Coastal shipping, while less documented than oceanic commerce, played a crucial role in European economic development and the British Industrial Revolution. The lack of physical infrastructure like roads and canals, combined with limited cartographic records of sea routes, has hindered our understanding of historical ship courses. However, shipping routes weren't random; they were shaped by climate, oceanography, technology, and economic geography. This study proposes a dynamic, multi-criteria simulation to model historical routing decisions and reveal the likely outlines of European shipping corridors during the age of sail. The model, applied to French and British ports, generates 12 monthly routing predictions for historical shipping corridors, tested against historical evidence. The research addresses a significant gap in our knowledge of pre-industrial maritime trade, providing much needed quantitative data on coastal shipping routes and journey times. This quantitative data is crucial for a more detailed understanding of the economic and social impact of coastal accessibility on early modern and pre-industrial societies.

Existing research using GIS and network analysis to model mobility often focuses on terrestrial movements, such as migration patterns, ancient road networks, and settlement locations. Least-cost path (LCP) analysis, a prominent method, has been applied to various historical and prehistorical contexts, from Roman roads to hominin dispersal. However, modeling historical sea routes is a relatively recent development. While archaeologists have made some contributions studying specific periods and regions (e.g., Viking seafaring, ancient Mediterranean routes, Northwest Pacific sailing), historians have been slower to adopt these computational techniques. This lag is partly due to the abundance of data available for early modern and modern periods, as well as a skills gap among historians in computational methods. Previous LCP modeling of sea routes used simplistic wind models and bathymetry data. More recent studies have incorporated anisotropic modeling (accounting for directional sailing conditions) and seasonal variables based on wind conditions. This study builds on these advancements, offering a more complex and realistic approach.

The model simulates rational routing behavior of an 18th-century pilot, based on: (i) thermodynamic principles of sailing adjusted for historical parameters; and (ii) long-term historical environmental time series for wind, waves, currents, visibility, and extreme weather. While 18th-century sailors lacked modern forecasting and GPS, the authors hypothesize that accumulated knowledge, experience, and early navigational charts provided a basis for informed routing decisions. The model incorporates several variables: bathymetry, coastal visibility, wind speed and direction, wind variability, frequency of extreme weather, wave height, direction and period, and current speed and direction. Climatic data were obtained from global meteorological datasets and normalized to hourly averages from 1950-1978 (except currents, 1992-2020), with the assumption that 20th-century average weather patterns can be applied to earlier periods. The study considered the limitations of using modern data to model the 'Little Ice Age', but argued that the variability of the North Atlantic Oscillation in the 20th century justifies this approach. The model initially used a stochastic approach (10,000 iterations for each port pair), but this was abandoned due to a theoretical flaw in ignoring the spatial co-dependency of environmental variables. A deterministic approach was adopted instead, calculating the optimal path for each direction, month, and historical port pair (24 routes per pair). Four routing models with varying complexity were created, adding variables sequentially. Model 1 used only bathymetry, visibility and wind data; Model 2 added wave data; Model 3 included currents; and Model 4 added wind variability and gusts. The model outputs were validated against historical logbooks (CLIWOC) and port books data, focusing on seasonal variations in route density and journey times. A modern navigation routing engine (QtVlm) was also used for a sanity check comparison. The methodology included detailed descriptions of data sources, parameters, and calculations for each variable, including ship characteristics, sailing speeds, historical port locations, visibility calculations (MPLVA), bathymetry, wind data, wave data, currents, and extreme weather events. Specific formulas and equations are used to estimate variables such as the maximum potential land-sight visibility area, wave lengths in both shallow and deep waters, and other important factors affecting sailing speed and navigational safety.

Model 1, using only wind data, showed a linear routing pattern due to the data's resolution. Adding wave data in Model 2 had a negligible impact on routing but increased journey times. Model 3, incorporating currents, produced more varied and less linear routes and reduced seasonal variation in route length, increasing overall lengths by 20%. Model 4, including wind variability and gusts, reinstated seasonal variation while adding more spatial complexity and further increasing route lengths. The effects of model complexity were not evenly distributed; Mediterranean ports were negatively impacted, while East Coast British ports benefited. Validation against CLIWOC data revealed similarities in seasonal routing patterns: an autumn shift south toward Morocco, summer concentration south-west of La Coruña, a summer shift toward the Portuguese coast, and spring/summer routes farther from the English East coast. Comparison with port book data showed consistency in monthly-to-annual journey time ratios. A sanity check against QtVlm results demonstrated similarities between simulated and modern routes, though the simulated routes exhibited greater complexity for longer distances. Analysis of CLIWOC data during wartime and peacetime revealed reduced traffic in the French Atlantic and Channel during wartime. The study empirically determined that the MPLVA + 12h outline best fit observed CLIWOC routes, indicating that ships generally maintained landsight at least once every 24 hours. Analysis of visibility data highlighted particularly hazardous areas, such as the East Anglian coast, where reduced visibility increased the risk of collisions.

The findings demonstrate the significant impact of seasonal and climatic factors on historical European shipping routes. The multi-criteria modeling approach provides a more accurate and dynamic representation of maritime corridors than previous models. The study's results highlight the importance of including currents, waves, and wind variability to accurately capture routing and journey times, particularly for heavily trafficked ports. The next steps involve refining data on coastal accessibility and freight costs to improve the accuracy of historical routing models, allowing calculations of precise journey times and costs for goods between points in France and Britain. This will enable new analyses of early industrial age trade routes and the impact of port connectivity on urban development. The limitations in using non-contemporaneous climate data and the inability to capture all the contingencies of historical navigation (e.g., weather events, breakdowns) must be kept in mind when interpreting the results; the findings offer an overview of the best-case scenario routing.

This study introduces a novel simulation method for modeling historical European coastal shipping routes during the age of sail. The results highlight the significance of various environmental factors and seasonal changes in shaping these routes. The model successfully identifies historical shipping corridors and provides quantitative journey-time metrics. Further refinement of this approach, integrating additional data on trade and cost variations, will lead to a more detailed understanding of the early modern European economy and transportation networks. Future research should focus on expanding the geographical scope and incorporating additional historical data sources to further enhance the model's accuracy and granularity.

The primary limitation is the reliance on non-contemporaneous 20th-century meteorological data to model earlier periods. While the study addresses this, the assumption of similar weather patterns across centuries remains a potential source of uncertainty. Additionally, the model simplifies complex historical factors such as individual ship capabilities, navigational decisions influenced by extreme weather, breakdowns, port delays, and piracy; therefore the model output should be considered a best-case scenario. The lack of detailed tidal current data might affect the accuracy of the model in specific areas.