio, 46, 400–409. doi:10.1007/s13280-017-

0956-0.

Theocharis, D., Rodrigues, V. S., Pettit, S., and Haider, J.

(2019). Feasibility of the Northern Sea Route: the role of

distance, fuel prices, ice breaking fees and ship size for the

product tanker market. Transportation Research Part E:

Logistics and Transportation Review, 129, 111–135.

doi:10.1016/j.tre.2019.07.003.

Topaj, A. G., Tarovik, O. V., Bakharev, A. A., and

Kondratenko, A. A. (2019). Optimal ice routing of a ship

with icebreaker assistance. Applied Ocean Research, 86,

177–187. doi:10.1016/j.apor.2019.02.021.

Wu, A., Zhang, Y., Chen, J., and Li, X. (2022). RouteView: an

intelligent route planning system for ships sailing in the

Arctic. International Journal of Digital Earth, 15(1), 2216–

2237. doi:10.1080/17538947.2022.2126016.